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/errno.h>
20 #include <linux/module.h>
21 #include <linux/types.h>
22 #include <linux/socket.h>
23 #include <linux/kernel.h>
24 #include <linux/timer.h>
25 #include <linux/string.h>
26 #include <linux/sockios.h>
27 #include <linux/net.h>
28 #include <linux/fcntl.h>
30 #include <linux/slab.h>
31 #include <linux/interrupt.h>
32 #include <linux/capability.h>
33 #include <linux/skbuff.h>
34 #include <linux/init.h>
35 #include <linux/security.h>
36 #include <linux/mutex.h>
37 #include <linux/if_addr.h>
38 #include <linux/if_bridge.h>
39 #include <linux/if_vlan.h>
40 #include <linux/pci.h>
41 #include <linux/etherdevice.h>
43 #include <asm/uaccess.h>
45 #include <linux/inet.h>
46 #include <linux/netdevice.h>
47 #include <net/switchdev.h>
49 #include <net/protocol.h>
51 #include <net/route.h>
55 #include <net/pkt_sched.h>
56 #include <net/fib_rules.h>
57 #include <net/rtnetlink.h>
58 #include <net/net_namespace.h>
62 rtnl_dumpit_func dumpit;
63 rtnl_calcit_func calcit;
66 static DEFINE_MUTEX(rtnl_mutex);
70 mutex_lock(&rtnl_mutex);
72 EXPORT_SYMBOL(rtnl_lock);
74 static struct sk_buff *defer_kfree_skb_list;
75 void rtnl_kfree_skbs(struct sk_buff *head, struct sk_buff *tail)
78 tail->next = defer_kfree_skb_list;
79 defer_kfree_skb_list = head;
82 EXPORT_SYMBOL(rtnl_kfree_skbs);
84 void __rtnl_unlock(void)
86 struct sk_buff *head = defer_kfree_skb_list;
88 defer_kfree_skb_list = NULL;
90 mutex_unlock(&rtnl_mutex);
93 struct sk_buff *next = head->next;
101 void rtnl_unlock(void)
103 /* This fellow will unlock it for us. */
106 EXPORT_SYMBOL(rtnl_unlock);
108 int rtnl_trylock(void)
110 return mutex_trylock(&rtnl_mutex);
112 EXPORT_SYMBOL(rtnl_trylock);
114 int rtnl_is_locked(void)
116 return mutex_is_locked(&rtnl_mutex);
118 EXPORT_SYMBOL(rtnl_is_locked);
120 #ifdef CONFIG_PROVE_LOCKING
121 bool lockdep_rtnl_is_held(void)
123 return lockdep_is_held(&rtnl_mutex);
125 EXPORT_SYMBOL(lockdep_rtnl_is_held);
126 #endif /* #ifdef CONFIG_PROVE_LOCKING */
128 static struct rtnl_link *rtnl_msg_handlers[RTNL_FAMILY_MAX + 1];
130 static inline int rtm_msgindex(int msgtype)
132 int msgindex = msgtype - RTM_BASE;
135 * msgindex < 0 implies someone tried to register a netlink
136 * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
137 * the message type has not been added to linux/rtnetlink.h
139 BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
144 static rtnl_doit_func rtnl_get_doit(int protocol, int msgindex)
146 struct rtnl_link *tab;
148 if (protocol <= RTNL_FAMILY_MAX)
149 tab = rtnl_msg_handlers[protocol];
153 if (tab == NULL || tab[msgindex].doit == NULL)
154 tab = rtnl_msg_handlers[PF_UNSPEC];
156 return tab[msgindex].doit;
159 static rtnl_dumpit_func rtnl_get_dumpit(int protocol, int msgindex)
161 struct rtnl_link *tab;
163 if (protocol <= RTNL_FAMILY_MAX)
164 tab = rtnl_msg_handlers[protocol];
168 if (tab == NULL || tab[msgindex].dumpit == NULL)
169 tab = rtnl_msg_handlers[PF_UNSPEC];
171 return tab[msgindex].dumpit;
174 static rtnl_calcit_func rtnl_get_calcit(int protocol, int msgindex)
176 struct rtnl_link *tab;
178 if (protocol <= RTNL_FAMILY_MAX)
179 tab = rtnl_msg_handlers[protocol];
183 if (tab == NULL || tab[msgindex].calcit == NULL)
184 tab = rtnl_msg_handlers[PF_UNSPEC];
186 return tab[msgindex].calcit;
190 * __rtnl_register - Register a rtnetlink message type
191 * @protocol: Protocol family or PF_UNSPEC
192 * @msgtype: rtnetlink message type
193 * @doit: Function pointer called for each request message
194 * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
195 * @calcit: Function pointer to calc size of dump message
197 * Registers the specified function pointers (at least one of them has
198 * to be non-NULL) to be called whenever a request message for the
199 * specified protocol family and message type is received.
201 * The special protocol family PF_UNSPEC may be used to define fallback
202 * function pointers for the case when no entry for the specific protocol
205 * Returns 0 on success or a negative error code.
207 int __rtnl_register(int protocol, int msgtype,
208 rtnl_doit_func doit, rtnl_dumpit_func dumpit,
209 rtnl_calcit_func calcit)
211 struct rtnl_link *tab;
214 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
215 msgindex = rtm_msgindex(msgtype);
217 tab = rtnl_msg_handlers[protocol];
219 tab = kcalloc(RTM_NR_MSGTYPES, sizeof(*tab), GFP_KERNEL);
223 rtnl_msg_handlers[protocol] = tab;
227 tab[msgindex].doit = doit;
230 tab[msgindex].dumpit = dumpit;
233 tab[msgindex].calcit = calcit;
237 EXPORT_SYMBOL_GPL(__rtnl_register);
240 * rtnl_register - Register a rtnetlink message type
242 * Identical to __rtnl_register() but panics on failure. This is useful
243 * as failure of this function is very unlikely, it can only happen due
244 * to lack of memory when allocating the chain to store all message
245 * handlers for a protocol. Meant for use in init functions where lack
246 * of memory implies no sense in continuing.
248 void rtnl_register(int protocol, int msgtype,
249 rtnl_doit_func doit, rtnl_dumpit_func dumpit,
250 rtnl_calcit_func calcit)
252 if (__rtnl_register(protocol, msgtype, doit, dumpit, calcit) < 0)
253 panic("Unable to register rtnetlink message handler, "
254 "protocol = %d, message type = %d\n",
257 EXPORT_SYMBOL_GPL(rtnl_register);
260 * rtnl_unregister - Unregister a rtnetlink message type
261 * @protocol: Protocol family or PF_UNSPEC
262 * @msgtype: rtnetlink message type
264 * Returns 0 on success or a negative error code.
266 int rtnl_unregister(int protocol, int msgtype)
270 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
271 msgindex = rtm_msgindex(msgtype);
273 if (rtnl_msg_handlers[protocol] == NULL)
276 rtnl_msg_handlers[protocol][msgindex].doit = NULL;
277 rtnl_msg_handlers[protocol][msgindex].dumpit = NULL;
278 rtnl_msg_handlers[protocol][msgindex].calcit = NULL;
282 EXPORT_SYMBOL_GPL(rtnl_unregister);
285 * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
286 * @protocol : Protocol family or PF_UNSPEC
288 * Identical to calling rtnl_unregster() for all registered message types
289 * of a certain protocol family.
291 void rtnl_unregister_all(int protocol)
293 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
295 kfree(rtnl_msg_handlers[protocol]);
296 rtnl_msg_handlers[protocol] = NULL;
298 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
300 static LIST_HEAD(link_ops);
302 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
304 const struct rtnl_link_ops *ops;
306 list_for_each_entry(ops, &link_ops, list) {
307 if (!strcmp(ops->kind, kind))
314 * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
315 * @ops: struct rtnl_link_ops * to register
317 * The caller must hold the rtnl_mutex. This function should be used
318 * by drivers that create devices during module initialization. It
319 * must be called before registering the devices.
321 * Returns 0 on success or a negative error code.
323 int __rtnl_link_register(struct rtnl_link_ops *ops)
325 if (rtnl_link_ops_get(ops->kind))
328 /* The check for setup is here because if ops
329 * does not have that filled up, it is not possible
330 * to use the ops for creating device. So do not
331 * fill up dellink as well. That disables rtnl_dellink.
333 if (ops->setup && !ops->dellink)
334 ops->dellink = unregister_netdevice_queue;
336 list_add_tail(&ops->list, &link_ops);
339 EXPORT_SYMBOL_GPL(__rtnl_link_register);
342 * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
343 * @ops: struct rtnl_link_ops * to register
345 * Returns 0 on success or a negative error code.
347 int rtnl_link_register(struct rtnl_link_ops *ops)
352 err = __rtnl_link_register(ops);
356 EXPORT_SYMBOL_GPL(rtnl_link_register);
358 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
360 struct net_device *dev;
361 LIST_HEAD(list_kill);
363 for_each_netdev(net, dev) {
364 if (dev->rtnl_link_ops == ops)
365 ops->dellink(dev, &list_kill);
367 unregister_netdevice_many(&list_kill);
371 * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
372 * @ops: struct rtnl_link_ops * to unregister
374 * The caller must hold the rtnl_mutex.
376 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
381 __rtnl_kill_links(net, ops);
383 list_del(&ops->list);
385 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
387 /* Return with the rtnl_lock held when there are no network
388 * devices unregistering in any network namespace.
390 static void rtnl_lock_unregistering_all(void)
394 DEFINE_WAIT_FUNC(wait, woken_wake_function);
396 add_wait_queue(&netdev_unregistering_wq, &wait);
398 unregistering = false;
401 if (net->dev_unreg_count > 0) {
402 unregistering = true;
410 wait_woken(&wait, TASK_UNINTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
412 remove_wait_queue(&netdev_unregistering_wq, &wait);
416 * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
417 * @ops: struct rtnl_link_ops * to unregister
419 void rtnl_link_unregister(struct rtnl_link_ops *ops)
421 /* Close the race with cleanup_net() */
422 mutex_lock(&net_mutex);
423 rtnl_lock_unregistering_all();
424 __rtnl_link_unregister(ops);
426 mutex_unlock(&net_mutex);
428 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
430 static size_t rtnl_link_get_slave_info_data_size(const struct net_device *dev)
432 struct net_device *master_dev;
433 const struct rtnl_link_ops *ops;
435 master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
438 ops = master_dev->rtnl_link_ops;
439 if (!ops || !ops->get_slave_size)
441 /* IFLA_INFO_SLAVE_DATA + nested data */
442 return nla_total_size(sizeof(struct nlattr)) +
443 ops->get_slave_size(master_dev, dev);
446 static size_t rtnl_link_get_size(const struct net_device *dev)
448 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
454 size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
455 nla_total_size(strlen(ops->kind) + 1); /* IFLA_INFO_KIND */
458 /* IFLA_INFO_DATA + nested data */
459 size += nla_total_size(sizeof(struct nlattr)) +
462 if (ops->get_xstats_size)
463 /* IFLA_INFO_XSTATS */
464 size += nla_total_size(ops->get_xstats_size(dev));
466 size += rtnl_link_get_slave_info_data_size(dev);
471 static LIST_HEAD(rtnl_af_ops);
473 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
475 const struct rtnl_af_ops *ops;
477 list_for_each_entry(ops, &rtnl_af_ops, list) {
478 if (ops->family == family)
486 * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
487 * @ops: struct rtnl_af_ops * to register
489 * Returns 0 on success or a negative error code.
491 void rtnl_af_register(struct rtnl_af_ops *ops)
494 list_add_tail(&ops->list, &rtnl_af_ops);
497 EXPORT_SYMBOL_GPL(rtnl_af_register);
500 * __rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
501 * @ops: struct rtnl_af_ops * to unregister
503 * The caller must hold the rtnl_mutex.
505 void __rtnl_af_unregister(struct rtnl_af_ops *ops)
507 list_del(&ops->list);
509 EXPORT_SYMBOL_GPL(__rtnl_af_unregister);
512 * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
513 * @ops: struct rtnl_af_ops * to unregister
515 void rtnl_af_unregister(struct rtnl_af_ops *ops)
518 __rtnl_af_unregister(ops);
521 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
523 static size_t rtnl_link_get_af_size(const struct net_device *dev,
526 struct rtnl_af_ops *af_ops;
530 size = nla_total_size(sizeof(struct nlattr));
532 list_for_each_entry(af_ops, &rtnl_af_ops, list) {
533 if (af_ops->get_link_af_size) {
534 /* AF_* + nested data */
535 size += nla_total_size(sizeof(struct nlattr)) +
536 af_ops->get_link_af_size(dev, ext_filter_mask);
543 static bool rtnl_have_link_slave_info(const struct net_device *dev)
545 struct net_device *master_dev;
547 master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
548 if (master_dev && master_dev->rtnl_link_ops)
553 static int rtnl_link_slave_info_fill(struct sk_buff *skb,
554 const struct net_device *dev)
556 struct net_device *master_dev;
557 const struct rtnl_link_ops *ops;
558 struct nlattr *slave_data;
561 master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
564 ops = master_dev->rtnl_link_ops;
567 if (nla_put_string(skb, IFLA_INFO_SLAVE_KIND, ops->kind) < 0)
569 if (ops->fill_slave_info) {
570 slave_data = nla_nest_start(skb, IFLA_INFO_SLAVE_DATA);
573 err = ops->fill_slave_info(skb, master_dev, dev);
575 goto err_cancel_slave_data;
576 nla_nest_end(skb, slave_data);
580 err_cancel_slave_data:
581 nla_nest_cancel(skb, slave_data);
585 static int rtnl_link_info_fill(struct sk_buff *skb,
586 const struct net_device *dev)
588 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
594 if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
596 if (ops->fill_xstats) {
597 err = ops->fill_xstats(skb, dev);
601 if (ops->fill_info) {
602 data = nla_nest_start(skb, IFLA_INFO_DATA);
605 err = ops->fill_info(skb, dev);
607 goto err_cancel_data;
608 nla_nest_end(skb, data);
613 nla_nest_cancel(skb, data);
617 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
619 struct nlattr *linkinfo;
622 linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
623 if (linkinfo == NULL)
626 err = rtnl_link_info_fill(skb, dev);
628 goto err_cancel_link;
630 err = rtnl_link_slave_info_fill(skb, dev);
632 goto err_cancel_link;
634 nla_nest_end(skb, linkinfo);
638 nla_nest_cancel(skb, linkinfo);
643 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
645 struct sock *rtnl = net->rtnl;
648 NETLINK_CB(skb).dst_group = group;
650 atomic_inc(&skb->users);
651 netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
653 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
657 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
659 struct sock *rtnl = net->rtnl;
661 return nlmsg_unicast(rtnl, skb, pid);
663 EXPORT_SYMBOL(rtnl_unicast);
665 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
666 struct nlmsghdr *nlh, gfp_t flags)
668 struct sock *rtnl = net->rtnl;
672 report = nlmsg_report(nlh);
674 nlmsg_notify(rtnl, skb, pid, group, report, flags);
676 EXPORT_SYMBOL(rtnl_notify);
678 void rtnl_set_sk_err(struct net *net, u32 group, int error)
680 struct sock *rtnl = net->rtnl;
682 netlink_set_err(rtnl, 0, group, error);
684 EXPORT_SYMBOL(rtnl_set_sk_err);
686 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
691 mx = nla_nest_start(skb, RTA_METRICS);
695 for (i = 0; i < RTAX_MAX; i++) {
697 if (i == RTAX_CC_ALGO - 1) {
698 char tmp[TCP_CA_NAME_MAX], *name;
700 name = tcp_ca_get_name_by_key(metrics[i], tmp);
703 if (nla_put_string(skb, i + 1, name))
704 goto nla_put_failure;
705 } else if (i == RTAX_FEATURES - 1) {
706 u32 user_features = metrics[i] & RTAX_FEATURE_MASK;
710 BUILD_BUG_ON(RTAX_FEATURE_MASK & DST_FEATURE_MASK);
711 if (nla_put_u32(skb, i + 1, user_features))
712 goto nla_put_failure;
714 if (nla_put_u32(skb, i + 1, metrics[i]))
715 goto nla_put_failure;
722 nla_nest_cancel(skb, mx);
726 return nla_nest_end(skb, mx);
729 nla_nest_cancel(skb, mx);
732 EXPORT_SYMBOL(rtnetlink_put_metrics);
734 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
735 long expires, u32 error)
737 struct rta_cacheinfo ci = {
738 .rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse),
739 .rta_used = dst->__use,
740 .rta_clntref = atomic_read(&(dst->__refcnt)),
748 clock = jiffies_to_clock_t(abs(expires));
749 clock = min_t(unsigned long, clock, INT_MAX);
750 ci.rta_expires = (expires > 0) ? clock : -clock;
752 return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
754 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
756 static void set_operstate(struct net_device *dev, unsigned char transition)
758 unsigned char operstate = dev->operstate;
760 switch (transition) {
762 if ((operstate == IF_OPER_DORMANT ||
763 operstate == IF_OPER_UNKNOWN) &&
765 operstate = IF_OPER_UP;
768 case IF_OPER_DORMANT:
769 if (operstate == IF_OPER_UP ||
770 operstate == IF_OPER_UNKNOWN)
771 operstate = IF_OPER_DORMANT;
775 if (dev->operstate != operstate) {
776 write_lock_bh(&dev_base_lock);
777 dev->operstate = operstate;
778 write_unlock_bh(&dev_base_lock);
779 netdev_state_change(dev);
783 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
785 return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
786 (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
789 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
790 const struct ifinfomsg *ifm)
792 unsigned int flags = ifm->ifi_flags;
794 /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
796 flags = (flags & ifm->ifi_change) |
797 (rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
802 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
803 const struct rtnl_link_stats64 *b)
805 a->rx_packets = b->rx_packets;
806 a->tx_packets = b->tx_packets;
807 a->rx_bytes = b->rx_bytes;
808 a->tx_bytes = b->tx_bytes;
809 a->rx_errors = b->rx_errors;
810 a->tx_errors = b->tx_errors;
811 a->rx_dropped = b->rx_dropped;
812 a->tx_dropped = b->tx_dropped;
814 a->multicast = b->multicast;
815 a->collisions = b->collisions;
817 a->rx_length_errors = b->rx_length_errors;
818 a->rx_over_errors = b->rx_over_errors;
819 a->rx_crc_errors = b->rx_crc_errors;
820 a->rx_frame_errors = b->rx_frame_errors;
821 a->rx_fifo_errors = b->rx_fifo_errors;
822 a->rx_missed_errors = b->rx_missed_errors;
824 a->tx_aborted_errors = b->tx_aborted_errors;
825 a->tx_carrier_errors = b->tx_carrier_errors;
826 a->tx_fifo_errors = b->tx_fifo_errors;
827 a->tx_heartbeat_errors = b->tx_heartbeat_errors;
828 a->tx_window_errors = b->tx_window_errors;
830 a->rx_compressed = b->rx_compressed;
831 a->tx_compressed = b->tx_compressed;
833 a->rx_nohandler = b->rx_nohandler;
837 static inline int rtnl_vfinfo_size(const struct net_device *dev,
840 if (dev->dev.parent && dev_is_pci(dev->dev.parent) &&
841 (ext_filter_mask & RTEXT_FILTER_VF)) {
842 int num_vfs = dev_num_vf(dev->dev.parent);
843 size_t size = nla_total_size(0);
846 nla_total_size(sizeof(struct ifla_vf_mac)) +
847 nla_total_size(sizeof(struct ifla_vf_vlan)) +
848 nla_total_size(0) + /* nest IFLA_VF_VLAN_LIST */
849 nla_total_size(MAX_VLAN_LIST_LEN *
850 sizeof(struct ifla_vf_vlan_info)) +
851 nla_total_size(sizeof(struct ifla_vf_spoofchk)) +
852 nla_total_size(sizeof(struct ifla_vf_tx_rate)) +
853 nla_total_size(sizeof(struct ifla_vf_rate)) +
854 nla_total_size(sizeof(struct ifla_vf_link_state)) +
855 nla_total_size(sizeof(struct ifla_vf_rss_query_en)) +
856 nla_total_size(0) + /* nest IFLA_VF_STATS */
857 /* IFLA_VF_STATS_RX_PACKETS */
858 nla_total_size_64bit(sizeof(__u64)) +
859 /* IFLA_VF_STATS_TX_PACKETS */
860 nla_total_size_64bit(sizeof(__u64)) +
861 /* IFLA_VF_STATS_RX_BYTES */
862 nla_total_size_64bit(sizeof(__u64)) +
863 /* IFLA_VF_STATS_TX_BYTES */
864 nla_total_size_64bit(sizeof(__u64)) +
865 /* IFLA_VF_STATS_BROADCAST */
866 nla_total_size_64bit(sizeof(__u64)) +
867 /* IFLA_VF_STATS_MULTICAST */
868 nla_total_size_64bit(sizeof(__u64)) +
869 nla_total_size(sizeof(struct ifla_vf_trust)));
875 static size_t rtnl_port_size(const struct net_device *dev,
878 size_t port_size = nla_total_size(4) /* PORT_VF */
879 + nla_total_size(PORT_PROFILE_MAX) /* PORT_PROFILE */
880 + nla_total_size(sizeof(struct ifla_port_vsi))
882 + nla_total_size(PORT_UUID_MAX) /* PORT_INSTANCE_UUID */
883 + nla_total_size(PORT_UUID_MAX) /* PORT_HOST_UUID */
884 + nla_total_size(1) /* PROT_VDP_REQUEST */
885 + nla_total_size(2); /* PORT_VDP_RESPONSE */
886 size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
887 size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
889 size_t port_self_size = nla_total_size(sizeof(struct nlattr))
892 if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
893 !(ext_filter_mask & RTEXT_FILTER_VF))
895 if (dev_num_vf(dev->dev.parent))
896 return port_self_size + vf_ports_size +
897 vf_port_size * dev_num_vf(dev->dev.parent);
899 return port_self_size;
902 static size_t rtnl_xdp_size(const struct net_device *dev)
904 size_t xdp_size = nla_total_size(0) + /* nest IFLA_XDP */
905 nla_total_size(1); /* XDP_ATTACHED */
907 if (!dev->netdev_ops->ndo_xdp)
913 static noinline size_t if_nlmsg_size(const struct net_device *dev,
916 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
917 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
918 + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
919 + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
920 + nla_total_size_64bit(sizeof(struct rtnl_link_ifmap))
921 + nla_total_size(sizeof(struct rtnl_link_stats))
922 + nla_total_size_64bit(sizeof(struct rtnl_link_stats64))
923 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
924 + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
925 + nla_total_size(4) /* IFLA_TXQLEN */
926 + nla_total_size(4) /* IFLA_WEIGHT */
927 + nla_total_size(4) /* IFLA_MTU */
928 + nla_total_size(4) /* IFLA_LINK */
929 + nla_total_size(4) /* IFLA_MASTER */
930 + nla_total_size(1) /* IFLA_CARRIER */
931 + nla_total_size(4) /* IFLA_PROMISCUITY */
932 + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
933 + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
934 + nla_total_size(4) /* IFLA_GSO_MAX_SEGS */
935 + nla_total_size(4) /* IFLA_GSO_MAX_SIZE */
936 + nla_total_size(1) /* IFLA_OPERSTATE */
937 + nla_total_size(1) /* IFLA_LINKMODE */
938 + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
939 + nla_total_size(4) /* IFLA_LINK_NETNSID */
940 + nla_total_size(4) /* IFLA_GROUP */
941 + nla_total_size(ext_filter_mask
942 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
943 + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
944 + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
945 + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
946 + rtnl_link_get_af_size(dev, ext_filter_mask) /* IFLA_AF_SPEC */
947 + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */
948 + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_SWITCH_ID */
949 + nla_total_size(IFNAMSIZ) /* IFLA_PHYS_PORT_NAME */
950 + rtnl_xdp_size(dev) /* IFLA_XDP */
951 + nla_total_size(1); /* IFLA_PROTO_DOWN */
955 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
957 struct nlattr *vf_ports;
958 struct nlattr *vf_port;
962 vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
966 for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
967 vf_port = nla_nest_start(skb, IFLA_VF_PORT);
969 goto nla_put_failure;
970 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
971 goto nla_put_failure;
972 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
973 if (err == -EMSGSIZE)
974 goto nla_put_failure;
976 nla_nest_cancel(skb, vf_port);
979 nla_nest_end(skb, vf_port);
982 nla_nest_end(skb, vf_ports);
987 nla_nest_cancel(skb, vf_ports);
991 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
993 struct nlattr *port_self;
996 port_self = nla_nest_start(skb, IFLA_PORT_SELF);
1000 err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
1002 nla_nest_cancel(skb, port_self);
1003 return (err == -EMSGSIZE) ? err : 0;
1006 nla_nest_end(skb, port_self);
1011 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
1012 u32 ext_filter_mask)
1016 if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
1017 !(ext_filter_mask & RTEXT_FILTER_VF))
1020 err = rtnl_port_self_fill(skb, dev);
1024 if (dev_num_vf(dev->dev.parent)) {
1025 err = rtnl_vf_ports_fill(skb, dev);
1033 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
1036 struct netdev_phys_item_id ppid;
1038 err = dev_get_phys_port_id(dev, &ppid);
1040 if (err == -EOPNOTSUPP)
1045 if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
1051 static int rtnl_phys_port_name_fill(struct sk_buff *skb, struct net_device *dev)
1053 char name[IFNAMSIZ];
1056 err = dev_get_phys_port_name(dev, name, sizeof(name));
1058 if (err == -EOPNOTSUPP)
1063 if (nla_put_string(skb, IFLA_PHYS_PORT_NAME, name))
1069 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev)
1072 struct switchdev_attr attr = {
1074 .id = SWITCHDEV_ATTR_ID_PORT_PARENT_ID,
1075 .flags = SWITCHDEV_F_NO_RECURSE,
1078 err = switchdev_port_attr_get(dev, &attr);
1080 if (err == -EOPNOTSUPP)
1085 if (nla_put(skb, IFLA_PHYS_SWITCH_ID, attr.u.ppid.id_len,
1092 static noinline_for_stack int rtnl_fill_stats(struct sk_buff *skb,
1093 struct net_device *dev)
1095 struct rtnl_link_stats64 *sp;
1096 struct nlattr *attr;
1098 attr = nla_reserve_64bit(skb, IFLA_STATS64,
1099 sizeof(struct rtnl_link_stats64), IFLA_PAD);
1103 sp = nla_data(attr);
1104 dev_get_stats(dev, sp);
1106 attr = nla_reserve(skb, IFLA_STATS,
1107 sizeof(struct rtnl_link_stats));
1111 copy_rtnl_link_stats(nla_data(attr), sp);
1116 static noinline_for_stack int rtnl_fill_vfinfo(struct sk_buff *skb,
1117 struct net_device *dev,
1119 struct nlattr *vfinfo)
1121 struct ifla_vf_rss_query_en vf_rss_query_en;
1122 struct nlattr *vf, *vfstats, *vfvlanlist;
1123 struct ifla_vf_link_state vf_linkstate;
1124 struct ifla_vf_vlan_info vf_vlan_info;
1125 struct ifla_vf_spoofchk vf_spoofchk;
1126 struct ifla_vf_tx_rate vf_tx_rate;
1127 struct ifla_vf_stats vf_stats;
1128 struct ifla_vf_trust vf_trust;
1129 struct ifla_vf_vlan vf_vlan;
1130 struct ifla_vf_rate vf_rate;
1131 struct ifla_vf_mac vf_mac;
1132 struct ifla_vf_info ivi;
1134 memset(&ivi, 0, sizeof(ivi));
1136 /* Not all SR-IOV capable drivers support the
1137 * spoofcheck and "RSS query enable" query. Preset to
1138 * -1 so the user space tool can detect that the driver
1139 * didn't report anything.
1142 ivi.rss_query_en = -1;
1144 /* The default value for VF link state is "auto"
1145 * IFLA_VF_LINK_STATE_AUTO which equals zero
1148 /* VLAN Protocol by default is 802.1Q */
1149 ivi.vlan_proto = htons(ETH_P_8021Q);
1150 if (dev->netdev_ops->ndo_get_vf_config(dev, vfs_num, &ivi))
1153 memset(&vf_vlan_info, 0, sizeof(vf_vlan_info));
1162 vf_rss_query_en.vf =
1163 vf_trust.vf = ivi.vf;
1165 memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1166 vf_vlan.vlan = ivi.vlan;
1167 vf_vlan.qos = ivi.qos;
1168 vf_vlan_info.vlan = ivi.vlan;
1169 vf_vlan_info.qos = ivi.qos;
1170 vf_vlan_info.vlan_proto = ivi.vlan_proto;
1171 vf_tx_rate.rate = ivi.max_tx_rate;
1172 vf_rate.min_tx_rate = ivi.min_tx_rate;
1173 vf_rate.max_tx_rate = ivi.max_tx_rate;
1174 vf_spoofchk.setting = ivi.spoofchk;
1175 vf_linkstate.link_state = ivi.linkstate;
1176 vf_rss_query_en.setting = ivi.rss_query_en;
1177 vf_trust.setting = ivi.trusted;
1178 vf = nla_nest_start(skb, IFLA_VF_INFO);
1180 goto nla_put_vfinfo_failure;
1181 if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1182 nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1183 nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1185 nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1187 nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1189 nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1191 nla_put(skb, IFLA_VF_RSS_QUERY_EN,
1192 sizeof(vf_rss_query_en),
1193 &vf_rss_query_en) ||
1194 nla_put(skb, IFLA_VF_TRUST,
1195 sizeof(vf_trust), &vf_trust))
1196 goto nla_put_vf_failure;
1197 vfvlanlist = nla_nest_start(skb, IFLA_VF_VLAN_LIST);
1199 goto nla_put_vf_failure;
1200 if (nla_put(skb, IFLA_VF_VLAN_INFO, sizeof(vf_vlan_info),
1202 nla_nest_cancel(skb, vfvlanlist);
1203 goto nla_put_vf_failure;
1205 nla_nest_end(skb, vfvlanlist);
1206 memset(&vf_stats, 0, sizeof(vf_stats));
1207 if (dev->netdev_ops->ndo_get_vf_stats)
1208 dev->netdev_ops->ndo_get_vf_stats(dev, vfs_num,
1210 vfstats = nla_nest_start(skb, IFLA_VF_STATS);
1212 goto nla_put_vf_failure;
1213 if (nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_PACKETS,
1214 vf_stats.rx_packets, IFLA_VF_STATS_PAD) ||
1215 nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_PACKETS,
1216 vf_stats.tx_packets, IFLA_VF_STATS_PAD) ||
1217 nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_BYTES,
1218 vf_stats.rx_bytes, IFLA_VF_STATS_PAD) ||
1219 nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_BYTES,
1220 vf_stats.tx_bytes, IFLA_VF_STATS_PAD) ||
1221 nla_put_u64_64bit(skb, IFLA_VF_STATS_BROADCAST,
1222 vf_stats.broadcast, IFLA_VF_STATS_PAD) ||
1223 nla_put_u64_64bit(skb, IFLA_VF_STATS_MULTICAST,
1224 vf_stats.multicast, IFLA_VF_STATS_PAD)) {
1225 nla_nest_cancel(skb, vfstats);
1226 goto nla_put_vf_failure;
1228 nla_nest_end(skb, vfstats);
1229 nla_nest_end(skb, vf);
1233 nla_nest_cancel(skb, vf);
1234 nla_put_vfinfo_failure:
1235 nla_nest_cancel(skb, vfinfo);
1239 static int rtnl_fill_link_ifmap(struct sk_buff *skb, struct net_device *dev)
1241 struct rtnl_link_ifmap map;
1243 memset(&map, 0, sizeof(map));
1244 map.mem_start = dev->mem_start;
1245 map.mem_end = dev->mem_end;
1246 map.base_addr = dev->base_addr;
1249 map.port = dev->if_port;
1251 if (nla_put_64bit(skb, IFLA_MAP, sizeof(map), &map, IFLA_PAD))
1257 static int rtnl_xdp_fill(struct sk_buff *skb, struct net_device *dev)
1259 struct netdev_xdp xdp_op = {};
1263 if (!dev->netdev_ops->ndo_xdp)
1265 xdp = nla_nest_start(skb, IFLA_XDP);
1268 xdp_op.command = XDP_QUERY_PROG;
1269 err = dev->netdev_ops->ndo_xdp(dev, &xdp_op);
1272 err = nla_put_u8(skb, IFLA_XDP_ATTACHED, xdp_op.prog_attached);
1276 nla_nest_end(skb, xdp);
1280 nla_nest_cancel(skb, xdp);
1284 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
1285 int type, u32 pid, u32 seq, u32 change,
1286 unsigned int flags, u32 ext_filter_mask)
1288 struct ifinfomsg *ifm;
1289 struct nlmsghdr *nlh;
1290 struct nlattr *af_spec;
1291 struct rtnl_af_ops *af_ops;
1292 struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1295 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
1299 ifm = nlmsg_data(nlh);
1300 ifm->ifi_family = AF_UNSPEC;
1302 ifm->ifi_type = dev->type;
1303 ifm->ifi_index = dev->ifindex;
1304 ifm->ifi_flags = dev_get_flags(dev);
1305 ifm->ifi_change = change;
1307 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
1308 nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
1309 nla_put_u8(skb, IFLA_OPERSTATE,
1310 netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
1311 nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
1312 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
1313 nla_put_u32(skb, IFLA_GROUP, dev->group) ||
1314 nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
1315 nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
1316 nla_put_u32(skb, IFLA_GSO_MAX_SEGS, dev->gso_max_segs) ||
1317 nla_put_u32(skb, IFLA_GSO_MAX_SIZE, dev->gso_max_size) ||
1319 nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1321 (dev->ifindex != dev_get_iflink(dev) &&
1322 nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
1324 nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex)) ||
1325 nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1327 nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
1329 nla_put_string(skb, IFLA_IFALIAS, dev->ifalias)) ||
1330 nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1331 atomic_read(&dev->carrier_changes)) ||
1332 nla_put_u8(skb, IFLA_PROTO_DOWN, dev->proto_down))
1333 goto nla_put_failure;
1335 if (rtnl_fill_link_ifmap(skb, dev))
1336 goto nla_put_failure;
1338 if (dev->addr_len) {
1339 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1340 nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1341 goto nla_put_failure;
1344 if (rtnl_phys_port_id_fill(skb, dev))
1345 goto nla_put_failure;
1347 if (rtnl_phys_port_name_fill(skb, dev))
1348 goto nla_put_failure;
1350 if (rtnl_phys_switch_id_fill(skb, dev))
1351 goto nla_put_failure;
1353 if (rtnl_fill_stats(skb, dev))
1354 goto nla_put_failure;
1356 if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF) &&
1357 nla_put_u32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent)))
1358 goto nla_put_failure;
1360 if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent &&
1361 ext_filter_mask & RTEXT_FILTER_VF) {
1363 struct nlattr *vfinfo;
1364 int num_vfs = dev_num_vf(dev->dev.parent);
1366 vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
1368 goto nla_put_failure;
1369 for (i = 0; i < num_vfs; i++) {
1370 if (rtnl_fill_vfinfo(skb, dev, i, vfinfo))
1371 goto nla_put_failure;
1374 nla_nest_end(skb, vfinfo);
1377 if (rtnl_port_fill(skb, dev, ext_filter_mask))
1378 goto nla_put_failure;
1380 if (rtnl_xdp_fill(skb, dev))
1381 goto nla_put_failure;
1383 if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1384 if (rtnl_link_fill(skb, dev) < 0)
1385 goto nla_put_failure;
1388 if (dev->rtnl_link_ops &&
1389 dev->rtnl_link_ops->get_link_net) {
1390 struct net *link_net = dev->rtnl_link_ops->get_link_net(dev);
1392 if (!net_eq(dev_net(dev), link_net)) {
1393 int id = peernet2id_alloc(dev_net(dev), link_net);
1395 if (nla_put_s32(skb, IFLA_LINK_NETNSID, id))
1396 goto nla_put_failure;
1400 if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1401 goto nla_put_failure;
1403 list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1404 if (af_ops->fill_link_af) {
1408 if (!(af = nla_nest_start(skb, af_ops->family)))
1409 goto nla_put_failure;
1411 err = af_ops->fill_link_af(skb, dev, ext_filter_mask);
1414 * Caller may return ENODATA to indicate that there
1415 * was no data to be dumped. This is not an error, it
1416 * means we should trim the attribute header and
1419 if (err == -ENODATA)
1420 nla_nest_cancel(skb, af);
1422 goto nla_put_failure;
1424 nla_nest_end(skb, af);
1428 nla_nest_end(skb, af_spec);
1430 nlmsg_end(skb, nlh);
1434 nlmsg_cancel(skb, nlh);
1438 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1439 [IFLA_IFNAME] = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1440 [IFLA_ADDRESS] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1441 [IFLA_BROADCAST] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1442 [IFLA_MAP] = { .len = sizeof(struct rtnl_link_ifmap) },
1443 [IFLA_MTU] = { .type = NLA_U32 },
1444 [IFLA_LINK] = { .type = NLA_U32 },
1445 [IFLA_MASTER] = { .type = NLA_U32 },
1446 [IFLA_CARRIER] = { .type = NLA_U8 },
1447 [IFLA_TXQLEN] = { .type = NLA_U32 },
1448 [IFLA_WEIGHT] = { .type = NLA_U32 },
1449 [IFLA_OPERSTATE] = { .type = NLA_U8 },
1450 [IFLA_LINKMODE] = { .type = NLA_U8 },
1451 [IFLA_LINKINFO] = { .type = NLA_NESTED },
1452 [IFLA_NET_NS_PID] = { .type = NLA_U32 },
1453 [IFLA_NET_NS_FD] = { .type = NLA_U32 },
1454 [IFLA_IFALIAS] = { .type = NLA_STRING, .len = IFALIASZ-1 },
1455 [IFLA_VFINFO_LIST] = {. type = NLA_NESTED },
1456 [IFLA_VF_PORTS] = { .type = NLA_NESTED },
1457 [IFLA_PORT_SELF] = { .type = NLA_NESTED },
1458 [IFLA_AF_SPEC] = { .type = NLA_NESTED },
1459 [IFLA_EXT_MASK] = { .type = NLA_U32 },
1460 [IFLA_PROMISCUITY] = { .type = NLA_U32 },
1461 [IFLA_NUM_TX_QUEUES] = { .type = NLA_U32 },
1462 [IFLA_NUM_RX_QUEUES] = { .type = NLA_U32 },
1463 [IFLA_PHYS_PORT_ID] = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1464 [IFLA_CARRIER_CHANGES] = { .type = NLA_U32 }, /* ignored */
1465 [IFLA_PHYS_SWITCH_ID] = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1466 [IFLA_LINK_NETNSID] = { .type = NLA_S32 },
1467 [IFLA_PROTO_DOWN] = { .type = NLA_U8 },
1468 [IFLA_XDP] = { .type = NLA_NESTED },
1469 [IFLA_GROUP] = { .type = NLA_U32 },
1472 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1473 [IFLA_INFO_KIND] = { .type = NLA_STRING },
1474 [IFLA_INFO_DATA] = { .type = NLA_NESTED },
1475 [IFLA_INFO_SLAVE_KIND] = { .type = NLA_STRING },
1476 [IFLA_INFO_SLAVE_DATA] = { .type = NLA_NESTED },
1479 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1480 [IFLA_VF_MAC] = { .len = sizeof(struct ifla_vf_mac) },
1481 [IFLA_VF_VLAN] = { .len = sizeof(struct ifla_vf_vlan) },
1482 [IFLA_VF_VLAN_LIST] = { .type = NLA_NESTED },
1483 [IFLA_VF_TX_RATE] = { .len = sizeof(struct ifla_vf_tx_rate) },
1484 [IFLA_VF_SPOOFCHK] = { .len = sizeof(struct ifla_vf_spoofchk) },
1485 [IFLA_VF_RATE] = { .len = sizeof(struct ifla_vf_rate) },
1486 [IFLA_VF_LINK_STATE] = { .len = sizeof(struct ifla_vf_link_state) },
1487 [IFLA_VF_RSS_QUERY_EN] = { .len = sizeof(struct ifla_vf_rss_query_en) },
1488 [IFLA_VF_STATS] = { .type = NLA_NESTED },
1489 [IFLA_VF_TRUST] = { .len = sizeof(struct ifla_vf_trust) },
1490 [IFLA_VF_IB_NODE_GUID] = { .len = sizeof(struct ifla_vf_guid) },
1491 [IFLA_VF_IB_PORT_GUID] = { .len = sizeof(struct ifla_vf_guid) },
1494 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1495 [IFLA_PORT_VF] = { .type = NLA_U32 },
1496 [IFLA_PORT_PROFILE] = { .type = NLA_STRING,
1497 .len = PORT_PROFILE_MAX },
1498 [IFLA_PORT_VSI_TYPE] = { .type = NLA_BINARY,
1499 .len = sizeof(struct ifla_port_vsi)},
1500 [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1501 .len = PORT_UUID_MAX },
1502 [IFLA_PORT_HOST_UUID] = { .type = NLA_STRING,
1503 .len = PORT_UUID_MAX },
1504 [IFLA_PORT_REQUEST] = { .type = NLA_U8, },
1505 [IFLA_PORT_RESPONSE] = { .type = NLA_U16, },
1508 static const struct nla_policy ifla_xdp_policy[IFLA_XDP_MAX + 1] = {
1509 [IFLA_XDP_FD] = { .type = NLA_S32 },
1510 [IFLA_XDP_ATTACHED] = { .type = NLA_U8 },
1513 static const struct rtnl_link_ops *linkinfo_to_kind_ops(const struct nlattr *nla)
1515 const struct rtnl_link_ops *ops = NULL;
1516 struct nlattr *linfo[IFLA_INFO_MAX + 1];
1518 if (nla_parse_nested(linfo, IFLA_INFO_MAX, nla, ifla_info_policy) < 0)
1521 if (linfo[IFLA_INFO_KIND]) {
1522 char kind[MODULE_NAME_LEN];
1524 nla_strlcpy(kind, linfo[IFLA_INFO_KIND], sizeof(kind));
1525 ops = rtnl_link_ops_get(kind);
1531 static bool link_master_filtered(struct net_device *dev, int master_idx)
1533 struct net_device *master;
1538 master = netdev_master_upper_dev_get(dev);
1539 if (!master || master->ifindex != master_idx)
1545 static bool link_kind_filtered(const struct net_device *dev,
1546 const struct rtnl_link_ops *kind_ops)
1548 if (kind_ops && dev->rtnl_link_ops != kind_ops)
1554 static bool link_dump_filtered(struct net_device *dev,
1556 const struct rtnl_link_ops *kind_ops)
1558 if (link_master_filtered(dev, master_idx) ||
1559 link_kind_filtered(dev, kind_ops))
1565 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1567 struct net *net = sock_net(skb->sk);
1570 struct net_device *dev;
1571 struct hlist_head *head;
1572 struct nlattr *tb[IFLA_MAX+1];
1573 u32 ext_filter_mask = 0;
1574 const struct rtnl_link_ops *kind_ops = NULL;
1575 unsigned int flags = NLM_F_MULTI;
1581 s_idx = cb->args[1];
1583 cb->seq = net->dev_base_seq;
1585 /* A hack to preserve kernel<->userspace interface.
1586 * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
1587 * However, before Linux v3.9 the code here assumed rtgenmsg and that's
1588 * what iproute2 < v3.9.0 used.
1589 * We can detect the old iproute2. Even including the IFLA_EXT_MASK
1590 * attribute, its netlink message is shorter than struct ifinfomsg.
1592 hdrlen = nlmsg_len(cb->nlh) < sizeof(struct ifinfomsg) ?
1593 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
1595 if (nlmsg_parse(cb->nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
1597 if (tb[IFLA_EXT_MASK])
1598 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1600 if (tb[IFLA_MASTER])
1601 master_idx = nla_get_u32(tb[IFLA_MASTER]);
1603 if (tb[IFLA_LINKINFO])
1604 kind_ops = linkinfo_to_kind_ops(tb[IFLA_LINKINFO]);
1606 if (master_idx || kind_ops)
1607 flags |= NLM_F_DUMP_FILTERED;
1610 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1612 head = &net->dev_index_head[h];
1613 hlist_for_each_entry(dev, head, index_hlist) {
1614 if (link_dump_filtered(dev, master_idx, kind_ops))
1618 err = rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1619 NETLINK_CB(cb->skb).portid,
1620 cb->nlh->nlmsg_seq, 0,
1625 if (likely(skb->len))
1631 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1645 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len)
1647 return nla_parse(tb, IFLA_MAX, head, len, ifla_policy);
1649 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
1651 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1654 /* Examine the link attributes and figure out which
1655 * network namespace we are talking about.
1657 if (tb[IFLA_NET_NS_PID])
1658 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1659 else if (tb[IFLA_NET_NS_FD])
1660 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1662 net = get_net(src_net);
1665 EXPORT_SYMBOL(rtnl_link_get_net);
1667 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1670 if (tb[IFLA_ADDRESS] &&
1671 nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1674 if (tb[IFLA_BROADCAST] &&
1675 nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1679 if (tb[IFLA_AF_SPEC]) {
1683 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1684 const struct rtnl_af_ops *af_ops;
1686 if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1687 return -EAFNOSUPPORT;
1689 if (!af_ops->set_link_af)
1692 if (af_ops->validate_link_af) {
1693 err = af_ops->validate_link_af(dev, af);
1703 static int handle_infiniband_guid(struct net_device *dev, struct ifla_vf_guid *ivt,
1706 const struct net_device_ops *ops = dev->netdev_ops;
1708 return ops->ndo_set_vf_guid(dev, ivt->vf, ivt->guid, guid_type);
1711 static int handle_vf_guid(struct net_device *dev, struct ifla_vf_guid *ivt, int guid_type)
1713 if (dev->type != ARPHRD_INFINIBAND)
1716 return handle_infiniband_guid(dev, ivt, guid_type);
1719 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb)
1721 const struct net_device_ops *ops = dev->netdev_ops;
1724 if (tb[IFLA_VF_MAC]) {
1725 struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]);
1727 if (ivm->vf >= INT_MAX)
1730 if (ops->ndo_set_vf_mac)
1731 err = ops->ndo_set_vf_mac(dev, ivm->vf,
1737 if (tb[IFLA_VF_VLAN]) {
1738 struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]);
1740 if (ivv->vf >= INT_MAX)
1743 if (ops->ndo_set_vf_vlan)
1744 err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan,
1746 htons(ETH_P_8021Q));
1751 if (tb[IFLA_VF_VLAN_LIST]) {
1752 struct ifla_vf_vlan_info *ivvl[MAX_VLAN_LIST_LEN];
1753 struct nlattr *attr;
1757 if (!ops->ndo_set_vf_vlan)
1760 nla_for_each_nested(attr, tb[IFLA_VF_VLAN_LIST], rem) {
1761 if (nla_type(attr) != IFLA_VF_VLAN_INFO ||
1762 nla_len(attr) < NLA_HDRLEN) {
1765 if (len >= MAX_VLAN_LIST_LEN)
1767 ivvl[len] = nla_data(attr);
1774 if (ivvl[0]->vf >= INT_MAX)
1776 err = ops->ndo_set_vf_vlan(dev, ivvl[0]->vf, ivvl[0]->vlan,
1777 ivvl[0]->qos, ivvl[0]->vlan_proto);
1782 if (tb[IFLA_VF_TX_RATE]) {
1783 struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]);
1784 struct ifla_vf_info ivf;
1786 if (ivt->vf >= INT_MAX)
1789 if (ops->ndo_get_vf_config)
1790 err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf);
1795 if (ops->ndo_set_vf_rate)
1796 err = ops->ndo_set_vf_rate(dev, ivt->vf,
1803 if (tb[IFLA_VF_RATE]) {
1804 struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]);
1806 if (ivt->vf >= INT_MAX)
1809 if (ops->ndo_set_vf_rate)
1810 err = ops->ndo_set_vf_rate(dev, ivt->vf,
1817 if (tb[IFLA_VF_SPOOFCHK]) {
1818 struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]);
1820 if (ivs->vf >= INT_MAX)
1823 if (ops->ndo_set_vf_spoofchk)
1824 err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1830 if (tb[IFLA_VF_LINK_STATE]) {
1831 struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]);
1833 if (ivl->vf >= INT_MAX)
1836 if (ops->ndo_set_vf_link_state)
1837 err = ops->ndo_set_vf_link_state(dev, ivl->vf,
1843 if (tb[IFLA_VF_RSS_QUERY_EN]) {
1844 struct ifla_vf_rss_query_en *ivrssq_en;
1847 ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]);
1848 if (ivrssq_en->vf >= INT_MAX)
1850 if (ops->ndo_set_vf_rss_query_en)
1851 err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf,
1852 ivrssq_en->setting);
1857 if (tb[IFLA_VF_TRUST]) {
1858 struct ifla_vf_trust *ivt = nla_data(tb[IFLA_VF_TRUST]);
1860 if (ivt->vf >= INT_MAX)
1863 if (ops->ndo_set_vf_trust)
1864 err = ops->ndo_set_vf_trust(dev, ivt->vf, ivt->setting);
1869 if (tb[IFLA_VF_IB_NODE_GUID]) {
1870 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_NODE_GUID]);
1872 if (ivt->vf >= INT_MAX)
1874 if (!ops->ndo_set_vf_guid)
1876 return handle_vf_guid(dev, ivt, IFLA_VF_IB_NODE_GUID);
1879 if (tb[IFLA_VF_IB_PORT_GUID]) {
1880 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_PORT_GUID]);
1882 if (ivt->vf >= INT_MAX)
1884 if (!ops->ndo_set_vf_guid)
1887 return handle_vf_guid(dev, ivt, IFLA_VF_IB_PORT_GUID);
1893 static int do_set_master(struct net_device *dev, int ifindex)
1895 struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1896 const struct net_device_ops *ops;
1900 if (upper_dev->ifindex == ifindex)
1902 ops = upper_dev->netdev_ops;
1903 if (ops->ndo_del_slave) {
1904 err = ops->ndo_del_slave(upper_dev, dev);
1913 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
1916 ops = upper_dev->netdev_ops;
1917 if (ops->ndo_add_slave) {
1918 err = ops->ndo_add_slave(upper_dev, dev);
1928 #define DO_SETLINK_MODIFIED 0x01
1929 /* notify flag means notify + modified. */
1930 #define DO_SETLINK_NOTIFY 0x03
1931 static int do_setlink(const struct sk_buff *skb,
1932 struct net_device *dev, struct ifinfomsg *ifm,
1933 struct nlattr **tb, char *ifname, int status)
1935 const struct net_device_ops *ops = dev->netdev_ops;
1938 err = validate_linkmsg(dev, tb);
1942 if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1943 struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1948 if (!netlink_ns_capable(skb, net->user_ns, CAP_NET_ADMIN)) {
1953 err = dev_change_net_namespace(dev, net, ifname);
1957 status |= DO_SETLINK_MODIFIED;
1961 struct rtnl_link_ifmap *u_map;
1964 if (!ops->ndo_set_config) {
1969 if (!netif_device_present(dev)) {
1974 u_map = nla_data(tb[IFLA_MAP]);
1975 k_map.mem_start = (unsigned long) u_map->mem_start;
1976 k_map.mem_end = (unsigned long) u_map->mem_end;
1977 k_map.base_addr = (unsigned short) u_map->base_addr;
1978 k_map.irq = (unsigned char) u_map->irq;
1979 k_map.dma = (unsigned char) u_map->dma;
1980 k_map.port = (unsigned char) u_map->port;
1982 err = ops->ndo_set_config(dev, &k_map);
1986 status |= DO_SETLINK_NOTIFY;
1989 if (tb[IFLA_ADDRESS]) {
1990 struct sockaddr *sa;
1993 len = sizeof(sa_family_t) + max_t(size_t, dev->addr_len,
1995 sa = kmalloc(len, GFP_KERNEL);
2000 sa->sa_family = dev->type;
2001 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
2003 err = dev_set_mac_address(dev, sa);
2007 status |= DO_SETLINK_MODIFIED;
2011 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
2014 status |= DO_SETLINK_MODIFIED;
2017 if (tb[IFLA_GROUP]) {
2018 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2019 status |= DO_SETLINK_NOTIFY;
2023 * Interface selected by interface index but interface
2024 * name provided implies that a name change has been
2027 if (ifm->ifi_index > 0 && ifname[0]) {
2028 err = dev_change_name(dev, ifname);
2031 status |= DO_SETLINK_MODIFIED;
2034 if (tb[IFLA_IFALIAS]) {
2035 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
2036 nla_len(tb[IFLA_IFALIAS]));
2039 status |= DO_SETLINK_NOTIFY;
2042 if (tb[IFLA_BROADCAST]) {
2043 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
2044 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
2047 if (ifm->ifi_flags || ifm->ifi_change) {
2048 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
2053 if (tb[IFLA_MASTER]) {
2054 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
2057 status |= DO_SETLINK_MODIFIED;
2060 if (tb[IFLA_CARRIER]) {
2061 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
2064 status |= DO_SETLINK_MODIFIED;
2067 if (tb[IFLA_TXQLEN]) {
2068 unsigned long value = nla_get_u32(tb[IFLA_TXQLEN]);
2069 unsigned long orig_len = dev->tx_queue_len;
2071 if (dev->tx_queue_len ^ value) {
2072 dev->tx_queue_len = value;
2073 err = call_netdevice_notifiers(
2074 NETDEV_CHANGE_TX_QUEUE_LEN, dev);
2075 err = notifier_to_errno(err);
2077 dev->tx_queue_len = orig_len;
2080 status |= DO_SETLINK_NOTIFY;
2084 if (tb[IFLA_OPERSTATE])
2085 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2087 if (tb[IFLA_LINKMODE]) {
2088 unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
2090 write_lock_bh(&dev_base_lock);
2091 if (dev->link_mode ^ value)
2092 status |= DO_SETLINK_NOTIFY;
2093 dev->link_mode = value;
2094 write_unlock_bh(&dev_base_lock);
2097 if (tb[IFLA_VFINFO_LIST]) {
2098 struct nlattr *vfinfo[IFLA_VF_MAX + 1];
2099 struct nlattr *attr;
2102 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
2103 if (nla_type(attr) != IFLA_VF_INFO ||
2104 nla_len(attr) < NLA_HDRLEN) {
2108 err = nla_parse_nested(vfinfo, IFLA_VF_MAX, attr,
2112 err = do_setvfinfo(dev, vfinfo);
2115 status |= DO_SETLINK_NOTIFY;
2120 if (tb[IFLA_VF_PORTS]) {
2121 struct nlattr *port[IFLA_PORT_MAX+1];
2122 struct nlattr *attr;
2127 if (!ops->ndo_set_vf_port)
2130 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
2131 if (nla_type(attr) != IFLA_VF_PORT ||
2132 nla_len(attr) < NLA_HDRLEN) {
2136 err = nla_parse_nested(port, IFLA_PORT_MAX, attr,
2140 if (!port[IFLA_PORT_VF]) {
2144 vf = nla_get_u32(port[IFLA_PORT_VF]);
2145 err = ops->ndo_set_vf_port(dev, vf, port);
2148 status |= DO_SETLINK_NOTIFY;
2153 if (tb[IFLA_PORT_SELF]) {
2154 struct nlattr *port[IFLA_PORT_MAX+1];
2156 err = nla_parse_nested(port, IFLA_PORT_MAX,
2157 tb[IFLA_PORT_SELF], ifla_port_policy);
2162 if (ops->ndo_set_vf_port)
2163 err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
2166 status |= DO_SETLINK_NOTIFY;
2169 if (tb[IFLA_AF_SPEC]) {
2173 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2174 const struct rtnl_af_ops *af_ops;
2176 if (!(af_ops = rtnl_af_lookup(nla_type(af))))
2179 err = af_ops->set_link_af(dev, af);
2183 status |= DO_SETLINK_NOTIFY;
2188 if (tb[IFLA_PROTO_DOWN]) {
2189 err = dev_change_proto_down(dev,
2190 nla_get_u8(tb[IFLA_PROTO_DOWN]));
2193 status |= DO_SETLINK_NOTIFY;
2197 struct nlattr *xdp[IFLA_XDP_MAX + 1];
2199 err = nla_parse_nested(xdp, IFLA_XDP_MAX, tb[IFLA_XDP],
2204 if (xdp[IFLA_XDP_ATTACHED]) {
2208 if (xdp[IFLA_XDP_FD]) {
2209 err = dev_change_xdp_fd(dev,
2210 nla_get_s32(xdp[IFLA_XDP_FD]));
2213 status |= DO_SETLINK_NOTIFY;
2218 if (status & DO_SETLINK_MODIFIED) {
2219 if (status & DO_SETLINK_NOTIFY)
2220 netdev_state_change(dev);
2223 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",
2230 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
2232 struct net *net = sock_net(skb->sk);
2233 struct ifinfomsg *ifm;
2234 struct net_device *dev;
2236 struct nlattr *tb[IFLA_MAX+1];
2237 char ifname[IFNAMSIZ];
2239 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2243 if (tb[IFLA_IFNAME])
2244 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2249 ifm = nlmsg_data(nlh);
2250 if (ifm->ifi_index > 0)
2251 dev = __dev_get_by_index(net, ifm->ifi_index);
2252 else if (tb[IFLA_IFNAME])
2253 dev = __dev_get_by_name(net, ifname);
2262 err = do_setlink(skb, dev, ifm, tb, ifname, 0);
2267 static int rtnl_group_dellink(const struct net *net, int group)
2269 struct net_device *dev, *aux;
2270 LIST_HEAD(list_kill);
2276 for_each_netdev(net, dev) {
2277 if (dev->group == group) {
2278 const struct rtnl_link_ops *ops;
2281 ops = dev->rtnl_link_ops;
2282 if (!ops || !ops->dellink)
2290 for_each_netdev_safe(net, dev, aux) {
2291 if (dev->group == group) {
2292 const struct rtnl_link_ops *ops;
2294 ops = dev->rtnl_link_ops;
2295 ops->dellink(dev, &list_kill);
2298 unregister_netdevice_many(&list_kill);
2303 int rtnl_delete_link(struct net_device *dev)
2305 const struct rtnl_link_ops *ops;
2306 LIST_HEAD(list_kill);
2308 ops = dev->rtnl_link_ops;
2309 if (!ops || !ops->dellink)
2312 ops->dellink(dev, &list_kill);
2313 unregister_netdevice_many(&list_kill);
2317 EXPORT_SYMBOL_GPL(rtnl_delete_link);
2319 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
2321 struct net *net = sock_net(skb->sk);
2322 struct net_device *dev;
2323 struct ifinfomsg *ifm;
2324 char ifname[IFNAMSIZ];
2325 struct nlattr *tb[IFLA_MAX+1];
2328 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2332 if (tb[IFLA_IFNAME])
2333 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2335 ifm = nlmsg_data(nlh);
2336 if (ifm->ifi_index > 0)
2337 dev = __dev_get_by_index(net, ifm->ifi_index);
2338 else if (tb[IFLA_IFNAME])
2339 dev = __dev_get_by_name(net, ifname);
2340 else if (tb[IFLA_GROUP])
2341 return rtnl_group_dellink(net, nla_get_u32(tb[IFLA_GROUP]));
2348 return rtnl_delete_link(dev);
2351 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
2353 unsigned int old_flags;
2356 old_flags = dev->flags;
2357 if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
2358 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
2363 if (dev->rtnl_link_state == RTNL_LINK_INITIALIZED) {
2364 __dev_notify_flags(dev, old_flags, (old_flags ^ dev->flags));
2366 dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
2367 __dev_notify_flags(dev, old_flags, ~0U);
2371 EXPORT_SYMBOL(rtnl_configure_link);
2373 struct net_device *rtnl_create_link(struct net *net,
2374 const char *ifname, unsigned char name_assign_type,
2375 const struct rtnl_link_ops *ops, struct nlattr *tb[])
2378 struct net_device *dev;
2379 unsigned int num_tx_queues = 1;
2380 unsigned int num_rx_queues = 1;
2382 if (tb[IFLA_NUM_TX_QUEUES])
2383 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
2384 else if (ops->get_num_tx_queues)
2385 num_tx_queues = ops->get_num_tx_queues();
2387 if (tb[IFLA_NUM_RX_QUEUES])
2388 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
2389 else if (ops->get_num_rx_queues)
2390 num_rx_queues = ops->get_num_rx_queues();
2392 if (num_tx_queues < 1 || num_tx_queues > 4096)
2393 return ERR_PTR(-EINVAL);
2395 if (num_rx_queues < 1 || num_rx_queues > 4096)
2396 return ERR_PTR(-EINVAL);
2399 dev = alloc_netdev_mqs(ops->priv_size, ifname, name_assign_type,
2400 ops->setup, num_tx_queues, num_rx_queues);
2404 dev_net_set(dev, net);
2405 dev->rtnl_link_ops = ops;
2406 dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
2409 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
2410 if (tb[IFLA_ADDRESS]) {
2411 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
2412 nla_len(tb[IFLA_ADDRESS]));
2413 dev->addr_assign_type = NET_ADDR_SET;
2415 if (tb[IFLA_BROADCAST])
2416 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
2417 nla_len(tb[IFLA_BROADCAST]));
2418 if (tb[IFLA_TXQLEN])
2419 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
2420 if (tb[IFLA_OPERSTATE])
2421 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2422 if (tb[IFLA_LINKMODE])
2423 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
2425 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2430 return ERR_PTR(err);
2432 EXPORT_SYMBOL(rtnl_create_link);
2434 static int rtnl_group_changelink(const struct sk_buff *skb,
2435 struct net *net, int group,
2436 struct ifinfomsg *ifm,
2439 struct net_device *dev, *aux;
2442 for_each_netdev_safe(net, dev, aux) {
2443 if (dev->group == group) {
2444 err = do_setlink(skb, dev, ifm, tb, NULL, 0);
2453 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh)
2455 struct net *net = sock_net(skb->sk);
2456 const struct rtnl_link_ops *ops;
2457 const struct rtnl_link_ops *m_ops = NULL;
2458 struct net_device *dev;
2459 struct net_device *master_dev = NULL;
2460 struct ifinfomsg *ifm;
2461 char kind[MODULE_NAME_LEN];
2462 char ifname[IFNAMSIZ];
2463 struct nlattr *tb[IFLA_MAX+1];
2464 struct nlattr *linkinfo[IFLA_INFO_MAX+1];
2465 unsigned char name_assign_type = NET_NAME_USER;
2468 #ifdef CONFIG_MODULES
2471 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2475 if (tb[IFLA_IFNAME])
2476 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2480 ifm = nlmsg_data(nlh);
2481 if (ifm->ifi_index > 0)
2482 dev = __dev_get_by_index(net, ifm->ifi_index);
2485 dev = __dev_get_by_name(net, ifname);
2491 master_dev = netdev_master_upper_dev_get(dev);
2493 m_ops = master_dev->rtnl_link_ops;
2496 err = validate_linkmsg(dev, tb);
2500 if (tb[IFLA_LINKINFO]) {
2501 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
2502 tb[IFLA_LINKINFO], ifla_info_policy);
2506 memset(linkinfo, 0, sizeof(linkinfo));
2508 if (linkinfo[IFLA_INFO_KIND]) {
2509 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
2510 ops = rtnl_link_ops_get(kind);
2517 struct nlattr *attr[ops ? ops->maxtype + 1 : 1];
2518 struct nlattr *slave_attr[m_ops ? m_ops->slave_maxtype + 1 : 1];
2519 struct nlattr **data = NULL;
2520 struct nlattr **slave_data = NULL;
2521 struct net *dest_net, *link_net = NULL;
2524 if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
2525 err = nla_parse_nested(attr, ops->maxtype,
2526 linkinfo[IFLA_INFO_DATA],
2532 if (ops->validate) {
2533 err = ops->validate(tb, data);
2540 if (m_ops->slave_maxtype &&
2541 linkinfo[IFLA_INFO_SLAVE_DATA]) {
2542 err = nla_parse_nested(slave_attr,
2543 m_ops->slave_maxtype,
2544 linkinfo[IFLA_INFO_SLAVE_DATA],
2545 m_ops->slave_policy);
2548 slave_data = slave_attr;
2550 if (m_ops->slave_validate) {
2551 err = m_ops->slave_validate(tb, slave_data);
2560 if (nlh->nlmsg_flags & NLM_F_EXCL)
2562 if (nlh->nlmsg_flags & NLM_F_REPLACE)
2565 if (linkinfo[IFLA_INFO_DATA]) {
2566 if (!ops || ops != dev->rtnl_link_ops ||
2570 err = ops->changelink(dev, tb, data);
2573 status |= DO_SETLINK_NOTIFY;
2576 if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
2577 if (!m_ops || !m_ops->slave_changelink)
2580 err = m_ops->slave_changelink(master_dev, dev,
2584 status |= DO_SETLINK_NOTIFY;
2587 return do_setlink(skb, dev, ifm, tb, ifname, status);
2590 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
2591 if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
2592 return rtnl_group_changelink(skb, net,
2593 nla_get_u32(tb[IFLA_GROUP]),
2598 if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
2602 #ifdef CONFIG_MODULES
2605 request_module("rtnl-link-%s", kind);
2607 ops = rtnl_link_ops_get(kind);
2619 snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
2620 name_assign_type = NET_NAME_ENUM;
2623 dest_net = rtnl_link_get_net(net, tb);
2624 if (IS_ERR(dest_net))
2625 return PTR_ERR(dest_net);
2628 if (!netlink_ns_capable(skb, dest_net->user_ns, CAP_NET_ADMIN))
2631 if (tb[IFLA_LINK_NETNSID]) {
2632 int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
2634 link_net = get_net_ns_by_id(dest_net, id);
2640 if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
2644 dev = rtnl_create_link(link_net ? : dest_net, ifname,
2645 name_assign_type, ops, tb);
2651 dev->ifindex = ifm->ifi_index;
2654 err = ops->newlink(link_net ? : net, dev, tb, data);
2655 /* Drivers should call free_netdev() in ->destructor
2656 * and unregister it on failure after registration
2657 * so that device could be finally freed in rtnl_unlock.
2660 /* If device is not registered at all, free it now */
2661 if (dev->reg_state == NETREG_UNINITIALIZED)
2666 err = register_netdevice(dev);
2672 err = rtnl_configure_link(dev, ifm);
2674 goto out_unregister;
2676 err = dev_change_net_namespace(dev, dest_net, ifname);
2678 goto out_unregister;
2687 LIST_HEAD(list_kill);
2689 ops->dellink(dev, &list_kill);
2690 unregister_netdevice_many(&list_kill);
2692 unregister_netdevice(dev);
2698 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh)
2700 struct net *net = sock_net(skb->sk);
2701 struct ifinfomsg *ifm;
2702 char ifname[IFNAMSIZ];
2703 struct nlattr *tb[IFLA_MAX+1];
2704 struct net_device *dev = NULL;
2705 struct sk_buff *nskb;
2707 u32 ext_filter_mask = 0;
2709 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2713 if (tb[IFLA_IFNAME])
2714 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2716 if (tb[IFLA_EXT_MASK])
2717 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2719 ifm = nlmsg_data(nlh);
2720 if (ifm->ifi_index > 0)
2721 dev = __dev_get_by_index(net, ifm->ifi_index);
2722 else if (tb[IFLA_IFNAME])
2723 dev = __dev_get_by_name(net, ifname);
2730 nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
2734 err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).portid,
2735 nlh->nlmsg_seq, 0, 0, ext_filter_mask);
2737 /* -EMSGSIZE implies BUG in if_nlmsg_size */
2738 WARN_ON(err == -EMSGSIZE);
2741 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
2746 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
2748 struct net *net = sock_net(skb->sk);
2749 struct net_device *dev;
2750 struct nlattr *tb[IFLA_MAX+1];
2751 u32 ext_filter_mask = 0;
2752 u16 min_ifinfo_dump_size = 0;
2755 /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
2756 hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
2757 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
2759 if (nlmsg_parse(nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
2760 if (tb[IFLA_EXT_MASK])
2761 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2764 if (!ext_filter_mask)
2765 return NLMSG_GOODSIZE;
2767 * traverse the list of net devices and compute the minimum
2768 * buffer size based upon the filter mask.
2770 list_for_each_entry(dev, &net->dev_base_head, dev_list) {
2771 min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
2776 return nlmsg_total_size(min_ifinfo_dump_size);
2779 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
2782 int s_idx = cb->family;
2786 for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
2787 int type = cb->nlh->nlmsg_type-RTM_BASE;
2788 if (idx < s_idx || idx == PF_PACKET)
2790 if (rtnl_msg_handlers[idx] == NULL ||
2791 rtnl_msg_handlers[idx][type].dumpit == NULL)
2794 memset(&cb->args[0], 0, sizeof(cb->args));
2798 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
2806 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
2807 unsigned int change, gfp_t flags)
2809 struct net *net = dev_net(dev);
2810 struct sk_buff *skb;
2812 size_t if_info_size;
2814 skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
2818 err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0);
2820 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
2821 WARN_ON(err == -EMSGSIZE);
2828 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2832 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags)
2834 struct net *net = dev_net(dev);
2836 rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
2839 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
2842 struct sk_buff *skb;
2844 if (dev->reg_state != NETREG_REGISTERED)
2847 skb = rtmsg_ifinfo_build_skb(type, dev, change, flags);
2849 rtmsg_ifinfo_send(skb, dev, flags);
2851 EXPORT_SYMBOL(rtmsg_ifinfo);
2853 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
2854 struct net_device *dev,
2855 u8 *addr, u16 vid, u32 pid, u32 seq,
2856 int type, unsigned int flags,
2857 int nlflags, u16 ndm_state)
2859 struct nlmsghdr *nlh;
2862 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
2866 ndm = nlmsg_data(nlh);
2867 ndm->ndm_family = AF_BRIDGE;
2870 ndm->ndm_flags = flags;
2872 ndm->ndm_ifindex = dev->ifindex;
2873 ndm->ndm_state = ndm_state;
2875 if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
2876 goto nla_put_failure;
2878 if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
2879 goto nla_put_failure;
2881 nlmsg_end(skb, nlh);
2885 nlmsg_cancel(skb, nlh);
2889 static inline size_t rtnl_fdb_nlmsg_size(void)
2891 return NLMSG_ALIGN(sizeof(struct ndmsg)) +
2892 nla_total_size(ETH_ALEN) + /* NDA_LLADDR */
2893 nla_total_size(sizeof(u16)) + /* NDA_VLAN */
2897 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type,
2900 struct net *net = dev_net(dev);
2901 struct sk_buff *skb;
2904 skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
2908 err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
2909 0, 0, type, NTF_SELF, 0, ndm_state);
2915 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2918 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
2922 * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
2924 int ndo_dflt_fdb_add(struct ndmsg *ndm,
2925 struct nlattr *tb[],
2926 struct net_device *dev,
2927 const unsigned char *addr, u16 vid,
2932 /* If aging addresses are supported device will need to
2933 * implement its own handler for this.
2935 if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
2936 pr_info("%s: FDB only supports static addresses\n", dev->name);
2941 pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
2945 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2946 err = dev_uc_add_excl(dev, addr);
2947 else if (is_multicast_ether_addr(addr))
2948 err = dev_mc_add_excl(dev, addr);
2950 /* Only return duplicate errors if NLM_F_EXCL is set */
2951 if (err == -EEXIST && !(flags & NLM_F_EXCL))
2956 EXPORT_SYMBOL(ndo_dflt_fdb_add);
2958 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid)
2963 if (nla_len(vlan_attr) != sizeof(u16)) {
2964 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan\n");
2968 vid = nla_get_u16(vlan_attr);
2970 if (!vid || vid >= VLAN_VID_MASK) {
2971 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan id %d\n",
2980 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh)
2982 struct net *net = sock_net(skb->sk);
2984 struct nlattr *tb[NDA_MAX+1];
2985 struct net_device *dev;
2990 err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2994 ndm = nlmsg_data(nlh);
2995 if (ndm->ndm_ifindex == 0) {
2996 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n");
3000 dev = __dev_get_by_index(net, ndm->ndm_ifindex);
3002 pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n");
3006 if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
3007 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n");
3011 if (dev->type != ARPHRD_ETHER) {
3012 pr_info("PF_BRIDGE: FDB add only supported for Ethernet devices");
3016 addr = nla_data(tb[NDA_LLADDR]);
3018 err = fdb_vid_parse(tb[NDA_VLAN], &vid);
3024 /* Support fdb on master device the net/bridge default case */
3025 if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
3026 (dev->priv_flags & IFF_BRIDGE_PORT)) {
3027 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3028 const struct net_device_ops *ops = br_dev->netdev_ops;
3030 err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
3035 ndm->ndm_flags &= ~NTF_MASTER;
3038 /* Embedded bridge, macvlan, and any other device support */
3039 if ((ndm->ndm_flags & NTF_SELF)) {
3040 if (dev->netdev_ops->ndo_fdb_add)
3041 err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
3045 err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
3049 rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH,
3051 ndm->ndm_flags &= ~NTF_SELF;
3059 * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
3061 int ndo_dflt_fdb_del(struct ndmsg *ndm,
3062 struct nlattr *tb[],
3063 struct net_device *dev,
3064 const unsigned char *addr, u16 vid)
3068 /* If aging addresses are supported device will need to
3069 * implement its own handler for this.
3071 if (!(ndm->ndm_state & NUD_PERMANENT)) {
3072 pr_info("%s: FDB only supports static addresses\n", dev->name);
3076 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
3077 err = dev_uc_del(dev, addr);
3078 else if (is_multicast_ether_addr(addr))
3079 err = dev_mc_del(dev, addr);
3083 EXPORT_SYMBOL(ndo_dflt_fdb_del);
3085 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh)
3087 struct net *net = sock_net(skb->sk);
3089 struct nlattr *tb[NDA_MAX+1];
3090 struct net_device *dev;
3095 if (!netlink_capable(skb, CAP_NET_ADMIN))
3098 err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
3102 ndm = nlmsg_data(nlh);
3103 if (ndm->ndm_ifindex == 0) {
3104 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n");
3108 dev = __dev_get_by_index(net, ndm->ndm_ifindex);
3110 pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n");
3114 if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
3115 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid address\n");
3119 if (dev->type != ARPHRD_ETHER) {
3120 pr_info("PF_BRIDGE: FDB delete only supported for Ethernet devices");
3124 addr = nla_data(tb[NDA_LLADDR]);
3126 err = fdb_vid_parse(tb[NDA_VLAN], &vid);
3132 /* Support fdb on master device the net/bridge default case */
3133 if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
3134 (dev->priv_flags & IFF_BRIDGE_PORT)) {
3135 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3136 const struct net_device_ops *ops = br_dev->netdev_ops;
3138 if (ops->ndo_fdb_del)
3139 err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid);
3144 ndm->ndm_flags &= ~NTF_MASTER;
3147 /* Embedded bridge, macvlan, and any other device support */
3148 if (ndm->ndm_flags & NTF_SELF) {
3149 if (dev->netdev_ops->ndo_fdb_del)
3150 err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr,
3153 err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
3156 rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH,
3158 ndm->ndm_flags &= ~NTF_SELF;
3165 static int nlmsg_populate_fdb(struct sk_buff *skb,
3166 struct netlink_callback *cb,
3167 struct net_device *dev,
3169 struct netdev_hw_addr_list *list)
3171 struct netdev_hw_addr *ha;
3175 portid = NETLINK_CB(cb->skb).portid;
3176 seq = cb->nlh->nlmsg_seq;
3178 list_for_each_entry(ha, &list->list, list) {
3179 if (*idx < cb->args[2])
3182 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
3184 RTM_NEWNEIGH, NTF_SELF,
3185 NLM_F_MULTI, NUD_PERMANENT);
3195 * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
3196 * @nlh: netlink message header
3199 * Default netdevice operation to dump the existing unicast address list.
3200 * Returns number of addresses from list put in skb.
3202 int ndo_dflt_fdb_dump(struct sk_buff *skb,
3203 struct netlink_callback *cb,
3204 struct net_device *dev,
3205 struct net_device *filter_dev,
3210 if (dev->type != ARPHRD_ETHER)
3213 netif_addr_lock_bh(dev);
3214 err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->uc);
3217 nlmsg_populate_fdb(skb, cb, dev, idx, &dev->mc);
3219 netif_addr_unlock_bh(dev);
3222 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
3224 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
3226 struct net_device *dev;
3227 struct nlattr *tb[IFLA_MAX+1];
3228 struct net_device *br_dev = NULL;
3229 const struct net_device_ops *ops = NULL;
3230 const struct net_device_ops *cops = NULL;
3231 struct ifinfomsg *ifm = nlmsg_data(cb->nlh);
3232 struct net *net = sock_net(skb->sk);
3233 struct hlist_head *head;
3241 if (nlmsg_parse(cb->nlh, sizeof(struct ifinfomsg), tb, IFLA_MAX,
3242 ifla_policy) == 0) {
3243 if (tb[IFLA_MASTER])
3244 br_idx = nla_get_u32(tb[IFLA_MASTER]);
3247 brport_idx = ifm->ifi_index;
3250 br_dev = __dev_get_by_index(net, br_idx);
3254 ops = br_dev->netdev_ops;
3258 s_idx = cb->args[1];
3260 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3262 head = &net->dev_index_head[h];
3263 hlist_for_each_entry(dev, head, index_hlist) {
3265 if (brport_idx && (dev->ifindex != brport_idx))
3268 if (!br_idx) { /* user did not specify a specific bridge */
3269 if (dev->priv_flags & IFF_BRIDGE_PORT) {
3270 br_dev = netdev_master_upper_dev_get(dev);
3271 cops = br_dev->netdev_ops;
3274 if (dev != br_dev &&
3275 !(dev->priv_flags & IFF_BRIDGE_PORT))
3278 if (br_dev != netdev_master_upper_dev_get(dev) &&
3279 !(dev->priv_flags & IFF_EBRIDGE))
3287 if (dev->priv_flags & IFF_BRIDGE_PORT) {
3288 if (cops && cops->ndo_fdb_dump) {
3289 err = cops->ndo_fdb_dump(skb, cb,
3292 if (err == -EMSGSIZE)
3297 if (dev->netdev_ops->ndo_fdb_dump)
3298 err = dev->netdev_ops->ndo_fdb_dump(skb, cb,
3302 err = ndo_dflt_fdb_dump(skb, cb, dev, NULL,
3304 if (err == -EMSGSIZE)
3309 /* reset fdb offset to 0 for rest of the interfaces */
3325 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
3326 unsigned int attrnum, unsigned int flag)
3329 return nla_put_u8(skb, attrnum, !!(flags & flag));
3333 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
3334 struct net_device *dev, u16 mode,
3335 u32 flags, u32 mask, int nlflags,
3337 int (*vlan_fill)(struct sk_buff *skb,
3338 struct net_device *dev,
3341 struct nlmsghdr *nlh;
3342 struct ifinfomsg *ifm;
3343 struct nlattr *br_afspec;
3344 struct nlattr *protinfo;
3345 u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
3346 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3349 nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
3353 ifm = nlmsg_data(nlh);
3354 ifm->ifi_family = AF_BRIDGE;
3356 ifm->ifi_type = dev->type;
3357 ifm->ifi_index = dev->ifindex;
3358 ifm->ifi_flags = dev_get_flags(dev);
3359 ifm->ifi_change = 0;
3362 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
3363 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
3364 nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
3366 nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
3368 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
3369 (dev->ifindex != dev_get_iflink(dev) &&
3370 nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
3371 goto nla_put_failure;
3373 br_afspec = nla_nest_start(skb, IFLA_AF_SPEC);
3375 goto nla_put_failure;
3377 if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
3378 nla_nest_cancel(skb, br_afspec);
3379 goto nla_put_failure;
3382 if (mode != BRIDGE_MODE_UNDEF) {
3383 if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
3384 nla_nest_cancel(skb, br_afspec);
3385 goto nla_put_failure;
3389 err = vlan_fill(skb, dev, filter_mask);
3391 nla_nest_cancel(skb, br_afspec);
3392 goto nla_put_failure;
3395 nla_nest_end(skb, br_afspec);
3397 protinfo = nla_nest_start(skb, IFLA_PROTINFO | NLA_F_NESTED);
3399 goto nla_put_failure;
3401 if (brport_nla_put_flag(skb, flags, mask,
3402 IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
3403 brport_nla_put_flag(skb, flags, mask,
3404 IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
3405 brport_nla_put_flag(skb, flags, mask,
3406 IFLA_BRPORT_FAST_LEAVE,
3407 BR_MULTICAST_FAST_LEAVE) ||
3408 brport_nla_put_flag(skb, flags, mask,
3409 IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
3410 brport_nla_put_flag(skb, flags, mask,
3411 IFLA_BRPORT_LEARNING, BR_LEARNING) ||
3412 brport_nla_put_flag(skb, flags, mask,
3413 IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
3414 brport_nla_put_flag(skb, flags, mask,
3415 IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
3416 brport_nla_put_flag(skb, flags, mask,
3417 IFLA_BRPORT_PROXYARP, BR_PROXYARP)) {
3418 nla_nest_cancel(skb, protinfo);
3419 goto nla_put_failure;
3422 nla_nest_end(skb, protinfo);
3424 nlmsg_end(skb, nlh);
3427 nlmsg_cancel(skb, nlh);
3428 return err ? err : -EMSGSIZE;
3430 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink);
3432 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
3434 struct net *net = sock_net(skb->sk);
3435 struct net_device *dev;
3437 u32 portid = NETLINK_CB(cb->skb).portid;
3438 u32 seq = cb->nlh->nlmsg_seq;
3439 u32 filter_mask = 0;
3442 if (nlmsg_len(cb->nlh) > sizeof(struct ifinfomsg)) {
3443 struct nlattr *extfilt;
3445 extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct ifinfomsg),
3448 if (nla_len(extfilt) < sizeof(filter_mask))
3451 filter_mask = nla_get_u32(extfilt);
3456 for_each_netdev_rcu(net, dev) {
3457 const struct net_device_ops *ops = dev->netdev_ops;
3458 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3460 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
3461 if (idx >= cb->args[0]) {
3462 err = br_dev->netdev_ops->ndo_bridge_getlink(
3463 skb, portid, seq, dev,
3464 filter_mask, NLM_F_MULTI);
3465 if (err < 0 && err != -EOPNOTSUPP) {
3466 if (likely(skb->len))
3475 if (ops->ndo_bridge_getlink) {
3476 if (idx >= cb->args[0]) {
3477 err = ops->ndo_bridge_getlink(skb, portid,
3481 if (err < 0 && err != -EOPNOTSUPP) {
3482 if (likely(skb->len))
3499 static inline size_t bridge_nlmsg_size(void)
3501 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3502 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3503 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3504 + nla_total_size(sizeof(u32)) /* IFLA_MASTER */
3505 + nla_total_size(sizeof(u32)) /* IFLA_MTU */
3506 + nla_total_size(sizeof(u32)) /* IFLA_LINK */
3507 + nla_total_size(sizeof(u32)) /* IFLA_OPERSTATE */
3508 + nla_total_size(sizeof(u8)) /* IFLA_PROTINFO */
3509 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
3510 + nla_total_size(sizeof(u16)) /* IFLA_BRIDGE_FLAGS */
3511 + nla_total_size(sizeof(u16)); /* IFLA_BRIDGE_MODE */
3514 static int rtnl_bridge_notify(struct net_device *dev)
3516 struct net *net = dev_net(dev);
3517 struct sk_buff *skb;
3518 int err = -EOPNOTSUPP;
3520 if (!dev->netdev_ops->ndo_bridge_getlink)
3523 skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
3529 err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
3533 /* Notification info is only filled for bridge ports, not the bridge
3534 * device itself. Therefore, a zero notification length is valid and
3535 * should not result in an error.
3540 rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
3543 WARN_ON(err == -EMSGSIZE);
3546 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
3550 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
3552 struct net *net = sock_net(skb->sk);
3553 struct ifinfomsg *ifm;
3554 struct net_device *dev;
3555 struct nlattr *br_spec, *attr = NULL;
3556 int rem, err = -EOPNOTSUPP;
3558 bool have_flags = false;
3560 if (nlmsg_len(nlh) < sizeof(*ifm))
3563 ifm = nlmsg_data(nlh);
3564 if (ifm->ifi_family != AF_BRIDGE)
3565 return -EPFNOSUPPORT;
3567 dev = __dev_get_by_index(net, ifm->ifi_index);
3569 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3573 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3575 nla_for_each_nested(attr, br_spec, rem) {
3576 if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3577 if (nla_len(attr) < sizeof(flags))
3581 flags = nla_get_u16(attr);
3587 if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3588 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3590 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
3595 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags);
3599 flags &= ~BRIDGE_FLAGS_MASTER;
3602 if ((flags & BRIDGE_FLAGS_SELF)) {
3603 if (!dev->netdev_ops->ndo_bridge_setlink)
3606 err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
3609 flags &= ~BRIDGE_FLAGS_SELF;
3611 /* Generate event to notify upper layer of bridge
3614 err = rtnl_bridge_notify(dev);
3619 memcpy(nla_data(attr), &flags, sizeof(flags));
3624 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
3626 struct net *net = sock_net(skb->sk);
3627 struct ifinfomsg *ifm;
3628 struct net_device *dev;
3629 struct nlattr *br_spec, *attr = NULL;
3630 int rem, err = -EOPNOTSUPP;
3632 bool have_flags = false;
3634 if (nlmsg_len(nlh) < sizeof(*ifm))
3637 ifm = nlmsg_data(nlh);
3638 if (ifm->ifi_family != AF_BRIDGE)
3639 return -EPFNOSUPPORT;
3641 dev = __dev_get_by_index(net, ifm->ifi_index);
3643 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3647 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3649 nla_for_each_nested(attr, br_spec, rem) {
3650 if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3651 if (nla_len(attr) < sizeof(flags))
3655 flags = nla_get_u16(attr);
3661 if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3662 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3664 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
3669 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
3673 flags &= ~BRIDGE_FLAGS_MASTER;
3676 if ((flags & BRIDGE_FLAGS_SELF)) {
3677 if (!dev->netdev_ops->ndo_bridge_dellink)
3680 err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
3684 flags &= ~BRIDGE_FLAGS_SELF;
3686 /* Generate event to notify upper layer of bridge
3689 err = rtnl_bridge_notify(dev);
3694 memcpy(nla_data(attr), &flags, sizeof(flags));
3699 static bool stats_attr_valid(unsigned int mask, int attrid, int idxattr)
3701 return (mask & IFLA_STATS_FILTER_BIT(attrid)) &&
3702 (!idxattr || idxattr == attrid);
3705 #define IFLA_OFFLOAD_XSTATS_FIRST (IFLA_OFFLOAD_XSTATS_UNSPEC + 1)
3706 static int rtnl_get_offload_stats_attr_size(int attr_id)
3709 case IFLA_OFFLOAD_XSTATS_CPU_HIT:
3710 return sizeof(struct rtnl_link_stats64);
3716 static int rtnl_get_offload_stats(struct sk_buff *skb, struct net_device *dev,
3719 struct nlattr *attr = NULL;
3724 if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
3725 dev->netdev_ops->ndo_get_offload_stats))
3728 for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
3729 attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
3730 if (attr_id < *prividx)
3733 size = rtnl_get_offload_stats_attr_size(attr_id);
3737 if (!dev->netdev_ops->ndo_has_offload_stats(attr_id))
3740 attr = nla_reserve_64bit(skb, attr_id, size,
3741 IFLA_OFFLOAD_XSTATS_UNSPEC);
3743 goto nla_put_failure;
3745 attr_data = nla_data(attr);
3746 memset(attr_data, 0, size);
3747 err = dev->netdev_ops->ndo_get_offload_stats(attr_id, dev,
3750 goto get_offload_stats_failure;
3761 get_offload_stats_failure:
3766 static int rtnl_get_offload_stats_size(const struct net_device *dev)
3772 if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
3773 dev->netdev_ops->ndo_get_offload_stats))
3776 for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
3777 attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
3778 if (!dev->netdev_ops->ndo_has_offload_stats(attr_id))
3780 size = rtnl_get_offload_stats_attr_size(attr_id);
3781 nla_size += nla_total_size_64bit(size);
3785 nla_size += nla_total_size(0);
3790 static int rtnl_fill_statsinfo(struct sk_buff *skb, struct net_device *dev,
3791 int type, u32 pid, u32 seq, u32 change,
3792 unsigned int flags, unsigned int filter_mask,
3793 int *idxattr, int *prividx)
3795 struct if_stats_msg *ifsm;
3796 struct nlmsghdr *nlh;
3797 struct nlattr *attr;
3798 int s_prividx = *prividx;
3803 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifsm), flags);
3807 ifsm = nlmsg_data(nlh);
3808 ifsm->family = PF_UNSPEC;
3811 ifsm->ifindex = dev->ifindex;
3812 ifsm->filter_mask = filter_mask;
3814 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, *idxattr)) {
3815 struct rtnl_link_stats64 *sp;
3817 attr = nla_reserve_64bit(skb, IFLA_STATS_LINK_64,
3818 sizeof(struct rtnl_link_stats64),
3821 goto nla_put_failure;
3823 sp = nla_data(attr);
3824 dev_get_stats(dev, sp);
3827 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, *idxattr)) {
3828 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
3830 if (ops && ops->fill_linkxstats) {
3831 *idxattr = IFLA_STATS_LINK_XSTATS;
3832 attr = nla_nest_start(skb,
3833 IFLA_STATS_LINK_XSTATS);
3835 goto nla_put_failure;
3837 err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
3838 nla_nest_end(skb, attr);
3840 goto nla_put_failure;
3845 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE,
3847 const struct rtnl_link_ops *ops = NULL;
3848 const struct net_device *master;
3850 master = netdev_master_upper_dev_get(dev);
3852 ops = master->rtnl_link_ops;
3853 if (ops && ops->fill_linkxstats) {
3854 *idxattr = IFLA_STATS_LINK_XSTATS_SLAVE;
3855 attr = nla_nest_start(skb,
3856 IFLA_STATS_LINK_XSTATS_SLAVE);
3858 goto nla_put_failure;
3860 err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
3861 nla_nest_end(skb, attr);
3863 goto nla_put_failure;
3868 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS,
3870 *idxattr = IFLA_STATS_LINK_OFFLOAD_XSTATS;
3871 attr = nla_nest_start(skb, IFLA_STATS_LINK_OFFLOAD_XSTATS);
3873 goto nla_put_failure;
3875 err = rtnl_get_offload_stats(skb, dev, prividx);
3876 if (err == -ENODATA)
3877 nla_nest_cancel(skb, attr);
3879 nla_nest_end(skb, attr);
3881 if (err && err != -ENODATA)
3882 goto nla_put_failure;
3886 nlmsg_end(skb, nlh);
3891 /* not a multi message or no progress mean a real error */
3892 if (!(flags & NLM_F_MULTI) || s_prividx == *prividx)
3893 nlmsg_cancel(skb, nlh);
3895 nlmsg_end(skb, nlh);
3900 static size_t if_nlmsg_stats_size(const struct net_device *dev,
3903 size_t size = NLMSG_ALIGN(sizeof(struct if_stats_msg));
3905 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, 0))
3906 size += nla_total_size_64bit(sizeof(struct rtnl_link_stats64));
3908 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, 0)) {
3909 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
3910 int attr = IFLA_STATS_LINK_XSTATS;
3912 if (ops && ops->get_linkxstats_size) {
3913 size += nla_total_size(ops->get_linkxstats_size(dev,
3915 /* for IFLA_STATS_LINK_XSTATS */
3916 size += nla_total_size(0);
3920 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE, 0)) {
3921 struct net_device *_dev = (struct net_device *)dev;
3922 const struct rtnl_link_ops *ops = NULL;
3923 const struct net_device *master;
3925 /* netdev_master_upper_dev_get can't take const */
3926 master = netdev_master_upper_dev_get(_dev);
3928 ops = master->rtnl_link_ops;
3929 if (ops && ops->get_linkxstats_size) {
3930 int attr = IFLA_STATS_LINK_XSTATS_SLAVE;
3932 size += nla_total_size(ops->get_linkxstats_size(dev,
3934 /* for IFLA_STATS_LINK_XSTATS_SLAVE */
3935 size += nla_total_size(0);
3939 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS, 0))
3940 size += rtnl_get_offload_stats_size(dev);
3945 static int rtnl_stats_get(struct sk_buff *skb, struct nlmsghdr *nlh)
3947 struct net *net = sock_net(skb->sk);
3948 struct net_device *dev = NULL;
3949 int idxattr = 0, prividx = 0;
3950 struct if_stats_msg *ifsm;
3951 struct sk_buff *nskb;
3955 if (nlmsg_len(nlh) < sizeof(*ifsm))
3958 ifsm = nlmsg_data(nlh);
3959 if (ifsm->ifindex > 0)
3960 dev = __dev_get_by_index(net, ifsm->ifindex);
3967 filter_mask = ifsm->filter_mask;
3971 nskb = nlmsg_new(if_nlmsg_stats_size(dev, filter_mask), GFP_KERNEL);
3975 err = rtnl_fill_statsinfo(nskb, dev, RTM_NEWSTATS,
3976 NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
3977 0, filter_mask, &idxattr, &prividx);
3979 /* -EMSGSIZE implies BUG in if_nlmsg_stats_size */
3980 WARN_ON(err == -EMSGSIZE);
3983 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
3989 static int rtnl_stats_dump(struct sk_buff *skb, struct netlink_callback *cb)
3991 int h, s_h, err, s_idx, s_idxattr, s_prividx;
3992 struct net *net = sock_net(skb->sk);
3993 unsigned int flags = NLM_F_MULTI;
3994 struct if_stats_msg *ifsm;
3995 struct hlist_head *head;
3996 struct net_device *dev;
3997 u32 filter_mask = 0;
4001 s_idx = cb->args[1];
4002 s_idxattr = cb->args[2];
4003 s_prividx = cb->args[3];
4005 cb->seq = net->dev_base_seq;
4007 if (nlmsg_len(cb->nlh) < sizeof(*ifsm))
4010 ifsm = nlmsg_data(cb->nlh);
4011 filter_mask = ifsm->filter_mask;
4015 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4017 head = &net->dev_index_head[h];
4018 hlist_for_each_entry(dev, head, index_hlist) {
4021 err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS,
4022 NETLINK_CB(cb->skb).portid,
4023 cb->nlh->nlmsg_seq, 0,
4025 &s_idxattr, &s_prividx);
4026 /* If we ran out of room on the first message,
4029 WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
4035 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4041 cb->args[3] = s_prividx;
4042 cb->args[2] = s_idxattr;
4049 /* Process one rtnetlink message. */
4051 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
4053 struct net *net = sock_net(skb->sk);
4054 rtnl_doit_func doit;
4060 type = nlh->nlmsg_type;
4066 /* All the messages must have at least 1 byte length */
4067 if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
4070 family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
4073 if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
4076 if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
4078 rtnl_dumpit_func dumpit;
4079 rtnl_calcit_func calcit;
4080 u16 min_dump_alloc = 0;
4082 dumpit = rtnl_get_dumpit(family, type);
4085 calcit = rtnl_get_calcit(family, type);
4087 min_dump_alloc = calcit(skb, nlh);
4092 struct netlink_dump_control c = {
4094 .min_dump_alloc = min_dump_alloc,
4096 err = netlink_dump_start(rtnl, skb, nlh, &c);
4102 doit = rtnl_get_doit(family, type);
4106 return doit(skb, nlh);
4109 static void rtnetlink_rcv(struct sk_buff *skb)
4112 netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
4116 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
4118 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
4124 case NETDEV_POST_INIT:
4125 case NETDEV_REGISTER:
4127 case NETDEV_PRE_TYPE_CHANGE:
4128 case NETDEV_GOING_DOWN:
4129 case NETDEV_UNREGISTER:
4130 case NETDEV_UNREGISTER_FINAL:
4131 case NETDEV_RELEASE:
4133 case NETDEV_BONDING_INFO:
4136 rtmsg_ifinfo(RTM_NEWLINK, dev, 0, GFP_KERNEL);
4142 static struct notifier_block rtnetlink_dev_notifier = {
4143 .notifier_call = rtnetlink_event,
4147 static int __net_init rtnetlink_net_init(struct net *net)
4150 struct netlink_kernel_cfg cfg = {
4151 .groups = RTNLGRP_MAX,
4152 .input = rtnetlink_rcv,
4153 .cb_mutex = &rtnl_mutex,
4154 .flags = NL_CFG_F_NONROOT_RECV,
4157 sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
4164 static void __net_exit rtnetlink_net_exit(struct net *net)
4166 netlink_kernel_release(net->rtnl);
4170 static struct pernet_operations rtnetlink_net_ops = {
4171 .init = rtnetlink_net_init,
4172 .exit = rtnetlink_net_exit,
4175 void __init rtnetlink_init(void)
4177 if (register_pernet_subsys(&rtnetlink_net_ops))
4178 panic("rtnetlink_init: cannot initialize rtnetlink\n");
4180 register_netdevice_notifier(&rtnetlink_dev_notifier);
4182 rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
4183 rtnl_dump_ifinfo, rtnl_calcit);
4184 rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
4185 rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
4186 rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
4188 rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
4189 rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
4191 rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, NULL);
4192 rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, NULL);
4193 rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, NULL);
4195 rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, NULL);
4196 rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, NULL);
4197 rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, NULL);
4199 rtnl_register(PF_UNSPEC, RTM_GETSTATS, rtnl_stats_get, rtnl_stats_dump,