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>
54 #include <net/pkt_sched.h>
55 #include <net/fib_rules.h>
56 #include <net/rtnetlink.h>
57 #include <net/net_namespace.h>
61 rtnl_dumpit_func dumpit;
62 rtnl_calcit_func calcit;
65 static DEFINE_MUTEX(rtnl_mutex);
69 mutex_lock(&rtnl_mutex);
71 EXPORT_SYMBOL(rtnl_lock);
73 void __rtnl_unlock(void)
75 mutex_unlock(&rtnl_mutex);
78 void rtnl_unlock(void)
80 /* This fellow will unlock it for us. */
83 EXPORT_SYMBOL(rtnl_unlock);
85 int rtnl_trylock(void)
87 return mutex_trylock(&rtnl_mutex);
89 EXPORT_SYMBOL(rtnl_trylock);
91 int rtnl_is_locked(void)
93 return mutex_is_locked(&rtnl_mutex);
95 EXPORT_SYMBOL(rtnl_is_locked);
97 #ifdef CONFIG_PROVE_LOCKING
98 int lockdep_rtnl_is_held(void)
100 return lockdep_is_held(&rtnl_mutex);
102 EXPORT_SYMBOL(lockdep_rtnl_is_held);
103 #endif /* #ifdef CONFIG_PROVE_LOCKING */
105 static struct rtnl_link *rtnl_msg_handlers[RTNL_FAMILY_MAX + 1];
107 static inline int rtm_msgindex(int msgtype)
109 int msgindex = msgtype - RTM_BASE;
112 * msgindex < 0 implies someone tried to register a netlink
113 * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
114 * the message type has not been added to linux/rtnetlink.h
116 BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
121 static rtnl_doit_func rtnl_get_doit(int protocol, int msgindex)
123 struct rtnl_link *tab;
125 if (protocol <= RTNL_FAMILY_MAX)
126 tab = rtnl_msg_handlers[protocol];
130 if (tab == NULL || tab[msgindex].doit == NULL)
131 tab = rtnl_msg_handlers[PF_UNSPEC];
133 return tab[msgindex].doit;
136 static rtnl_dumpit_func rtnl_get_dumpit(int protocol, int msgindex)
138 struct rtnl_link *tab;
140 if (protocol <= RTNL_FAMILY_MAX)
141 tab = rtnl_msg_handlers[protocol];
145 if (tab == NULL || tab[msgindex].dumpit == NULL)
146 tab = rtnl_msg_handlers[PF_UNSPEC];
148 return tab[msgindex].dumpit;
151 static rtnl_calcit_func rtnl_get_calcit(int protocol, int msgindex)
153 struct rtnl_link *tab;
155 if (protocol <= RTNL_FAMILY_MAX)
156 tab = rtnl_msg_handlers[protocol];
160 if (tab == NULL || tab[msgindex].calcit == NULL)
161 tab = rtnl_msg_handlers[PF_UNSPEC];
163 return tab[msgindex].calcit;
167 * __rtnl_register - Register a rtnetlink message type
168 * @protocol: Protocol family or PF_UNSPEC
169 * @msgtype: rtnetlink message type
170 * @doit: Function pointer called for each request message
171 * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
172 * @calcit: Function pointer to calc size of dump message
174 * Registers the specified function pointers (at least one of them has
175 * to be non-NULL) to be called whenever a request message for the
176 * specified protocol family and message type is received.
178 * The special protocol family PF_UNSPEC may be used to define fallback
179 * function pointers for the case when no entry for the specific protocol
182 * Returns 0 on success or a negative error code.
184 int __rtnl_register(int protocol, int msgtype,
185 rtnl_doit_func doit, rtnl_dumpit_func dumpit,
186 rtnl_calcit_func calcit)
188 struct rtnl_link *tab;
191 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
192 msgindex = rtm_msgindex(msgtype);
194 tab = rtnl_msg_handlers[protocol];
196 tab = kcalloc(RTM_NR_MSGTYPES, sizeof(*tab), GFP_KERNEL);
200 rtnl_msg_handlers[protocol] = tab;
204 tab[msgindex].doit = doit;
207 tab[msgindex].dumpit = dumpit;
210 tab[msgindex].calcit = calcit;
214 EXPORT_SYMBOL_GPL(__rtnl_register);
217 * rtnl_register - Register a rtnetlink message type
219 * Identical to __rtnl_register() but panics on failure. This is useful
220 * as failure of this function is very unlikely, it can only happen due
221 * to lack of memory when allocating the chain to store all message
222 * handlers for a protocol. Meant for use in init functions where lack
223 * of memory implies no sense in continuing.
225 void rtnl_register(int protocol, int msgtype,
226 rtnl_doit_func doit, rtnl_dumpit_func dumpit,
227 rtnl_calcit_func calcit)
229 if (__rtnl_register(protocol, msgtype, doit, dumpit, calcit) < 0)
230 panic("Unable to register rtnetlink message handler, "
231 "protocol = %d, message type = %d\n",
234 EXPORT_SYMBOL_GPL(rtnl_register);
237 * rtnl_unregister - Unregister a rtnetlink message type
238 * @protocol: Protocol family or PF_UNSPEC
239 * @msgtype: rtnetlink message type
241 * Returns 0 on success or a negative error code.
243 int rtnl_unregister(int protocol, int msgtype)
247 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
248 msgindex = rtm_msgindex(msgtype);
250 if (rtnl_msg_handlers[protocol] == NULL)
253 rtnl_msg_handlers[protocol][msgindex].doit = NULL;
254 rtnl_msg_handlers[protocol][msgindex].dumpit = NULL;
258 EXPORT_SYMBOL_GPL(rtnl_unregister);
261 * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
262 * @protocol : Protocol family or PF_UNSPEC
264 * Identical to calling rtnl_unregster() for all registered message types
265 * of a certain protocol family.
267 void rtnl_unregister_all(int protocol)
269 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
271 kfree(rtnl_msg_handlers[protocol]);
272 rtnl_msg_handlers[protocol] = NULL;
274 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
276 static LIST_HEAD(link_ops);
278 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
280 const struct rtnl_link_ops *ops;
282 list_for_each_entry(ops, &link_ops, list) {
283 if (!strcmp(ops->kind, kind))
290 * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
291 * @ops: struct rtnl_link_ops * to register
293 * The caller must hold the rtnl_mutex. This function should be used
294 * by drivers that create devices during module initialization. It
295 * must be called before registering the devices.
297 * Returns 0 on success or a negative error code.
299 int __rtnl_link_register(struct rtnl_link_ops *ops)
301 if (rtnl_link_ops_get(ops->kind))
304 /* The check for setup is here because if ops
305 * does not have that filled up, it is not possible
306 * to use the ops for creating device. So do not
307 * fill up dellink as well. That disables rtnl_dellink.
309 if (ops->setup && !ops->dellink)
310 ops->dellink = unregister_netdevice_queue;
312 list_add_tail(&ops->list, &link_ops);
315 EXPORT_SYMBOL_GPL(__rtnl_link_register);
318 * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
319 * @ops: struct rtnl_link_ops * to register
321 * Returns 0 on success or a negative error code.
323 int rtnl_link_register(struct rtnl_link_ops *ops)
328 err = __rtnl_link_register(ops);
332 EXPORT_SYMBOL_GPL(rtnl_link_register);
334 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
336 struct net_device *dev;
337 LIST_HEAD(list_kill);
339 for_each_netdev(net, dev) {
340 if (dev->rtnl_link_ops == ops)
341 ops->dellink(dev, &list_kill);
343 unregister_netdevice_many(&list_kill);
347 * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
348 * @ops: struct rtnl_link_ops * to unregister
350 * The caller must hold the rtnl_mutex.
352 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
357 __rtnl_kill_links(net, ops);
359 list_del(&ops->list);
361 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
363 /* Return with the rtnl_lock held when there are no network
364 * devices unregistering in any network namespace.
366 static void rtnl_lock_unregistering_all(void)
370 DEFINE_WAIT_FUNC(wait, woken_wake_function);
372 add_wait_queue(&netdev_unregistering_wq, &wait);
374 unregistering = false;
377 if (net->dev_unreg_count > 0) {
378 unregistering = true;
386 wait_woken(&wait, TASK_UNINTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
388 remove_wait_queue(&netdev_unregistering_wq, &wait);
392 * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
393 * @ops: struct rtnl_link_ops * to unregister
395 void rtnl_link_unregister(struct rtnl_link_ops *ops)
397 /* Close the race with cleanup_net() */
398 mutex_lock(&net_mutex);
399 rtnl_lock_unregistering_all();
400 __rtnl_link_unregister(ops);
402 mutex_unlock(&net_mutex);
404 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
406 static size_t rtnl_link_get_slave_info_data_size(const struct net_device *dev)
408 struct net_device *master_dev;
409 const struct rtnl_link_ops *ops;
411 master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
414 ops = master_dev->rtnl_link_ops;
415 if (!ops || !ops->get_slave_size)
417 /* IFLA_INFO_SLAVE_DATA + nested data */
418 return nla_total_size(sizeof(struct nlattr)) +
419 ops->get_slave_size(master_dev, dev);
422 static size_t rtnl_link_get_size(const struct net_device *dev)
424 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
430 size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
431 nla_total_size(strlen(ops->kind) + 1); /* IFLA_INFO_KIND */
434 /* IFLA_INFO_DATA + nested data */
435 size += nla_total_size(sizeof(struct nlattr)) +
438 if (ops->get_xstats_size)
439 /* IFLA_INFO_XSTATS */
440 size += nla_total_size(ops->get_xstats_size(dev));
442 size += rtnl_link_get_slave_info_data_size(dev);
447 static LIST_HEAD(rtnl_af_ops);
449 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
451 const struct rtnl_af_ops *ops;
453 list_for_each_entry(ops, &rtnl_af_ops, list) {
454 if (ops->family == family)
462 * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
463 * @ops: struct rtnl_af_ops * to register
465 * Returns 0 on success or a negative error code.
467 void rtnl_af_register(struct rtnl_af_ops *ops)
470 list_add_tail(&ops->list, &rtnl_af_ops);
473 EXPORT_SYMBOL_GPL(rtnl_af_register);
476 * __rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
477 * @ops: struct rtnl_af_ops * to unregister
479 * The caller must hold the rtnl_mutex.
481 void __rtnl_af_unregister(struct rtnl_af_ops *ops)
483 list_del(&ops->list);
485 EXPORT_SYMBOL_GPL(__rtnl_af_unregister);
488 * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
489 * @ops: struct rtnl_af_ops * to unregister
491 void rtnl_af_unregister(struct rtnl_af_ops *ops)
494 __rtnl_af_unregister(ops);
497 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
499 static size_t rtnl_link_get_af_size(const struct net_device *dev)
501 struct rtnl_af_ops *af_ops;
505 size = nla_total_size(sizeof(struct nlattr));
507 list_for_each_entry(af_ops, &rtnl_af_ops, list) {
508 if (af_ops->get_link_af_size) {
509 /* AF_* + nested data */
510 size += nla_total_size(sizeof(struct nlattr)) +
511 af_ops->get_link_af_size(dev);
518 static bool rtnl_have_link_slave_info(const struct net_device *dev)
520 struct net_device *master_dev;
522 master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
523 if (master_dev && master_dev->rtnl_link_ops)
528 static int rtnl_link_slave_info_fill(struct sk_buff *skb,
529 const struct net_device *dev)
531 struct net_device *master_dev;
532 const struct rtnl_link_ops *ops;
533 struct nlattr *slave_data;
536 master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
539 ops = master_dev->rtnl_link_ops;
542 if (nla_put_string(skb, IFLA_INFO_SLAVE_KIND, ops->kind) < 0)
544 if (ops->fill_slave_info) {
545 slave_data = nla_nest_start(skb, IFLA_INFO_SLAVE_DATA);
548 err = ops->fill_slave_info(skb, master_dev, dev);
550 goto err_cancel_slave_data;
551 nla_nest_end(skb, slave_data);
555 err_cancel_slave_data:
556 nla_nest_cancel(skb, slave_data);
560 static int rtnl_link_info_fill(struct sk_buff *skb,
561 const struct net_device *dev)
563 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
569 if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
571 if (ops->fill_xstats) {
572 err = ops->fill_xstats(skb, dev);
576 if (ops->fill_info) {
577 data = nla_nest_start(skb, IFLA_INFO_DATA);
580 err = ops->fill_info(skb, dev);
582 goto err_cancel_data;
583 nla_nest_end(skb, data);
588 nla_nest_cancel(skb, data);
592 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
594 struct nlattr *linkinfo;
597 linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
598 if (linkinfo == NULL)
601 err = rtnl_link_info_fill(skb, dev);
603 goto err_cancel_link;
605 err = rtnl_link_slave_info_fill(skb, dev);
607 goto err_cancel_link;
609 nla_nest_end(skb, linkinfo);
613 nla_nest_cancel(skb, linkinfo);
618 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
620 struct sock *rtnl = net->rtnl;
623 NETLINK_CB(skb).dst_group = group;
625 atomic_inc(&skb->users);
626 netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
628 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
632 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
634 struct sock *rtnl = net->rtnl;
636 return nlmsg_unicast(rtnl, skb, pid);
638 EXPORT_SYMBOL(rtnl_unicast);
640 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
641 struct nlmsghdr *nlh, gfp_t flags)
643 struct sock *rtnl = net->rtnl;
647 report = nlmsg_report(nlh);
649 nlmsg_notify(rtnl, skb, pid, group, report, flags);
651 EXPORT_SYMBOL(rtnl_notify);
653 void rtnl_set_sk_err(struct net *net, u32 group, int error)
655 struct sock *rtnl = net->rtnl;
657 netlink_set_err(rtnl, 0, group, error);
659 EXPORT_SYMBOL(rtnl_set_sk_err);
661 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
666 mx = nla_nest_start(skb, RTA_METRICS);
670 for (i = 0; i < RTAX_MAX; i++) {
673 if (nla_put_u32(skb, i+1, metrics[i]))
674 goto nla_put_failure;
679 nla_nest_cancel(skb, mx);
683 return nla_nest_end(skb, mx);
686 nla_nest_cancel(skb, mx);
689 EXPORT_SYMBOL(rtnetlink_put_metrics);
691 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
692 long expires, u32 error)
694 struct rta_cacheinfo ci = {
695 .rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse),
696 .rta_used = dst->__use,
697 .rta_clntref = atomic_read(&(dst->__refcnt)),
705 clock = jiffies_to_clock_t(abs(expires));
706 clock = min_t(unsigned long, clock, INT_MAX);
707 ci.rta_expires = (expires > 0) ? clock : -clock;
709 return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
711 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
713 static void set_operstate(struct net_device *dev, unsigned char transition)
715 unsigned char operstate = dev->operstate;
717 switch (transition) {
719 if ((operstate == IF_OPER_DORMANT ||
720 operstate == IF_OPER_UNKNOWN) &&
722 operstate = IF_OPER_UP;
725 case IF_OPER_DORMANT:
726 if (operstate == IF_OPER_UP ||
727 operstate == IF_OPER_UNKNOWN)
728 operstate = IF_OPER_DORMANT;
732 if (dev->operstate != operstate) {
733 write_lock_bh(&dev_base_lock);
734 dev->operstate = operstate;
735 write_unlock_bh(&dev_base_lock);
736 netdev_state_change(dev);
740 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
742 return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
743 (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
746 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
747 const struct ifinfomsg *ifm)
749 unsigned int flags = ifm->ifi_flags;
751 /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
753 flags = (flags & ifm->ifi_change) |
754 (rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
759 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
760 const struct rtnl_link_stats64 *b)
762 a->rx_packets = b->rx_packets;
763 a->tx_packets = b->tx_packets;
764 a->rx_bytes = b->rx_bytes;
765 a->tx_bytes = b->tx_bytes;
766 a->rx_errors = b->rx_errors;
767 a->tx_errors = b->tx_errors;
768 a->rx_dropped = b->rx_dropped;
769 a->tx_dropped = b->tx_dropped;
771 a->multicast = b->multicast;
772 a->collisions = b->collisions;
774 a->rx_length_errors = b->rx_length_errors;
775 a->rx_over_errors = b->rx_over_errors;
776 a->rx_crc_errors = b->rx_crc_errors;
777 a->rx_frame_errors = b->rx_frame_errors;
778 a->rx_fifo_errors = b->rx_fifo_errors;
779 a->rx_missed_errors = b->rx_missed_errors;
781 a->tx_aborted_errors = b->tx_aborted_errors;
782 a->tx_carrier_errors = b->tx_carrier_errors;
783 a->tx_fifo_errors = b->tx_fifo_errors;
784 a->tx_heartbeat_errors = b->tx_heartbeat_errors;
785 a->tx_window_errors = b->tx_window_errors;
787 a->rx_compressed = b->rx_compressed;
788 a->tx_compressed = b->tx_compressed;
791 static void copy_rtnl_link_stats64(void *v, const struct rtnl_link_stats64 *b)
793 memcpy(v, b, sizeof(*b));
797 static inline int rtnl_vfinfo_size(const struct net_device *dev,
800 if (dev->dev.parent && dev_is_pci(dev->dev.parent) &&
801 (ext_filter_mask & RTEXT_FILTER_VF)) {
802 int num_vfs = dev_num_vf(dev->dev.parent);
803 size_t size = nla_total_size(sizeof(struct nlattr));
804 size += nla_total_size(num_vfs * sizeof(struct nlattr));
806 (nla_total_size(sizeof(struct ifla_vf_mac)) +
807 nla_total_size(sizeof(struct ifla_vf_vlan)) +
808 nla_total_size(sizeof(struct ifla_vf_spoofchk)) +
809 nla_total_size(sizeof(struct ifla_vf_rate)) +
810 nla_total_size(sizeof(struct ifla_vf_link_state)));
816 static size_t rtnl_port_size(const struct net_device *dev,
819 size_t port_size = nla_total_size(4) /* PORT_VF */
820 + nla_total_size(PORT_PROFILE_MAX) /* PORT_PROFILE */
821 + nla_total_size(sizeof(struct ifla_port_vsi))
823 + nla_total_size(PORT_UUID_MAX) /* PORT_INSTANCE_UUID */
824 + nla_total_size(PORT_UUID_MAX) /* PORT_HOST_UUID */
825 + nla_total_size(1) /* PROT_VDP_REQUEST */
826 + nla_total_size(2); /* PORT_VDP_RESPONSE */
827 size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
828 size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
830 size_t port_self_size = nla_total_size(sizeof(struct nlattr))
833 if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
834 !(ext_filter_mask & RTEXT_FILTER_VF))
836 if (dev_num_vf(dev->dev.parent))
837 return port_self_size + vf_ports_size +
838 vf_port_size * dev_num_vf(dev->dev.parent);
840 return port_self_size;
843 static noinline size_t if_nlmsg_size(const struct net_device *dev,
846 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
847 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
848 + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
849 + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
850 + nla_total_size(sizeof(struct rtnl_link_ifmap))
851 + nla_total_size(sizeof(struct rtnl_link_stats))
852 + nla_total_size(sizeof(struct rtnl_link_stats64))
853 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
854 + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
855 + nla_total_size(4) /* IFLA_TXQLEN */
856 + nla_total_size(4) /* IFLA_WEIGHT */
857 + nla_total_size(4) /* IFLA_MTU */
858 + nla_total_size(4) /* IFLA_LINK */
859 + nla_total_size(4) /* IFLA_MASTER */
860 + nla_total_size(1) /* IFLA_CARRIER */
861 + nla_total_size(4) /* IFLA_PROMISCUITY */
862 + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
863 + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
864 + nla_total_size(1) /* IFLA_OPERSTATE */
865 + nla_total_size(1) /* IFLA_LINKMODE */
866 + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
867 + nla_total_size(ext_filter_mask
868 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
869 + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
870 + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
871 + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
872 + rtnl_link_get_af_size(dev) /* IFLA_AF_SPEC */
873 + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */
874 + nla_total_size(MAX_PHYS_ITEM_ID_LEN); /* IFLA_PHYS_SWITCH_ID */
877 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
879 struct nlattr *vf_ports;
880 struct nlattr *vf_port;
884 vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
888 for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
889 vf_port = nla_nest_start(skb, IFLA_VF_PORT);
891 goto nla_put_failure;
892 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
893 goto nla_put_failure;
894 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
895 if (err == -EMSGSIZE)
896 goto nla_put_failure;
898 nla_nest_cancel(skb, vf_port);
901 nla_nest_end(skb, vf_port);
904 nla_nest_end(skb, vf_ports);
909 nla_nest_cancel(skb, vf_ports);
913 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
915 struct nlattr *port_self;
918 port_self = nla_nest_start(skb, IFLA_PORT_SELF);
922 err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
924 nla_nest_cancel(skb, port_self);
925 return (err == -EMSGSIZE) ? err : 0;
928 nla_nest_end(skb, port_self);
933 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
938 if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
939 !(ext_filter_mask & RTEXT_FILTER_VF))
942 err = rtnl_port_self_fill(skb, dev);
946 if (dev_num_vf(dev->dev.parent)) {
947 err = rtnl_vf_ports_fill(skb, dev);
955 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
958 struct netdev_phys_item_id ppid;
960 err = dev_get_phys_port_id(dev, &ppid);
962 if (err == -EOPNOTSUPP)
967 if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
973 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev)
976 struct netdev_phys_item_id psid;
978 err = netdev_switch_parent_id_get(dev, &psid);
980 if (err == -EOPNOTSUPP)
985 if (nla_put(skb, IFLA_PHYS_SWITCH_ID, psid.id_len, psid.id))
991 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
992 int type, u32 pid, u32 seq, u32 change,
993 unsigned int flags, u32 ext_filter_mask)
995 struct ifinfomsg *ifm;
996 struct nlmsghdr *nlh;
997 struct rtnl_link_stats64 temp;
998 const struct rtnl_link_stats64 *stats;
999 struct nlattr *attr, *af_spec;
1000 struct rtnl_af_ops *af_ops;
1001 struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1004 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
1008 ifm = nlmsg_data(nlh);
1009 ifm->ifi_family = AF_UNSPEC;
1011 ifm->ifi_type = dev->type;
1012 ifm->ifi_index = dev->ifindex;
1013 ifm->ifi_flags = dev_get_flags(dev);
1014 ifm->ifi_change = change;
1016 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
1017 nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
1018 nla_put_u8(skb, IFLA_OPERSTATE,
1019 netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
1020 nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
1021 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
1022 nla_put_u32(skb, IFLA_GROUP, dev->group) ||
1023 nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
1024 nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
1026 nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1028 (dev->ifindex != dev->iflink &&
1029 nla_put_u32(skb, IFLA_LINK, dev->iflink)) ||
1031 nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex)) ||
1032 nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1034 nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
1036 nla_put_string(skb, IFLA_IFALIAS, dev->ifalias)) ||
1037 nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1038 atomic_read(&dev->carrier_changes)))
1039 goto nla_put_failure;
1042 struct rtnl_link_ifmap map = {
1043 .mem_start = dev->mem_start,
1044 .mem_end = dev->mem_end,
1045 .base_addr = dev->base_addr,
1048 .port = dev->if_port,
1050 if (nla_put(skb, IFLA_MAP, sizeof(map), &map))
1051 goto nla_put_failure;
1054 if (dev->addr_len) {
1055 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1056 nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1057 goto nla_put_failure;
1060 if (rtnl_phys_port_id_fill(skb, dev))
1061 goto nla_put_failure;
1063 if (rtnl_phys_switch_id_fill(skb, dev))
1064 goto nla_put_failure;
1066 attr = nla_reserve(skb, IFLA_STATS,
1067 sizeof(struct rtnl_link_stats));
1069 goto nla_put_failure;
1071 stats = dev_get_stats(dev, &temp);
1072 copy_rtnl_link_stats(nla_data(attr), stats);
1074 attr = nla_reserve(skb, IFLA_STATS64,
1075 sizeof(struct rtnl_link_stats64));
1077 goto nla_put_failure;
1078 copy_rtnl_link_stats64(nla_data(attr), stats);
1080 if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF) &&
1081 nla_put_u32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent)))
1082 goto nla_put_failure;
1084 if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent
1085 && (ext_filter_mask & RTEXT_FILTER_VF)) {
1088 struct nlattr *vfinfo, *vf;
1089 int num_vfs = dev_num_vf(dev->dev.parent);
1091 vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
1093 goto nla_put_failure;
1094 for (i = 0; i < num_vfs; i++) {
1095 struct ifla_vf_info ivi;
1096 struct ifla_vf_mac vf_mac;
1097 struct ifla_vf_vlan vf_vlan;
1098 struct ifla_vf_rate vf_rate;
1099 struct ifla_vf_tx_rate vf_tx_rate;
1100 struct ifla_vf_spoofchk vf_spoofchk;
1101 struct ifla_vf_link_state vf_linkstate;
1104 * Not all SR-IOV capable drivers support the
1105 * spoofcheck query. Preset to -1 so the user
1106 * space tool can detect that the driver didn't
1110 memset(ivi.mac, 0, sizeof(ivi.mac));
1111 /* The default value for VF link state is "auto"
1112 * IFLA_VF_LINK_STATE_AUTO which equals zero
1115 if (dev->netdev_ops->ndo_get_vf_config(dev, i, &ivi))
1122 vf_linkstate.vf = ivi.vf;
1124 memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1125 vf_vlan.vlan = ivi.vlan;
1126 vf_vlan.qos = ivi.qos;
1127 vf_tx_rate.rate = ivi.max_tx_rate;
1128 vf_rate.min_tx_rate = ivi.min_tx_rate;
1129 vf_rate.max_tx_rate = ivi.max_tx_rate;
1130 vf_spoofchk.setting = ivi.spoofchk;
1131 vf_linkstate.link_state = ivi.linkstate;
1132 vf = nla_nest_start(skb, IFLA_VF_INFO);
1134 nla_nest_cancel(skb, vfinfo);
1135 goto nla_put_failure;
1137 if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1138 nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1139 nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1141 nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1143 nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1145 nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1147 goto nla_put_failure;
1148 nla_nest_end(skb, vf);
1150 nla_nest_end(skb, vfinfo);
1153 if (rtnl_port_fill(skb, dev, ext_filter_mask))
1154 goto nla_put_failure;
1156 if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1157 if (rtnl_link_fill(skb, dev) < 0)
1158 goto nla_put_failure;
1161 if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1162 goto nla_put_failure;
1164 list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1165 if (af_ops->fill_link_af) {
1169 if (!(af = nla_nest_start(skb, af_ops->family)))
1170 goto nla_put_failure;
1172 err = af_ops->fill_link_af(skb, dev);
1175 * Caller may return ENODATA to indicate that there
1176 * was no data to be dumped. This is not an error, it
1177 * means we should trim the attribute header and
1180 if (err == -ENODATA)
1181 nla_nest_cancel(skb, af);
1183 goto nla_put_failure;
1185 nla_nest_end(skb, af);
1189 nla_nest_end(skb, af_spec);
1191 return nlmsg_end(skb, nlh);
1194 nlmsg_cancel(skb, nlh);
1198 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1199 [IFLA_IFNAME] = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1200 [IFLA_ADDRESS] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1201 [IFLA_BROADCAST] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1202 [IFLA_MAP] = { .len = sizeof(struct rtnl_link_ifmap) },
1203 [IFLA_MTU] = { .type = NLA_U32 },
1204 [IFLA_LINK] = { .type = NLA_U32 },
1205 [IFLA_MASTER] = { .type = NLA_U32 },
1206 [IFLA_CARRIER] = { .type = NLA_U8 },
1207 [IFLA_TXQLEN] = { .type = NLA_U32 },
1208 [IFLA_WEIGHT] = { .type = NLA_U32 },
1209 [IFLA_OPERSTATE] = { .type = NLA_U8 },
1210 [IFLA_LINKMODE] = { .type = NLA_U8 },
1211 [IFLA_LINKINFO] = { .type = NLA_NESTED },
1212 [IFLA_NET_NS_PID] = { .type = NLA_U32 },
1213 [IFLA_NET_NS_FD] = { .type = NLA_U32 },
1214 [IFLA_IFALIAS] = { .type = NLA_STRING, .len = IFALIASZ-1 },
1215 [IFLA_VFINFO_LIST] = {. type = NLA_NESTED },
1216 [IFLA_VF_PORTS] = { .type = NLA_NESTED },
1217 [IFLA_PORT_SELF] = { .type = NLA_NESTED },
1218 [IFLA_AF_SPEC] = { .type = NLA_NESTED },
1219 [IFLA_EXT_MASK] = { .type = NLA_U32 },
1220 [IFLA_PROMISCUITY] = { .type = NLA_U32 },
1221 [IFLA_NUM_TX_QUEUES] = { .type = NLA_U32 },
1222 [IFLA_NUM_RX_QUEUES] = { .type = NLA_U32 },
1223 [IFLA_PHYS_PORT_ID] = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1224 [IFLA_CARRIER_CHANGES] = { .type = NLA_U32 }, /* ignored */
1225 [IFLA_PHYS_SWITCH_ID] = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1228 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1229 [IFLA_INFO_KIND] = { .type = NLA_STRING },
1230 [IFLA_INFO_DATA] = { .type = NLA_NESTED },
1231 [IFLA_INFO_SLAVE_KIND] = { .type = NLA_STRING },
1232 [IFLA_INFO_SLAVE_DATA] = { .type = NLA_NESTED },
1235 static const struct nla_policy ifla_vfinfo_policy[IFLA_VF_INFO_MAX+1] = {
1236 [IFLA_VF_INFO] = { .type = NLA_NESTED },
1239 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1240 [IFLA_VF_MAC] = { .type = NLA_BINARY,
1241 .len = sizeof(struct ifla_vf_mac) },
1242 [IFLA_VF_VLAN] = { .type = NLA_BINARY,
1243 .len = sizeof(struct ifla_vf_vlan) },
1244 [IFLA_VF_TX_RATE] = { .type = NLA_BINARY,
1245 .len = sizeof(struct ifla_vf_tx_rate) },
1246 [IFLA_VF_SPOOFCHK] = { .type = NLA_BINARY,
1247 .len = sizeof(struct ifla_vf_spoofchk) },
1248 [IFLA_VF_RATE] = { .type = NLA_BINARY,
1249 .len = sizeof(struct ifla_vf_rate) },
1250 [IFLA_VF_LINK_STATE] = { .type = NLA_BINARY,
1251 .len = sizeof(struct ifla_vf_link_state) },
1254 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1255 [IFLA_PORT_VF] = { .type = NLA_U32 },
1256 [IFLA_PORT_PROFILE] = { .type = NLA_STRING,
1257 .len = PORT_PROFILE_MAX },
1258 [IFLA_PORT_VSI_TYPE] = { .type = NLA_BINARY,
1259 .len = sizeof(struct ifla_port_vsi)},
1260 [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1261 .len = PORT_UUID_MAX },
1262 [IFLA_PORT_HOST_UUID] = { .type = NLA_STRING,
1263 .len = PORT_UUID_MAX },
1264 [IFLA_PORT_REQUEST] = { .type = NLA_U8, },
1265 [IFLA_PORT_RESPONSE] = { .type = NLA_U16, },
1268 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1270 struct net *net = sock_net(skb->sk);
1273 struct net_device *dev;
1274 struct hlist_head *head;
1275 struct nlattr *tb[IFLA_MAX+1];
1276 u32 ext_filter_mask = 0;
1281 s_idx = cb->args[1];
1284 cb->seq = net->dev_base_seq;
1286 /* A hack to preserve kernel<->userspace interface.
1287 * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
1288 * However, before Linux v3.9 the code here assumed rtgenmsg and that's
1289 * what iproute2 < v3.9.0 used.
1290 * We can detect the old iproute2. Even including the IFLA_EXT_MASK
1291 * attribute, its netlink message is shorter than struct ifinfomsg.
1293 hdrlen = nlmsg_len(cb->nlh) < sizeof(struct ifinfomsg) ?
1294 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
1296 if (nlmsg_parse(cb->nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
1298 if (tb[IFLA_EXT_MASK])
1299 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1302 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1304 head = &net->dev_index_head[h];
1305 hlist_for_each_entry_rcu(dev, head, index_hlist) {
1308 err = rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1309 NETLINK_CB(cb->skb).portid,
1310 cb->nlh->nlmsg_seq, 0,
1313 /* If we ran out of room on the first message,
1316 WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
1321 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1334 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len)
1336 return nla_parse(tb, IFLA_MAX, head, len, ifla_policy);
1338 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
1340 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1343 /* Examine the link attributes and figure out which
1344 * network namespace we are talking about.
1346 if (tb[IFLA_NET_NS_PID])
1347 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1348 else if (tb[IFLA_NET_NS_FD])
1349 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1351 net = get_net(src_net);
1354 EXPORT_SYMBOL(rtnl_link_get_net);
1356 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1359 if (tb[IFLA_ADDRESS] &&
1360 nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1363 if (tb[IFLA_BROADCAST] &&
1364 nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1368 if (tb[IFLA_AF_SPEC]) {
1372 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1373 const struct rtnl_af_ops *af_ops;
1375 if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1376 return -EAFNOSUPPORT;
1378 if (!af_ops->set_link_af)
1381 if (af_ops->validate_link_af) {
1382 err = af_ops->validate_link_af(dev, af);
1392 static int do_setvfinfo(struct net_device *dev, struct nlattr *attr)
1394 int rem, err = -EINVAL;
1396 const struct net_device_ops *ops = dev->netdev_ops;
1398 nla_for_each_nested(vf, attr, rem) {
1399 switch (nla_type(vf)) {
1401 struct ifla_vf_mac *ivm;
1404 if (ops->ndo_set_vf_mac)
1405 err = ops->ndo_set_vf_mac(dev, ivm->vf,
1409 case IFLA_VF_VLAN: {
1410 struct ifla_vf_vlan *ivv;
1413 if (ops->ndo_set_vf_vlan)
1414 err = ops->ndo_set_vf_vlan(dev, ivv->vf,
1419 case IFLA_VF_TX_RATE: {
1420 struct ifla_vf_tx_rate *ivt;
1421 struct ifla_vf_info ivf;
1424 if (ops->ndo_get_vf_config)
1425 err = ops->ndo_get_vf_config(dev, ivt->vf,
1430 if (ops->ndo_set_vf_rate)
1431 err = ops->ndo_set_vf_rate(dev, ivt->vf,
1436 case IFLA_VF_RATE: {
1437 struct ifla_vf_rate *ivt;
1440 if (ops->ndo_set_vf_rate)
1441 err = ops->ndo_set_vf_rate(dev, ivt->vf,
1446 case IFLA_VF_SPOOFCHK: {
1447 struct ifla_vf_spoofchk *ivs;
1450 if (ops->ndo_set_vf_spoofchk)
1451 err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1455 case IFLA_VF_LINK_STATE: {
1456 struct ifla_vf_link_state *ivl;
1459 if (ops->ndo_set_vf_link_state)
1460 err = ops->ndo_set_vf_link_state(dev, ivl->vf,
1474 static int do_set_master(struct net_device *dev, int ifindex)
1476 struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1477 const struct net_device_ops *ops;
1481 if (upper_dev->ifindex == ifindex)
1483 ops = upper_dev->netdev_ops;
1484 if (ops->ndo_del_slave) {
1485 err = ops->ndo_del_slave(upper_dev, dev);
1494 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
1497 ops = upper_dev->netdev_ops;
1498 if (ops->ndo_add_slave) {
1499 err = ops->ndo_add_slave(upper_dev, dev);
1509 #define DO_SETLINK_MODIFIED 0x01
1510 /* notify flag means notify + modified. */
1511 #define DO_SETLINK_NOTIFY 0x03
1512 static int do_setlink(const struct sk_buff *skb,
1513 struct net_device *dev, struct ifinfomsg *ifm,
1514 struct nlattr **tb, char *ifname, int status)
1516 const struct net_device_ops *ops = dev->netdev_ops;
1519 if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1520 struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1525 if (!netlink_ns_capable(skb, net->user_ns, CAP_NET_ADMIN)) {
1530 err = dev_change_net_namespace(dev, net, ifname);
1534 status |= DO_SETLINK_MODIFIED;
1538 struct rtnl_link_ifmap *u_map;
1541 if (!ops->ndo_set_config) {
1546 if (!netif_device_present(dev)) {
1551 u_map = nla_data(tb[IFLA_MAP]);
1552 k_map.mem_start = (unsigned long) u_map->mem_start;
1553 k_map.mem_end = (unsigned long) u_map->mem_end;
1554 k_map.base_addr = (unsigned short) u_map->base_addr;
1555 k_map.irq = (unsigned char) u_map->irq;
1556 k_map.dma = (unsigned char) u_map->dma;
1557 k_map.port = (unsigned char) u_map->port;
1559 err = ops->ndo_set_config(dev, &k_map);
1563 status |= DO_SETLINK_NOTIFY;
1566 if (tb[IFLA_ADDRESS]) {
1567 struct sockaddr *sa;
1570 len = sizeof(sa_family_t) + dev->addr_len;
1571 sa = kmalloc(len, GFP_KERNEL);
1576 sa->sa_family = dev->type;
1577 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1579 err = dev_set_mac_address(dev, sa);
1583 status |= DO_SETLINK_MODIFIED;
1587 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1590 status |= DO_SETLINK_MODIFIED;
1593 if (tb[IFLA_GROUP]) {
1594 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1595 status |= DO_SETLINK_NOTIFY;
1599 * Interface selected by interface index but interface
1600 * name provided implies that a name change has been
1603 if (ifm->ifi_index > 0 && ifname[0]) {
1604 err = dev_change_name(dev, ifname);
1607 status |= DO_SETLINK_MODIFIED;
1610 if (tb[IFLA_IFALIAS]) {
1611 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
1612 nla_len(tb[IFLA_IFALIAS]));
1615 status |= DO_SETLINK_NOTIFY;
1618 if (tb[IFLA_BROADCAST]) {
1619 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
1620 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
1623 if (ifm->ifi_flags || ifm->ifi_change) {
1624 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1629 if (tb[IFLA_MASTER]) {
1630 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
1633 status |= DO_SETLINK_MODIFIED;
1636 if (tb[IFLA_CARRIER]) {
1637 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
1640 status |= DO_SETLINK_MODIFIED;
1643 if (tb[IFLA_TXQLEN]) {
1644 unsigned long value = nla_get_u32(tb[IFLA_TXQLEN]);
1646 if (dev->tx_queue_len ^ value)
1647 status |= DO_SETLINK_NOTIFY;
1649 dev->tx_queue_len = value;
1652 if (tb[IFLA_OPERSTATE])
1653 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1655 if (tb[IFLA_LINKMODE]) {
1656 unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
1658 write_lock_bh(&dev_base_lock);
1659 if (dev->link_mode ^ value)
1660 status |= DO_SETLINK_NOTIFY;
1661 dev->link_mode = value;
1662 write_unlock_bh(&dev_base_lock);
1665 if (tb[IFLA_VFINFO_LIST]) {
1666 struct nlattr *attr;
1668 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
1669 if (nla_type(attr) != IFLA_VF_INFO) {
1673 err = do_setvfinfo(dev, attr);
1676 status |= DO_SETLINK_NOTIFY;
1681 if (tb[IFLA_VF_PORTS]) {
1682 struct nlattr *port[IFLA_PORT_MAX+1];
1683 struct nlattr *attr;
1688 if (!ops->ndo_set_vf_port)
1691 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
1692 if (nla_type(attr) != IFLA_VF_PORT)
1694 err = nla_parse_nested(port, IFLA_PORT_MAX,
1695 attr, ifla_port_policy);
1698 if (!port[IFLA_PORT_VF]) {
1702 vf = nla_get_u32(port[IFLA_PORT_VF]);
1703 err = ops->ndo_set_vf_port(dev, vf, port);
1706 status |= DO_SETLINK_NOTIFY;
1711 if (tb[IFLA_PORT_SELF]) {
1712 struct nlattr *port[IFLA_PORT_MAX+1];
1714 err = nla_parse_nested(port, IFLA_PORT_MAX,
1715 tb[IFLA_PORT_SELF], ifla_port_policy);
1720 if (ops->ndo_set_vf_port)
1721 err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
1724 status |= DO_SETLINK_NOTIFY;
1727 if (tb[IFLA_AF_SPEC]) {
1731 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1732 const struct rtnl_af_ops *af_ops;
1734 if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1737 err = af_ops->set_link_af(dev, af);
1741 status |= DO_SETLINK_NOTIFY;
1747 if (status & DO_SETLINK_MODIFIED) {
1748 if (status & DO_SETLINK_NOTIFY)
1749 netdev_state_change(dev);
1752 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",
1759 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
1761 struct net *net = sock_net(skb->sk);
1762 struct ifinfomsg *ifm;
1763 struct net_device *dev;
1765 struct nlattr *tb[IFLA_MAX+1];
1766 char ifname[IFNAMSIZ];
1768 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1772 if (tb[IFLA_IFNAME])
1773 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1778 ifm = nlmsg_data(nlh);
1779 if (ifm->ifi_index > 0)
1780 dev = __dev_get_by_index(net, ifm->ifi_index);
1781 else if (tb[IFLA_IFNAME])
1782 dev = __dev_get_by_name(net, ifname);
1791 err = validate_linkmsg(dev, tb);
1795 err = do_setlink(skb, dev, ifm, tb, ifname, 0);
1800 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
1802 struct net *net = sock_net(skb->sk);
1803 const struct rtnl_link_ops *ops;
1804 struct net_device *dev;
1805 struct ifinfomsg *ifm;
1806 char ifname[IFNAMSIZ];
1807 struct nlattr *tb[IFLA_MAX+1];
1809 LIST_HEAD(list_kill);
1811 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1815 if (tb[IFLA_IFNAME])
1816 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1818 ifm = nlmsg_data(nlh);
1819 if (ifm->ifi_index > 0)
1820 dev = __dev_get_by_index(net, ifm->ifi_index);
1821 else if (tb[IFLA_IFNAME])
1822 dev = __dev_get_by_name(net, ifname);
1829 ops = dev->rtnl_link_ops;
1830 if (!ops || !ops->dellink)
1833 ops->dellink(dev, &list_kill);
1834 unregister_netdevice_many(&list_kill);
1838 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
1840 unsigned int old_flags;
1843 old_flags = dev->flags;
1844 if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
1845 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1850 dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
1852 __dev_notify_flags(dev, old_flags, ~0U);
1855 EXPORT_SYMBOL(rtnl_configure_link);
1857 struct net_device *rtnl_create_link(struct net *net,
1858 char *ifname, unsigned char name_assign_type,
1859 const struct rtnl_link_ops *ops, struct nlattr *tb[])
1862 struct net_device *dev;
1863 unsigned int num_tx_queues = 1;
1864 unsigned int num_rx_queues = 1;
1866 if (tb[IFLA_NUM_TX_QUEUES])
1867 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
1868 else if (ops->get_num_tx_queues)
1869 num_tx_queues = ops->get_num_tx_queues();
1871 if (tb[IFLA_NUM_RX_QUEUES])
1872 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
1873 else if (ops->get_num_rx_queues)
1874 num_rx_queues = ops->get_num_rx_queues();
1877 dev = alloc_netdev_mqs(ops->priv_size, ifname, name_assign_type,
1878 ops->setup, num_tx_queues, num_rx_queues);
1882 dev_net_set(dev, net);
1883 dev->rtnl_link_ops = ops;
1884 dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
1887 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
1888 if (tb[IFLA_ADDRESS]) {
1889 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
1890 nla_len(tb[IFLA_ADDRESS]));
1891 dev->addr_assign_type = NET_ADDR_SET;
1893 if (tb[IFLA_BROADCAST])
1894 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
1895 nla_len(tb[IFLA_BROADCAST]));
1896 if (tb[IFLA_TXQLEN])
1897 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1898 if (tb[IFLA_OPERSTATE])
1899 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1900 if (tb[IFLA_LINKMODE])
1901 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1903 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1908 return ERR_PTR(err);
1910 EXPORT_SYMBOL(rtnl_create_link);
1912 static int rtnl_group_changelink(const struct sk_buff *skb,
1913 struct net *net, int group,
1914 struct ifinfomsg *ifm,
1917 struct net_device *dev;
1920 for_each_netdev(net, dev) {
1921 if (dev->group == group) {
1922 err = do_setlink(skb, dev, ifm, tb, NULL, 0);
1931 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh)
1933 struct net *net = sock_net(skb->sk);
1934 const struct rtnl_link_ops *ops;
1935 const struct rtnl_link_ops *m_ops = NULL;
1936 struct net_device *dev;
1937 struct net_device *master_dev = NULL;
1938 struct ifinfomsg *ifm;
1939 char kind[MODULE_NAME_LEN];
1940 char ifname[IFNAMSIZ];
1941 struct nlattr *tb[IFLA_MAX+1];
1942 struct nlattr *linkinfo[IFLA_INFO_MAX+1];
1943 unsigned char name_assign_type = NET_NAME_USER;
1946 #ifdef CONFIG_MODULES
1949 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1953 if (tb[IFLA_IFNAME])
1954 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1958 ifm = nlmsg_data(nlh);
1959 if (ifm->ifi_index > 0)
1960 dev = __dev_get_by_index(net, ifm->ifi_index);
1963 dev = __dev_get_by_name(net, ifname);
1969 master_dev = netdev_master_upper_dev_get(dev);
1971 m_ops = master_dev->rtnl_link_ops;
1974 err = validate_linkmsg(dev, tb);
1978 if (tb[IFLA_LINKINFO]) {
1979 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
1980 tb[IFLA_LINKINFO], ifla_info_policy);
1984 memset(linkinfo, 0, sizeof(linkinfo));
1986 if (linkinfo[IFLA_INFO_KIND]) {
1987 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
1988 ops = rtnl_link_ops_get(kind);
1995 struct nlattr *attr[ops ? ops->maxtype + 1 : 0];
1996 struct nlattr *slave_attr[m_ops ? m_ops->slave_maxtype + 1 : 0];
1997 struct nlattr **data = NULL;
1998 struct nlattr **slave_data = NULL;
1999 struct net *dest_net;
2002 if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
2003 err = nla_parse_nested(attr, ops->maxtype,
2004 linkinfo[IFLA_INFO_DATA],
2010 if (ops->validate) {
2011 err = ops->validate(tb, data);
2018 if (m_ops->slave_maxtype &&
2019 linkinfo[IFLA_INFO_SLAVE_DATA]) {
2020 err = nla_parse_nested(slave_attr,
2021 m_ops->slave_maxtype,
2022 linkinfo[IFLA_INFO_SLAVE_DATA],
2023 m_ops->slave_policy);
2026 slave_data = slave_attr;
2028 if (m_ops->slave_validate) {
2029 err = m_ops->slave_validate(tb, slave_data);
2038 if (nlh->nlmsg_flags & NLM_F_EXCL)
2040 if (nlh->nlmsg_flags & NLM_F_REPLACE)
2043 if (linkinfo[IFLA_INFO_DATA]) {
2044 if (!ops || ops != dev->rtnl_link_ops ||
2048 err = ops->changelink(dev, tb, data);
2051 status |= DO_SETLINK_NOTIFY;
2054 if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
2055 if (!m_ops || !m_ops->slave_changelink)
2058 err = m_ops->slave_changelink(master_dev, dev,
2062 status |= DO_SETLINK_NOTIFY;
2065 return do_setlink(skb, dev, ifm, tb, ifname, status);
2068 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
2069 if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
2070 return rtnl_group_changelink(skb, net,
2071 nla_get_u32(tb[IFLA_GROUP]),
2076 if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
2080 #ifdef CONFIG_MODULES
2083 request_module("rtnl-link-%s", kind);
2085 ops = rtnl_link_ops_get(kind);
2097 snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
2098 name_assign_type = NET_NAME_ENUM;
2101 dest_net = rtnl_link_get_net(net, tb);
2102 if (IS_ERR(dest_net))
2103 return PTR_ERR(dest_net);
2105 dev = rtnl_create_link(dest_net, ifname, name_assign_type, ops, tb);
2111 dev->ifindex = ifm->ifi_index;
2114 err = ops->newlink(net, dev, tb, data);
2115 /* Drivers should call free_netdev() in ->destructor
2116 * and unregister it on failure after registration
2117 * so that device could be finally freed in rtnl_unlock.
2120 /* If device is not registered at all, free it now */
2121 if (dev->reg_state == NETREG_UNINITIALIZED)
2126 err = register_netdevice(dev);
2132 err = rtnl_configure_link(dev, ifm);
2134 unregister_netdevice(dev);
2141 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh)
2143 struct net *net = sock_net(skb->sk);
2144 struct ifinfomsg *ifm;
2145 char ifname[IFNAMSIZ];
2146 struct nlattr *tb[IFLA_MAX+1];
2147 struct net_device *dev = NULL;
2148 struct sk_buff *nskb;
2150 u32 ext_filter_mask = 0;
2152 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2156 if (tb[IFLA_IFNAME])
2157 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2159 if (tb[IFLA_EXT_MASK])
2160 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2162 ifm = nlmsg_data(nlh);
2163 if (ifm->ifi_index > 0)
2164 dev = __dev_get_by_index(net, ifm->ifi_index);
2165 else if (tb[IFLA_IFNAME])
2166 dev = __dev_get_by_name(net, ifname);
2173 nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
2177 err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).portid,
2178 nlh->nlmsg_seq, 0, 0, ext_filter_mask);
2180 /* -EMSGSIZE implies BUG in if_nlmsg_size */
2181 WARN_ON(err == -EMSGSIZE);
2184 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
2189 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
2191 struct net *net = sock_net(skb->sk);
2192 struct net_device *dev;
2193 struct nlattr *tb[IFLA_MAX+1];
2194 u32 ext_filter_mask = 0;
2195 u16 min_ifinfo_dump_size = 0;
2198 /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
2199 hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
2200 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
2202 if (nlmsg_parse(nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
2203 if (tb[IFLA_EXT_MASK])
2204 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2207 if (!ext_filter_mask)
2208 return NLMSG_GOODSIZE;
2210 * traverse the list of net devices and compute the minimum
2211 * buffer size based upon the filter mask.
2213 list_for_each_entry(dev, &net->dev_base_head, dev_list) {
2214 min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
2219 return min_ifinfo_dump_size;
2222 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
2225 int s_idx = cb->family;
2229 for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
2230 int type = cb->nlh->nlmsg_type-RTM_BASE;
2231 if (idx < s_idx || idx == PF_PACKET)
2233 if (rtnl_msg_handlers[idx] == NULL ||
2234 rtnl_msg_handlers[idx][type].dumpit == NULL)
2237 memset(&cb->args[0], 0, sizeof(cb->args));
2241 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
2249 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
2250 unsigned int change, gfp_t flags)
2252 struct net *net = dev_net(dev);
2253 struct sk_buff *skb;
2255 size_t if_info_size;
2257 skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
2261 err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0);
2263 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
2264 WARN_ON(err == -EMSGSIZE);
2271 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2275 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags)
2277 struct net *net = dev_net(dev);
2279 rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
2282 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
2285 struct sk_buff *skb;
2287 skb = rtmsg_ifinfo_build_skb(type, dev, change, flags);
2289 rtmsg_ifinfo_send(skb, dev, flags);
2291 EXPORT_SYMBOL(rtmsg_ifinfo);
2293 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
2294 struct net_device *dev,
2295 u8 *addr, u32 pid, u32 seq,
2296 int type, unsigned int flags,
2299 struct nlmsghdr *nlh;
2302 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
2306 ndm = nlmsg_data(nlh);
2307 ndm->ndm_family = AF_BRIDGE;
2310 ndm->ndm_flags = flags;
2312 ndm->ndm_ifindex = dev->ifindex;
2313 ndm->ndm_state = NUD_PERMANENT;
2315 if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
2316 goto nla_put_failure;
2318 return nlmsg_end(skb, nlh);
2321 nlmsg_cancel(skb, nlh);
2325 static inline size_t rtnl_fdb_nlmsg_size(void)
2327 return NLMSG_ALIGN(sizeof(struct ndmsg)) + nla_total_size(ETH_ALEN);
2330 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, int type)
2332 struct net *net = dev_net(dev);
2333 struct sk_buff *skb;
2336 skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
2340 err = nlmsg_populate_fdb_fill(skb, dev, addr, 0, 0, type, NTF_SELF, 0);
2346 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2349 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
2353 * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
2355 int ndo_dflt_fdb_add(struct ndmsg *ndm,
2356 struct nlattr *tb[],
2357 struct net_device *dev,
2358 const unsigned char *addr, u16 vid,
2363 /* If aging addresses are supported device will need to
2364 * implement its own handler for this.
2366 if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
2367 pr_info("%s: FDB only supports static addresses\n", dev->name);
2372 pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
2376 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2377 err = dev_uc_add_excl(dev, addr);
2378 else if (is_multicast_ether_addr(addr))
2379 err = dev_mc_add_excl(dev, addr);
2381 /* Only return duplicate errors if NLM_F_EXCL is set */
2382 if (err == -EEXIST && !(flags & NLM_F_EXCL))
2387 EXPORT_SYMBOL(ndo_dflt_fdb_add);
2389 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid)
2394 if (nla_len(vlan_attr) != sizeof(u16)) {
2395 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan\n");
2399 vid = nla_get_u16(vlan_attr);
2401 if (!vid || vid >= VLAN_VID_MASK) {
2402 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan id %d\n",
2411 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh)
2413 struct net *net = sock_net(skb->sk);
2415 struct nlattr *tb[NDA_MAX+1];
2416 struct net_device *dev;
2421 err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2425 ndm = nlmsg_data(nlh);
2426 if (ndm->ndm_ifindex == 0) {
2427 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n");
2431 dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2433 pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n");
2437 if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2438 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n");
2442 addr = nla_data(tb[NDA_LLADDR]);
2444 err = fdb_vid_parse(tb[NDA_VLAN], &vid);
2450 /* Support fdb on master device the net/bridge default case */
2451 if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2452 (dev->priv_flags & IFF_BRIDGE_PORT)) {
2453 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2454 const struct net_device_ops *ops = br_dev->netdev_ops;
2456 err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
2461 ndm->ndm_flags &= ~NTF_MASTER;
2464 /* Embedded bridge, macvlan, and any other device support */
2465 if ((ndm->ndm_flags & NTF_SELF)) {
2466 if (dev->netdev_ops->ndo_fdb_add)
2467 err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
2471 err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
2475 rtnl_fdb_notify(dev, addr, RTM_NEWNEIGH);
2476 ndm->ndm_flags &= ~NTF_SELF;
2484 * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
2486 int ndo_dflt_fdb_del(struct ndmsg *ndm,
2487 struct nlattr *tb[],
2488 struct net_device *dev,
2489 const unsigned char *addr, u16 vid)
2493 /* If aging addresses are supported device will need to
2494 * implement its own handler for this.
2496 if (!(ndm->ndm_state & NUD_PERMANENT)) {
2497 pr_info("%s: FDB only supports static addresses\n", dev->name);
2501 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2502 err = dev_uc_del(dev, addr);
2503 else if (is_multicast_ether_addr(addr))
2504 err = dev_mc_del(dev, addr);
2508 EXPORT_SYMBOL(ndo_dflt_fdb_del);
2510 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh)
2512 struct net *net = sock_net(skb->sk);
2514 struct nlattr *tb[NDA_MAX+1];
2515 struct net_device *dev;
2520 if (!netlink_capable(skb, CAP_NET_ADMIN))
2523 err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2527 ndm = nlmsg_data(nlh);
2528 if (ndm->ndm_ifindex == 0) {
2529 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n");
2533 dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2535 pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n");
2539 if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2540 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid address\n");
2544 addr = nla_data(tb[NDA_LLADDR]);
2546 err = fdb_vid_parse(tb[NDA_VLAN], &vid);
2552 /* Support fdb on master device the net/bridge default case */
2553 if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2554 (dev->priv_flags & IFF_BRIDGE_PORT)) {
2555 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2556 const struct net_device_ops *ops = br_dev->netdev_ops;
2558 if (ops->ndo_fdb_del)
2559 err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid);
2564 ndm->ndm_flags &= ~NTF_MASTER;
2567 /* Embedded bridge, macvlan, and any other device support */
2568 if (ndm->ndm_flags & NTF_SELF) {
2569 if (dev->netdev_ops->ndo_fdb_del)
2570 err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr,
2573 err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
2576 rtnl_fdb_notify(dev, addr, RTM_DELNEIGH);
2577 ndm->ndm_flags &= ~NTF_SELF;
2584 static int nlmsg_populate_fdb(struct sk_buff *skb,
2585 struct netlink_callback *cb,
2586 struct net_device *dev,
2588 struct netdev_hw_addr_list *list)
2590 struct netdev_hw_addr *ha;
2594 portid = NETLINK_CB(cb->skb).portid;
2595 seq = cb->nlh->nlmsg_seq;
2597 list_for_each_entry(ha, &list->list, list) {
2598 if (*idx < cb->args[0])
2601 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr,
2603 RTM_NEWNEIGH, NTF_SELF,
2614 * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
2615 * @nlh: netlink message header
2618 * Default netdevice operation to dump the existing unicast address list.
2619 * Returns number of addresses from list put in skb.
2621 int ndo_dflt_fdb_dump(struct sk_buff *skb,
2622 struct netlink_callback *cb,
2623 struct net_device *dev,
2624 struct net_device *filter_dev,
2629 netif_addr_lock_bh(dev);
2630 err = nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->uc);
2633 nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->mc);
2635 netif_addr_unlock_bh(dev);
2638 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
2640 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
2642 struct net_device *dev;
2643 struct nlattr *tb[IFLA_MAX+1];
2644 struct net_device *bdev = NULL;
2645 struct net_device *br_dev = NULL;
2646 const struct net_device_ops *ops = NULL;
2647 const struct net_device_ops *cops = NULL;
2648 struct ifinfomsg *ifm = nlmsg_data(cb->nlh);
2649 struct net *net = sock_net(skb->sk);
2654 if (nlmsg_parse(cb->nlh, sizeof(struct ifinfomsg), tb, IFLA_MAX,
2655 ifla_policy) == 0) {
2656 if (tb[IFLA_MASTER])
2657 br_idx = nla_get_u32(tb[IFLA_MASTER]);
2660 brport_idx = ifm->ifi_index;
2663 br_dev = __dev_get_by_index(net, br_idx);
2667 ops = br_dev->netdev_ops;
2671 for_each_netdev(net, dev) {
2672 if (brport_idx && (dev->ifindex != brport_idx))
2675 if (!br_idx) { /* user did not specify a specific bridge */
2676 if (dev->priv_flags & IFF_BRIDGE_PORT) {
2677 br_dev = netdev_master_upper_dev_get(dev);
2678 cops = br_dev->netdev_ops;
2683 if (dev != br_dev &&
2684 !(dev->priv_flags & IFF_BRIDGE_PORT))
2687 if (br_dev != netdev_master_upper_dev_get(dev) &&
2688 !(dev->priv_flags & IFF_EBRIDGE))
2695 if (dev->priv_flags & IFF_BRIDGE_PORT) {
2696 if (cops && cops->ndo_fdb_dump)
2697 idx = cops->ndo_fdb_dump(skb, cb, br_dev, dev,
2701 idx = ndo_dflt_fdb_dump(skb, cb, dev, NULL, idx);
2702 if (dev->netdev_ops->ndo_fdb_dump)
2703 idx = dev->netdev_ops->ndo_fdb_dump(skb, cb, bdev, dev,
2713 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
2714 unsigned int attrnum, unsigned int flag)
2717 return nla_put_u8(skb, attrnum, !!(flags & flag));
2721 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
2722 struct net_device *dev, u16 mode,
2723 u32 flags, u32 mask)
2725 struct nlmsghdr *nlh;
2726 struct ifinfomsg *ifm;
2727 struct nlattr *br_afspec;
2728 struct nlattr *protinfo;
2729 u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
2730 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2732 nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), NLM_F_MULTI);
2736 ifm = nlmsg_data(nlh);
2737 ifm->ifi_family = AF_BRIDGE;
2739 ifm->ifi_type = dev->type;
2740 ifm->ifi_index = dev->ifindex;
2741 ifm->ifi_flags = dev_get_flags(dev);
2742 ifm->ifi_change = 0;
2745 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
2746 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
2747 nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
2749 nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
2751 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
2752 (dev->ifindex != dev->iflink &&
2753 nla_put_u32(skb, IFLA_LINK, dev->iflink)))
2754 goto nla_put_failure;
2756 br_afspec = nla_nest_start(skb, IFLA_AF_SPEC);
2758 goto nla_put_failure;
2760 if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
2761 nla_nest_cancel(skb, br_afspec);
2762 goto nla_put_failure;
2765 if (mode != BRIDGE_MODE_UNDEF) {
2766 if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
2767 nla_nest_cancel(skb, br_afspec);
2768 goto nla_put_failure;
2771 nla_nest_end(skb, br_afspec);
2773 protinfo = nla_nest_start(skb, IFLA_PROTINFO | NLA_F_NESTED);
2775 goto nla_put_failure;
2777 if (brport_nla_put_flag(skb, flags, mask,
2778 IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
2779 brport_nla_put_flag(skb, flags, mask,
2780 IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
2781 brport_nla_put_flag(skb, flags, mask,
2782 IFLA_BRPORT_FAST_LEAVE,
2783 BR_MULTICAST_FAST_LEAVE) ||
2784 brport_nla_put_flag(skb, flags, mask,
2785 IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
2786 brport_nla_put_flag(skb, flags, mask,
2787 IFLA_BRPORT_LEARNING, BR_LEARNING) ||
2788 brport_nla_put_flag(skb, flags, mask,
2789 IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
2790 brport_nla_put_flag(skb, flags, mask,
2791 IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
2792 brport_nla_put_flag(skb, flags, mask,
2793 IFLA_BRPORT_PROXYARP, BR_PROXYARP)) {
2794 nla_nest_cancel(skb, protinfo);
2795 goto nla_put_failure;
2798 nla_nest_end(skb, protinfo);
2800 return nlmsg_end(skb, nlh);
2802 nlmsg_cancel(skb, nlh);
2805 EXPORT_SYMBOL(ndo_dflt_bridge_getlink);
2807 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
2809 struct net *net = sock_net(skb->sk);
2810 struct net_device *dev;
2812 u32 portid = NETLINK_CB(cb->skb).portid;
2813 u32 seq = cb->nlh->nlmsg_seq;
2814 u32 filter_mask = 0;
2816 if (nlmsg_len(cb->nlh) > sizeof(struct ifinfomsg)) {
2817 struct nlattr *extfilt;
2819 extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct ifinfomsg),
2822 if (nla_len(extfilt) < sizeof(filter_mask))
2825 filter_mask = nla_get_u32(extfilt);
2830 for_each_netdev_rcu(net, dev) {
2831 const struct net_device_ops *ops = dev->netdev_ops;
2832 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2834 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
2835 if (idx >= cb->args[0] &&
2836 br_dev->netdev_ops->ndo_bridge_getlink(
2837 skb, portid, seq, dev, filter_mask) < 0)
2842 if (ops->ndo_bridge_getlink) {
2843 if (idx >= cb->args[0] &&
2844 ops->ndo_bridge_getlink(skb, portid, seq, dev,
2856 static inline size_t bridge_nlmsg_size(void)
2858 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
2859 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
2860 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
2861 + nla_total_size(sizeof(u32)) /* IFLA_MASTER */
2862 + nla_total_size(sizeof(u32)) /* IFLA_MTU */
2863 + nla_total_size(sizeof(u32)) /* IFLA_LINK */
2864 + nla_total_size(sizeof(u32)) /* IFLA_OPERSTATE */
2865 + nla_total_size(sizeof(u8)) /* IFLA_PROTINFO */
2866 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
2867 + nla_total_size(sizeof(u16)) /* IFLA_BRIDGE_FLAGS */
2868 + nla_total_size(sizeof(u16)); /* IFLA_BRIDGE_MODE */
2871 static int rtnl_bridge_notify(struct net_device *dev, u16 flags)
2873 struct net *net = dev_net(dev);
2874 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2875 struct sk_buff *skb;
2876 int err = -EOPNOTSUPP;
2878 skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
2884 if ((!flags || (flags & BRIDGE_FLAGS_MASTER)) &&
2885 br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
2886 err = br_dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0);
2891 if ((flags & BRIDGE_FLAGS_SELF) &&
2892 dev->netdev_ops->ndo_bridge_getlink) {
2893 err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0);
2901 rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
2904 WARN_ON(err == -EMSGSIZE);
2907 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2911 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
2913 struct net *net = sock_net(skb->sk);
2914 struct ifinfomsg *ifm;
2915 struct net_device *dev;
2916 struct nlattr *br_spec, *attr = NULL;
2917 int rem, err = -EOPNOTSUPP;
2918 u16 oflags, flags = 0;
2919 bool have_flags = false;
2921 if (nlmsg_len(nlh) < sizeof(*ifm))
2924 ifm = nlmsg_data(nlh);
2925 if (ifm->ifi_family != AF_BRIDGE)
2926 return -EPFNOSUPPORT;
2928 dev = __dev_get_by_index(net, ifm->ifi_index);
2930 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
2934 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
2936 nla_for_each_nested(attr, br_spec, rem) {
2937 if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
2938 if (nla_len(attr) < sizeof(flags))
2942 flags = nla_get_u16(attr);
2950 if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
2951 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2953 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
2958 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh);
2962 flags &= ~BRIDGE_FLAGS_MASTER;
2965 if ((flags & BRIDGE_FLAGS_SELF)) {
2966 if (!dev->netdev_ops->ndo_bridge_setlink)
2969 err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh);
2972 flags &= ~BRIDGE_FLAGS_SELF;
2976 memcpy(nla_data(attr), &flags, sizeof(flags));
2977 /* Generate event to notify upper layer of bridge change */
2979 err = rtnl_bridge_notify(dev, oflags);
2984 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
2986 struct net *net = sock_net(skb->sk);
2987 struct ifinfomsg *ifm;
2988 struct net_device *dev;
2989 struct nlattr *br_spec, *attr = NULL;
2990 int rem, err = -EOPNOTSUPP;
2991 u16 oflags, flags = 0;
2992 bool have_flags = false;
2994 if (nlmsg_len(nlh) < sizeof(*ifm))
2997 ifm = nlmsg_data(nlh);
2998 if (ifm->ifi_family != AF_BRIDGE)
2999 return -EPFNOSUPPORT;
3001 dev = __dev_get_by_index(net, ifm->ifi_index);
3003 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3007 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3009 nla_for_each_nested(attr, br_spec, rem) {
3010 if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3011 if (nla_len(attr) < sizeof(flags))
3015 flags = nla_get_u16(attr);
3023 if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3024 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3026 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
3031 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh);
3035 flags &= ~BRIDGE_FLAGS_MASTER;
3038 if ((flags & BRIDGE_FLAGS_SELF)) {
3039 if (!dev->netdev_ops->ndo_bridge_dellink)
3042 err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh);
3045 flags &= ~BRIDGE_FLAGS_SELF;
3049 memcpy(nla_data(attr), &flags, sizeof(flags));
3050 /* Generate event to notify upper layer of bridge change */
3052 err = rtnl_bridge_notify(dev, oflags);
3057 /* Process one rtnetlink message. */
3059 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
3061 struct net *net = sock_net(skb->sk);
3062 rtnl_doit_func doit;
3068 type = nlh->nlmsg_type;
3074 /* All the messages must have at least 1 byte length */
3075 if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
3078 family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
3082 if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
3085 if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
3087 rtnl_dumpit_func dumpit;
3088 rtnl_calcit_func calcit;
3089 u16 min_dump_alloc = 0;
3091 dumpit = rtnl_get_dumpit(family, type);
3094 calcit = rtnl_get_calcit(family, type);
3096 min_dump_alloc = calcit(skb, nlh);
3101 struct netlink_dump_control c = {
3103 .min_dump_alloc = min_dump_alloc,
3105 err = netlink_dump_start(rtnl, skb, nlh, &c);
3111 doit = rtnl_get_doit(family, type);
3115 return doit(skb, nlh);
3118 static void rtnetlink_rcv(struct sk_buff *skb)
3121 netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
3125 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
3127 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3133 case NETDEV_POST_INIT:
3134 case NETDEV_REGISTER:
3136 case NETDEV_PRE_TYPE_CHANGE:
3137 case NETDEV_GOING_DOWN:
3138 case NETDEV_UNREGISTER:
3139 case NETDEV_UNREGISTER_FINAL:
3140 case NETDEV_RELEASE:
3144 rtmsg_ifinfo(RTM_NEWLINK, dev, 0, GFP_KERNEL);
3150 static struct notifier_block rtnetlink_dev_notifier = {
3151 .notifier_call = rtnetlink_event,
3155 static int __net_init rtnetlink_net_init(struct net *net)
3158 struct netlink_kernel_cfg cfg = {
3159 .groups = RTNLGRP_MAX,
3160 .input = rtnetlink_rcv,
3161 .cb_mutex = &rtnl_mutex,
3162 .flags = NL_CFG_F_NONROOT_RECV,
3165 sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
3172 static void __net_exit rtnetlink_net_exit(struct net *net)
3174 netlink_kernel_release(net->rtnl);
3178 static struct pernet_operations rtnetlink_net_ops = {
3179 .init = rtnetlink_net_init,
3180 .exit = rtnetlink_net_exit,
3183 void __init rtnetlink_init(void)
3185 if (register_pernet_subsys(&rtnetlink_net_ops))
3186 panic("rtnetlink_init: cannot initialize rtnetlink\n");
3188 register_netdevice_notifier(&rtnetlink_dev_notifier);
3190 rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
3191 rtnl_dump_ifinfo, rtnl_calcit);
3192 rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
3193 rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
3194 rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
3196 rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
3197 rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
3199 rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, NULL);
3200 rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, NULL);
3201 rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, NULL);
3203 rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, NULL);
3204 rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, NULL);
3205 rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, NULL);