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/pci.h>
40 #include <linux/etherdevice.h>
42 #include <asm/uaccess.h>
44 #include <linux/inet.h>
45 #include <linux/netdevice.h>
47 #include <net/protocol.h>
49 #include <net/route.h>
52 #include <net/pkt_sched.h>
53 #include <net/fib_rules.h>
54 #include <net/rtnetlink.h>
55 #include <net/net_namespace.h>
59 rtnl_dumpit_func dumpit;
60 rtnl_calcit_func calcit;
63 static DEFINE_MUTEX(rtnl_mutex);
67 mutex_lock(&rtnl_mutex);
69 EXPORT_SYMBOL(rtnl_lock);
71 void __rtnl_unlock(void)
73 mutex_unlock(&rtnl_mutex);
76 void rtnl_unlock(void)
78 /* This fellow will unlock it for us. */
81 EXPORT_SYMBOL(rtnl_unlock);
83 int rtnl_trylock(void)
85 return mutex_trylock(&rtnl_mutex);
87 EXPORT_SYMBOL(rtnl_trylock);
89 int rtnl_is_locked(void)
91 return mutex_is_locked(&rtnl_mutex);
93 EXPORT_SYMBOL(rtnl_is_locked);
95 #ifdef CONFIG_PROVE_LOCKING
96 int lockdep_rtnl_is_held(void)
98 return lockdep_is_held(&rtnl_mutex);
100 EXPORT_SYMBOL(lockdep_rtnl_is_held);
101 #endif /* #ifdef CONFIG_PROVE_LOCKING */
103 static struct rtnl_link *rtnl_msg_handlers[RTNL_FAMILY_MAX + 1];
105 static inline int rtm_msgindex(int msgtype)
107 int msgindex = msgtype - RTM_BASE;
110 * msgindex < 0 implies someone tried to register a netlink
111 * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
112 * the message type has not been added to linux/rtnetlink.h
114 BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
119 static rtnl_doit_func rtnl_get_doit(int protocol, int msgindex)
121 struct rtnl_link *tab;
123 if (protocol <= RTNL_FAMILY_MAX)
124 tab = rtnl_msg_handlers[protocol];
128 if (tab == NULL || tab[msgindex].doit == NULL)
129 tab = rtnl_msg_handlers[PF_UNSPEC];
131 return tab[msgindex].doit;
134 static rtnl_dumpit_func rtnl_get_dumpit(int protocol, int msgindex)
136 struct rtnl_link *tab;
138 if (protocol <= RTNL_FAMILY_MAX)
139 tab = rtnl_msg_handlers[protocol];
143 if (tab == NULL || tab[msgindex].dumpit == NULL)
144 tab = rtnl_msg_handlers[PF_UNSPEC];
146 return tab[msgindex].dumpit;
149 static rtnl_calcit_func rtnl_get_calcit(int protocol, int msgindex)
151 struct rtnl_link *tab;
153 if (protocol <= RTNL_FAMILY_MAX)
154 tab = rtnl_msg_handlers[protocol];
158 if (tab == NULL || tab[msgindex].calcit == NULL)
159 tab = rtnl_msg_handlers[PF_UNSPEC];
161 return tab[msgindex].calcit;
165 * __rtnl_register - Register a rtnetlink message type
166 * @protocol: Protocol family or PF_UNSPEC
167 * @msgtype: rtnetlink message type
168 * @doit: Function pointer called for each request message
169 * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
170 * @calcit: Function pointer to calc size of dump message
172 * Registers the specified function pointers (at least one of them has
173 * to be non-NULL) to be called whenever a request message for the
174 * specified protocol family and message type is received.
176 * The special protocol family PF_UNSPEC may be used to define fallback
177 * function pointers for the case when no entry for the specific protocol
180 * Returns 0 on success or a negative error code.
182 int __rtnl_register(int protocol, int msgtype,
183 rtnl_doit_func doit, rtnl_dumpit_func dumpit,
184 rtnl_calcit_func calcit)
186 struct rtnl_link *tab;
189 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
190 msgindex = rtm_msgindex(msgtype);
192 tab = rtnl_msg_handlers[protocol];
194 tab = kcalloc(RTM_NR_MSGTYPES, sizeof(*tab), GFP_KERNEL);
198 rtnl_msg_handlers[protocol] = tab;
202 tab[msgindex].doit = doit;
205 tab[msgindex].dumpit = dumpit;
208 tab[msgindex].calcit = calcit;
212 EXPORT_SYMBOL_GPL(__rtnl_register);
215 * rtnl_register - Register a rtnetlink message type
217 * Identical to __rtnl_register() but panics on failure. This is useful
218 * as failure of this function is very unlikely, it can only happen due
219 * to lack of memory when allocating the chain to store all message
220 * handlers for a protocol. Meant for use in init functions where lack
221 * of memory implies no sense in continuing.
223 void rtnl_register(int protocol, int msgtype,
224 rtnl_doit_func doit, rtnl_dumpit_func dumpit,
225 rtnl_calcit_func calcit)
227 if (__rtnl_register(protocol, msgtype, doit, dumpit, calcit) < 0)
228 panic("Unable to register rtnetlink message handler, "
229 "protocol = %d, message type = %d\n",
232 EXPORT_SYMBOL_GPL(rtnl_register);
235 * rtnl_unregister - Unregister a rtnetlink message type
236 * @protocol: Protocol family or PF_UNSPEC
237 * @msgtype: rtnetlink message type
239 * Returns 0 on success or a negative error code.
241 int rtnl_unregister(int protocol, int msgtype)
245 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
246 msgindex = rtm_msgindex(msgtype);
248 if (rtnl_msg_handlers[protocol] == NULL)
251 rtnl_msg_handlers[protocol][msgindex].doit = NULL;
252 rtnl_msg_handlers[protocol][msgindex].dumpit = NULL;
256 EXPORT_SYMBOL_GPL(rtnl_unregister);
259 * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
260 * @protocol : Protocol family or PF_UNSPEC
262 * Identical to calling rtnl_unregster() for all registered message types
263 * of a certain protocol family.
265 void rtnl_unregister_all(int protocol)
267 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
269 kfree(rtnl_msg_handlers[protocol]);
270 rtnl_msg_handlers[protocol] = NULL;
272 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
274 static LIST_HEAD(link_ops);
276 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
278 const struct rtnl_link_ops *ops;
280 list_for_each_entry(ops, &link_ops, list) {
281 if (!strcmp(ops->kind, kind))
288 * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
289 * @ops: struct rtnl_link_ops * to register
291 * The caller must hold the rtnl_mutex. This function should be used
292 * by drivers that create devices during module initialization. It
293 * must be called before registering the devices.
295 * Returns 0 on success or a negative error code.
297 int __rtnl_link_register(struct rtnl_link_ops *ops)
299 if (rtnl_link_ops_get(ops->kind))
303 ops->dellink = unregister_netdevice_queue;
305 list_add_tail(&ops->list, &link_ops);
308 EXPORT_SYMBOL_GPL(__rtnl_link_register);
311 * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
312 * @ops: struct rtnl_link_ops * to register
314 * Returns 0 on success or a negative error code.
316 int rtnl_link_register(struct rtnl_link_ops *ops)
321 err = __rtnl_link_register(ops);
325 EXPORT_SYMBOL_GPL(rtnl_link_register);
327 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
329 struct net_device *dev;
330 LIST_HEAD(list_kill);
332 for_each_netdev(net, dev) {
333 if (dev->rtnl_link_ops == ops)
334 ops->dellink(dev, &list_kill);
336 unregister_netdevice_many(&list_kill);
340 * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
341 * @ops: struct rtnl_link_ops * to unregister
343 * The caller must hold the rtnl_mutex.
345 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
350 __rtnl_kill_links(net, ops);
352 list_del(&ops->list);
354 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
356 /* Return with the rtnl_lock held when there are no network
357 * devices unregistering in any network namespace.
359 static void rtnl_lock_unregistering_all(void)
366 prepare_to_wait(&netdev_unregistering_wq, &wait,
367 TASK_UNINTERRUPTIBLE);
368 unregistering = false;
371 if (net->dev_unreg_count > 0) {
372 unregistering = true;
381 finish_wait(&netdev_unregistering_wq, &wait);
385 * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
386 * @ops: struct rtnl_link_ops * to unregister
388 void rtnl_link_unregister(struct rtnl_link_ops *ops)
390 /* Close the race with cleanup_net() */
391 mutex_lock(&net_mutex);
392 rtnl_lock_unregistering_all();
393 __rtnl_link_unregister(ops);
395 mutex_unlock(&net_mutex);
397 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
399 static size_t rtnl_link_get_slave_info_data_size(const struct net_device *dev)
401 struct net_device *master_dev;
402 const struct rtnl_link_ops *ops;
404 master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
407 ops = master_dev->rtnl_link_ops;
408 if (!ops || !ops->get_slave_size)
410 /* IFLA_INFO_SLAVE_DATA + nested data */
411 return nla_total_size(sizeof(struct nlattr)) +
412 ops->get_slave_size(master_dev, dev);
415 static size_t rtnl_link_get_size(const struct net_device *dev)
417 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
423 size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
424 nla_total_size(strlen(ops->kind) + 1); /* IFLA_INFO_KIND */
427 /* IFLA_INFO_DATA + nested data */
428 size += nla_total_size(sizeof(struct nlattr)) +
431 if (ops->get_xstats_size)
432 /* IFLA_INFO_XSTATS */
433 size += nla_total_size(ops->get_xstats_size(dev));
435 size += rtnl_link_get_slave_info_data_size(dev);
440 static LIST_HEAD(rtnl_af_ops);
442 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
444 const struct rtnl_af_ops *ops;
446 list_for_each_entry(ops, &rtnl_af_ops, list) {
447 if (ops->family == family)
455 * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
456 * @ops: struct rtnl_af_ops * to register
458 * Returns 0 on success or a negative error code.
460 void rtnl_af_register(struct rtnl_af_ops *ops)
463 list_add_tail(&ops->list, &rtnl_af_ops);
466 EXPORT_SYMBOL_GPL(rtnl_af_register);
469 * __rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
470 * @ops: struct rtnl_af_ops * to unregister
472 * The caller must hold the rtnl_mutex.
474 void __rtnl_af_unregister(struct rtnl_af_ops *ops)
476 list_del(&ops->list);
478 EXPORT_SYMBOL_GPL(__rtnl_af_unregister);
481 * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
482 * @ops: struct rtnl_af_ops * to unregister
484 void rtnl_af_unregister(struct rtnl_af_ops *ops)
487 __rtnl_af_unregister(ops);
490 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
492 static size_t rtnl_link_get_af_size(const struct net_device *dev)
494 struct rtnl_af_ops *af_ops;
498 size = nla_total_size(sizeof(struct nlattr));
500 list_for_each_entry(af_ops, &rtnl_af_ops, list) {
501 if (af_ops->get_link_af_size) {
502 /* AF_* + nested data */
503 size += nla_total_size(sizeof(struct nlattr)) +
504 af_ops->get_link_af_size(dev);
511 static bool rtnl_have_link_slave_info(const struct net_device *dev)
513 struct net_device *master_dev;
515 master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
516 if (master_dev && master_dev->rtnl_link_ops)
521 static int rtnl_link_slave_info_fill(struct sk_buff *skb,
522 const struct net_device *dev)
524 struct net_device *master_dev;
525 const struct rtnl_link_ops *ops;
526 struct nlattr *slave_data;
529 master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
532 ops = master_dev->rtnl_link_ops;
535 if (nla_put_string(skb, IFLA_INFO_SLAVE_KIND, ops->kind) < 0)
537 if (ops->fill_slave_info) {
538 slave_data = nla_nest_start(skb, IFLA_INFO_SLAVE_DATA);
541 err = ops->fill_slave_info(skb, master_dev, dev);
543 goto err_cancel_slave_data;
544 nla_nest_end(skb, slave_data);
548 err_cancel_slave_data:
549 nla_nest_cancel(skb, slave_data);
553 static int rtnl_link_info_fill(struct sk_buff *skb,
554 const struct net_device *dev)
556 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
562 if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
564 if (ops->fill_xstats) {
565 err = ops->fill_xstats(skb, dev);
569 if (ops->fill_info) {
570 data = nla_nest_start(skb, IFLA_INFO_DATA);
573 err = ops->fill_info(skb, dev);
575 goto err_cancel_data;
576 nla_nest_end(skb, data);
581 nla_nest_cancel(skb, data);
585 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
587 struct nlattr *linkinfo;
590 linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
591 if (linkinfo == NULL)
594 err = rtnl_link_info_fill(skb, dev);
596 goto err_cancel_link;
598 err = rtnl_link_slave_info_fill(skb, dev);
600 goto err_cancel_link;
602 nla_nest_end(skb, linkinfo);
606 nla_nest_cancel(skb, linkinfo);
611 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
613 struct sock *rtnl = net->rtnl;
616 NETLINK_CB(skb).dst_group = group;
618 atomic_inc(&skb->users);
619 netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
621 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
625 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
627 struct sock *rtnl = net->rtnl;
629 return nlmsg_unicast(rtnl, skb, pid);
631 EXPORT_SYMBOL(rtnl_unicast);
633 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
634 struct nlmsghdr *nlh, gfp_t flags)
636 struct sock *rtnl = net->rtnl;
640 report = nlmsg_report(nlh);
642 nlmsg_notify(rtnl, skb, pid, group, report, flags);
644 EXPORT_SYMBOL(rtnl_notify);
646 void rtnl_set_sk_err(struct net *net, u32 group, int error)
648 struct sock *rtnl = net->rtnl;
650 netlink_set_err(rtnl, 0, group, error);
652 EXPORT_SYMBOL(rtnl_set_sk_err);
654 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
659 mx = nla_nest_start(skb, RTA_METRICS);
663 for (i = 0; i < RTAX_MAX; i++) {
666 if (nla_put_u32(skb, i+1, metrics[i]))
667 goto nla_put_failure;
672 nla_nest_cancel(skb, mx);
676 return nla_nest_end(skb, mx);
679 nla_nest_cancel(skb, mx);
682 EXPORT_SYMBOL(rtnetlink_put_metrics);
684 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
685 long expires, u32 error)
687 struct rta_cacheinfo ci = {
688 .rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse),
689 .rta_used = dst->__use,
690 .rta_clntref = atomic_read(&(dst->__refcnt)),
698 clock = jiffies_to_clock_t(abs(expires));
699 clock = min_t(unsigned long, clock, INT_MAX);
700 ci.rta_expires = (expires > 0) ? clock : -clock;
702 return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
704 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
706 static void set_operstate(struct net_device *dev, unsigned char transition)
708 unsigned char operstate = dev->operstate;
710 switch (transition) {
712 if ((operstate == IF_OPER_DORMANT ||
713 operstate == IF_OPER_UNKNOWN) &&
715 operstate = IF_OPER_UP;
718 case IF_OPER_DORMANT:
719 if (operstate == IF_OPER_UP ||
720 operstate == IF_OPER_UNKNOWN)
721 operstate = IF_OPER_DORMANT;
725 if (dev->operstate != operstate) {
726 write_lock_bh(&dev_base_lock);
727 dev->operstate = operstate;
728 write_unlock_bh(&dev_base_lock);
729 netdev_state_change(dev);
733 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
735 return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
736 (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
739 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
740 const struct ifinfomsg *ifm)
742 unsigned int flags = ifm->ifi_flags;
744 /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
746 flags = (flags & ifm->ifi_change) |
747 (rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
752 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
753 const struct rtnl_link_stats64 *b)
755 a->rx_packets = b->rx_packets;
756 a->tx_packets = b->tx_packets;
757 a->rx_bytes = b->rx_bytes;
758 a->tx_bytes = b->tx_bytes;
759 a->rx_errors = b->rx_errors;
760 a->tx_errors = b->tx_errors;
761 a->rx_dropped = b->rx_dropped;
762 a->tx_dropped = b->tx_dropped;
764 a->multicast = b->multicast;
765 a->collisions = b->collisions;
767 a->rx_length_errors = b->rx_length_errors;
768 a->rx_over_errors = b->rx_over_errors;
769 a->rx_crc_errors = b->rx_crc_errors;
770 a->rx_frame_errors = b->rx_frame_errors;
771 a->rx_fifo_errors = b->rx_fifo_errors;
772 a->rx_missed_errors = b->rx_missed_errors;
774 a->tx_aborted_errors = b->tx_aborted_errors;
775 a->tx_carrier_errors = b->tx_carrier_errors;
776 a->tx_fifo_errors = b->tx_fifo_errors;
777 a->tx_heartbeat_errors = b->tx_heartbeat_errors;
778 a->tx_window_errors = b->tx_window_errors;
780 a->rx_compressed = b->rx_compressed;
781 a->tx_compressed = b->tx_compressed;
784 static void copy_rtnl_link_stats64(void *v, const struct rtnl_link_stats64 *b)
786 memcpy(v, b, sizeof(*b));
790 static inline int rtnl_vfinfo_size(const struct net_device *dev,
793 if (dev->dev.parent && dev_is_pci(dev->dev.parent) &&
794 (ext_filter_mask & RTEXT_FILTER_VF)) {
795 int num_vfs = dev_num_vf(dev->dev.parent);
796 size_t size = nla_total_size(sizeof(struct nlattr));
797 size += nla_total_size(num_vfs * sizeof(struct nlattr));
799 (nla_total_size(sizeof(struct ifla_vf_mac)) +
800 nla_total_size(sizeof(struct ifla_vf_vlan)) +
801 nla_total_size(sizeof(struct ifla_vf_tx_rate)) +
802 nla_total_size(sizeof(struct ifla_vf_spoofchk)));
808 static size_t rtnl_port_size(const struct net_device *dev,
811 size_t port_size = nla_total_size(4) /* PORT_VF */
812 + nla_total_size(PORT_PROFILE_MAX) /* PORT_PROFILE */
813 + nla_total_size(sizeof(struct ifla_port_vsi))
815 + nla_total_size(PORT_UUID_MAX) /* PORT_INSTANCE_UUID */
816 + nla_total_size(PORT_UUID_MAX) /* PORT_HOST_UUID */
817 + nla_total_size(1) /* PROT_VDP_REQUEST */
818 + nla_total_size(2); /* PORT_VDP_RESPONSE */
819 size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
820 size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
822 size_t port_self_size = nla_total_size(sizeof(struct nlattr))
825 if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
826 !(ext_filter_mask & RTEXT_FILTER_VF))
828 if (dev_num_vf(dev->dev.parent))
829 return port_self_size + vf_ports_size +
830 vf_port_size * dev_num_vf(dev->dev.parent);
832 return port_self_size;
835 static noinline size_t if_nlmsg_size(const struct net_device *dev,
838 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
839 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
840 + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
841 + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
842 + nla_total_size(sizeof(struct rtnl_link_ifmap))
843 + nla_total_size(sizeof(struct rtnl_link_stats))
844 + nla_total_size(sizeof(struct rtnl_link_stats64))
845 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
846 + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
847 + nla_total_size(4) /* IFLA_TXQLEN */
848 + nla_total_size(4) /* IFLA_WEIGHT */
849 + nla_total_size(4) /* IFLA_MTU */
850 + nla_total_size(4) /* IFLA_LINK */
851 + nla_total_size(4) /* IFLA_MASTER */
852 + nla_total_size(1) /* IFLA_CARRIER */
853 + nla_total_size(4) /* IFLA_PROMISCUITY */
854 + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
855 + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
856 + nla_total_size(1) /* IFLA_OPERSTATE */
857 + nla_total_size(1) /* IFLA_LINKMODE */
858 + nla_total_size(ext_filter_mask
859 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
860 + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
861 + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
862 + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
863 + rtnl_link_get_af_size(dev) /* IFLA_AF_SPEC */
864 + nla_total_size(MAX_PHYS_PORT_ID_LEN); /* IFLA_PHYS_PORT_ID */
867 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
869 struct nlattr *vf_ports;
870 struct nlattr *vf_port;
874 vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
878 for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
879 vf_port = nla_nest_start(skb, IFLA_VF_PORT);
881 goto nla_put_failure;
882 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
883 goto nla_put_failure;
884 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
885 if (err == -EMSGSIZE)
886 goto nla_put_failure;
888 nla_nest_cancel(skb, vf_port);
891 nla_nest_end(skb, vf_port);
894 nla_nest_end(skb, vf_ports);
899 nla_nest_cancel(skb, vf_ports);
903 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
905 struct nlattr *port_self;
908 port_self = nla_nest_start(skb, IFLA_PORT_SELF);
912 err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
914 nla_nest_cancel(skb, port_self);
915 return (err == -EMSGSIZE) ? err : 0;
918 nla_nest_end(skb, port_self);
923 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
928 if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
929 !(ext_filter_mask & RTEXT_FILTER_VF))
932 err = rtnl_port_self_fill(skb, dev);
936 if (dev_num_vf(dev->dev.parent)) {
937 err = rtnl_vf_ports_fill(skb, dev);
945 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
948 struct netdev_phys_port_id ppid;
950 err = dev_get_phys_port_id(dev, &ppid);
952 if (err == -EOPNOTSUPP)
957 if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
963 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
964 int type, u32 pid, u32 seq, u32 change,
965 unsigned int flags, u32 ext_filter_mask)
967 struct ifinfomsg *ifm;
968 struct nlmsghdr *nlh;
969 struct rtnl_link_stats64 temp;
970 const struct rtnl_link_stats64 *stats;
971 struct nlattr *attr, *af_spec;
972 struct rtnl_af_ops *af_ops;
973 struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
976 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
980 ifm = nlmsg_data(nlh);
981 ifm->ifi_family = AF_UNSPEC;
983 ifm->ifi_type = dev->type;
984 ifm->ifi_index = dev->ifindex;
985 ifm->ifi_flags = dev_get_flags(dev);
986 ifm->ifi_change = change;
988 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
989 nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
990 nla_put_u8(skb, IFLA_OPERSTATE,
991 netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
992 nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
993 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
994 nla_put_u32(skb, IFLA_GROUP, dev->group) ||
995 nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
996 nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
998 nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1000 (dev->ifindex != dev->iflink &&
1001 nla_put_u32(skb, IFLA_LINK, dev->iflink)) ||
1003 nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex)) ||
1004 nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1006 nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
1008 nla_put_string(skb, IFLA_IFALIAS, dev->ifalias)))
1009 goto nla_put_failure;
1012 struct rtnl_link_ifmap map = {
1013 .mem_start = dev->mem_start,
1014 .mem_end = dev->mem_end,
1015 .base_addr = dev->base_addr,
1018 .port = dev->if_port,
1020 if (nla_put(skb, IFLA_MAP, sizeof(map), &map))
1021 goto nla_put_failure;
1024 if (dev->addr_len) {
1025 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1026 nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1027 goto nla_put_failure;
1030 if (rtnl_phys_port_id_fill(skb, dev))
1031 goto nla_put_failure;
1033 attr = nla_reserve(skb, IFLA_STATS,
1034 sizeof(struct rtnl_link_stats));
1036 goto nla_put_failure;
1038 stats = dev_get_stats(dev, &temp);
1039 copy_rtnl_link_stats(nla_data(attr), stats);
1041 attr = nla_reserve(skb, IFLA_STATS64,
1042 sizeof(struct rtnl_link_stats64));
1044 goto nla_put_failure;
1045 copy_rtnl_link_stats64(nla_data(attr), stats);
1047 if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF) &&
1048 nla_put_u32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent)))
1049 goto nla_put_failure;
1051 if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent
1052 && (ext_filter_mask & RTEXT_FILTER_VF)) {
1055 struct nlattr *vfinfo, *vf;
1056 int num_vfs = dev_num_vf(dev->dev.parent);
1058 vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
1060 goto nla_put_failure;
1061 for (i = 0; i < num_vfs; i++) {
1062 struct ifla_vf_info ivi;
1063 struct ifla_vf_mac vf_mac;
1064 struct ifla_vf_vlan vf_vlan;
1065 struct ifla_vf_tx_rate vf_tx_rate;
1066 struct ifla_vf_spoofchk vf_spoofchk;
1067 struct ifla_vf_link_state vf_linkstate;
1070 * Not all SR-IOV capable drivers support the
1071 * spoofcheck query. Preset to -1 so the user
1072 * space tool can detect that the driver didn't
1076 memset(ivi.mac, 0, sizeof(ivi.mac));
1077 /* The default value for VF link state is "auto"
1078 * IFLA_VF_LINK_STATE_AUTO which equals zero
1081 if (dev->netdev_ops->ndo_get_vf_config(dev, i, &ivi))
1087 vf_linkstate.vf = ivi.vf;
1089 memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1090 vf_vlan.vlan = ivi.vlan;
1091 vf_vlan.qos = ivi.qos;
1092 vf_tx_rate.rate = ivi.tx_rate;
1093 vf_spoofchk.setting = ivi.spoofchk;
1094 vf_linkstate.link_state = ivi.linkstate;
1095 vf = nla_nest_start(skb, IFLA_VF_INFO);
1097 nla_nest_cancel(skb, vfinfo);
1098 goto nla_put_failure;
1100 if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1101 nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1102 nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1104 nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1106 nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1108 goto nla_put_failure;
1109 nla_nest_end(skb, vf);
1111 nla_nest_end(skb, vfinfo);
1114 if (rtnl_port_fill(skb, dev, ext_filter_mask))
1115 goto nla_put_failure;
1117 if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1118 if (rtnl_link_fill(skb, dev) < 0)
1119 goto nla_put_failure;
1122 if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1123 goto nla_put_failure;
1125 list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1126 if (af_ops->fill_link_af) {
1130 if (!(af = nla_nest_start(skb, af_ops->family)))
1131 goto nla_put_failure;
1133 err = af_ops->fill_link_af(skb, dev);
1136 * Caller may return ENODATA to indicate that there
1137 * was no data to be dumped. This is not an error, it
1138 * means we should trim the attribute header and
1141 if (err == -ENODATA)
1142 nla_nest_cancel(skb, af);
1144 goto nla_put_failure;
1146 nla_nest_end(skb, af);
1150 nla_nest_end(skb, af_spec);
1152 return nlmsg_end(skb, nlh);
1155 nlmsg_cancel(skb, nlh);
1159 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1161 struct net *net = sock_net(skb->sk);
1164 struct net_device *dev;
1165 struct hlist_head *head;
1166 struct nlattr *tb[IFLA_MAX+1];
1167 u32 ext_filter_mask = 0;
1172 s_idx = cb->args[1];
1175 cb->seq = net->dev_base_seq;
1177 /* A hack to preserve kernel<->userspace interface.
1178 * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
1179 * However, before Linux v3.9 the code here assumed rtgenmsg and that's
1180 * what iproute2 < v3.9.0 used.
1181 * We can detect the old iproute2. Even including the IFLA_EXT_MASK
1182 * attribute, its netlink message is shorter than struct ifinfomsg.
1184 hdrlen = nlmsg_len(cb->nlh) < sizeof(struct ifinfomsg) ?
1185 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
1187 if (nlmsg_parse(cb->nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
1189 if (tb[IFLA_EXT_MASK])
1190 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1193 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1195 head = &net->dev_index_head[h];
1196 hlist_for_each_entry_rcu(dev, head, index_hlist) {
1199 err = rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1200 NETLINK_CB(cb->skb).portid,
1201 cb->nlh->nlmsg_seq, 0,
1204 /* If we ran out of room on the first message,
1207 WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
1212 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1225 const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1226 [IFLA_IFNAME] = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1227 [IFLA_ADDRESS] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1228 [IFLA_BROADCAST] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1229 [IFLA_MAP] = { .len = sizeof(struct rtnl_link_ifmap) },
1230 [IFLA_MTU] = { .type = NLA_U32 },
1231 [IFLA_LINK] = { .type = NLA_U32 },
1232 [IFLA_MASTER] = { .type = NLA_U32 },
1233 [IFLA_CARRIER] = { .type = NLA_U8 },
1234 [IFLA_TXQLEN] = { .type = NLA_U32 },
1235 [IFLA_WEIGHT] = { .type = NLA_U32 },
1236 [IFLA_OPERSTATE] = { .type = NLA_U8 },
1237 [IFLA_LINKMODE] = { .type = NLA_U8 },
1238 [IFLA_LINKINFO] = { .type = NLA_NESTED },
1239 [IFLA_NET_NS_PID] = { .type = NLA_U32 },
1240 [IFLA_NET_NS_FD] = { .type = NLA_U32 },
1241 [IFLA_IFALIAS] = { .type = NLA_STRING, .len = IFALIASZ-1 },
1242 [IFLA_VFINFO_LIST] = {. type = NLA_NESTED },
1243 [IFLA_VF_PORTS] = { .type = NLA_NESTED },
1244 [IFLA_PORT_SELF] = { .type = NLA_NESTED },
1245 [IFLA_AF_SPEC] = { .type = NLA_NESTED },
1246 [IFLA_EXT_MASK] = { .type = NLA_U32 },
1247 [IFLA_PROMISCUITY] = { .type = NLA_U32 },
1248 [IFLA_NUM_TX_QUEUES] = { .type = NLA_U32 },
1249 [IFLA_NUM_RX_QUEUES] = { .type = NLA_U32 },
1250 [IFLA_PHYS_PORT_ID] = { .type = NLA_BINARY, .len = MAX_PHYS_PORT_ID_LEN },
1252 EXPORT_SYMBOL(ifla_policy);
1254 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1255 [IFLA_INFO_KIND] = { .type = NLA_STRING },
1256 [IFLA_INFO_DATA] = { .type = NLA_NESTED },
1257 [IFLA_INFO_SLAVE_KIND] = { .type = NLA_STRING },
1258 [IFLA_INFO_SLAVE_DATA] = { .type = NLA_NESTED },
1261 static const struct nla_policy ifla_vfinfo_policy[IFLA_VF_INFO_MAX+1] = {
1262 [IFLA_VF_INFO] = { .type = NLA_NESTED },
1265 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1266 [IFLA_VF_MAC] = { .type = NLA_BINARY,
1267 .len = sizeof(struct ifla_vf_mac) },
1268 [IFLA_VF_VLAN] = { .type = NLA_BINARY,
1269 .len = sizeof(struct ifla_vf_vlan) },
1270 [IFLA_VF_TX_RATE] = { .type = NLA_BINARY,
1271 .len = sizeof(struct ifla_vf_tx_rate) },
1272 [IFLA_VF_SPOOFCHK] = { .type = NLA_BINARY,
1273 .len = sizeof(struct ifla_vf_spoofchk) },
1276 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1277 [IFLA_PORT_VF] = { .type = NLA_U32 },
1278 [IFLA_PORT_PROFILE] = { .type = NLA_STRING,
1279 .len = PORT_PROFILE_MAX },
1280 [IFLA_PORT_VSI_TYPE] = { .type = NLA_BINARY,
1281 .len = sizeof(struct ifla_port_vsi)},
1282 [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1283 .len = PORT_UUID_MAX },
1284 [IFLA_PORT_HOST_UUID] = { .type = NLA_STRING,
1285 .len = PORT_UUID_MAX },
1286 [IFLA_PORT_REQUEST] = { .type = NLA_U8, },
1287 [IFLA_PORT_RESPONSE] = { .type = NLA_U16, },
1290 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1293 /* Examine the link attributes and figure out which
1294 * network namespace we are talking about.
1296 if (tb[IFLA_NET_NS_PID])
1297 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1298 else if (tb[IFLA_NET_NS_FD])
1299 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1301 net = get_net(src_net);
1304 EXPORT_SYMBOL(rtnl_link_get_net);
1306 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1309 if (tb[IFLA_ADDRESS] &&
1310 nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1313 if (tb[IFLA_BROADCAST] &&
1314 nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1318 if (tb[IFLA_AF_SPEC]) {
1322 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1323 const struct rtnl_af_ops *af_ops;
1325 if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1326 return -EAFNOSUPPORT;
1328 if (!af_ops->set_link_af)
1331 if (af_ops->validate_link_af) {
1332 err = af_ops->validate_link_af(dev, af);
1342 static int do_setvfinfo(struct net_device *dev, struct nlattr *attr)
1344 int rem, err = -EINVAL;
1346 const struct net_device_ops *ops = dev->netdev_ops;
1348 nla_for_each_nested(vf, attr, rem) {
1349 switch (nla_type(vf)) {
1351 struct ifla_vf_mac *ivm;
1354 if (ops->ndo_set_vf_mac)
1355 err = ops->ndo_set_vf_mac(dev, ivm->vf,
1359 case IFLA_VF_VLAN: {
1360 struct ifla_vf_vlan *ivv;
1363 if (ops->ndo_set_vf_vlan)
1364 err = ops->ndo_set_vf_vlan(dev, ivv->vf,
1369 case IFLA_VF_TX_RATE: {
1370 struct ifla_vf_tx_rate *ivt;
1373 if (ops->ndo_set_vf_tx_rate)
1374 err = ops->ndo_set_vf_tx_rate(dev, ivt->vf,
1378 case IFLA_VF_SPOOFCHK: {
1379 struct ifla_vf_spoofchk *ivs;
1382 if (ops->ndo_set_vf_spoofchk)
1383 err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1387 case IFLA_VF_LINK_STATE: {
1388 struct ifla_vf_link_state *ivl;
1391 if (ops->ndo_set_vf_link_state)
1392 err = ops->ndo_set_vf_link_state(dev, ivl->vf,
1406 static int do_set_master(struct net_device *dev, int ifindex)
1408 struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1409 const struct net_device_ops *ops;
1413 if (upper_dev->ifindex == ifindex)
1415 ops = upper_dev->netdev_ops;
1416 if (ops->ndo_del_slave) {
1417 err = ops->ndo_del_slave(upper_dev, dev);
1426 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
1429 ops = upper_dev->netdev_ops;
1430 if (ops->ndo_add_slave) {
1431 err = ops->ndo_add_slave(upper_dev, dev);
1441 static int do_setlink(const struct sk_buff *skb,
1442 struct net_device *dev, struct ifinfomsg *ifm,
1443 struct nlattr **tb, char *ifname, int modified)
1445 const struct net_device_ops *ops = dev->netdev_ops;
1448 if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1449 struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1454 if (!netlink_ns_capable(skb, net->user_ns, CAP_NET_ADMIN)) {
1458 err = dev_change_net_namespace(dev, net, ifname);
1466 struct rtnl_link_ifmap *u_map;
1469 if (!ops->ndo_set_config) {
1474 if (!netif_device_present(dev)) {
1479 u_map = nla_data(tb[IFLA_MAP]);
1480 k_map.mem_start = (unsigned long) u_map->mem_start;
1481 k_map.mem_end = (unsigned long) u_map->mem_end;
1482 k_map.base_addr = (unsigned short) u_map->base_addr;
1483 k_map.irq = (unsigned char) u_map->irq;
1484 k_map.dma = (unsigned char) u_map->dma;
1485 k_map.port = (unsigned char) u_map->port;
1487 err = ops->ndo_set_config(dev, &k_map);
1494 if (tb[IFLA_ADDRESS]) {
1495 struct sockaddr *sa;
1498 len = sizeof(sa_family_t) + dev->addr_len;
1499 sa = kmalloc(len, GFP_KERNEL);
1504 sa->sa_family = dev->type;
1505 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1507 err = dev_set_mac_address(dev, sa);
1515 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1521 if (tb[IFLA_GROUP]) {
1522 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1527 * Interface selected by interface index but interface
1528 * name provided implies that a name change has been
1531 if (ifm->ifi_index > 0 && ifname[0]) {
1532 err = dev_change_name(dev, ifname);
1538 if (tb[IFLA_IFALIAS]) {
1539 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
1540 nla_len(tb[IFLA_IFALIAS]));
1546 if (tb[IFLA_BROADCAST]) {
1547 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
1548 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
1551 if (ifm->ifi_flags || ifm->ifi_change) {
1552 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1557 if (tb[IFLA_MASTER]) {
1558 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
1564 if (tb[IFLA_CARRIER]) {
1565 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
1571 if (tb[IFLA_TXQLEN])
1572 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1574 if (tb[IFLA_OPERSTATE])
1575 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1577 if (tb[IFLA_LINKMODE]) {
1578 write_lock_bh(&dev_base_lock);
1579 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1580 write_unlock_bh(&dev_base_lock);
1583 if (tb[IFLA_VFINFO_LIST]) {
1584 struct nlattr *attr;
1586 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
1587 if (nla_type(attr) != IFLA_VF_INFO) {
1591 err = do_setvfinfo(dev, attr);
1599 if (tb[IFLA_VF_PORTS]) {
1600 struct nlattr *port[IFLA_PORT_MAX+1];
1601 struct nlattr *attr;
1606 if (!ops->ndo_set_vf_port)
1609 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
1610 if (nla_type(attr) != IFLA_VF_PORT)
1612 err = nla_parse_nested(port, IFLA_PORT_MAX,
1613 attr, ifla_port_policy);
1616 if (!port[IFLA_PORT_VF]) {
1620 vf = nla_get_u32(port[IFLA_PORT_VF]);
1621 err = ops->ndo_set_vf_port(dev, vf, port);
1629 if (tb[IFLA_PORT_SELF]) {
1630 struct nlattr *port[IFLA_PORT_MAX+1];
1632 err = nla_parse_nested(port, IFLA_PORT_MAX,
1633 tb[IFLA_PORT_SELF], ifla_port_policy);
1638 if (ops->ndo_set_vf_port)
1639 err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
1645 if (tb[IFLA_AF_SPEC]) {
1649 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1650 const struct rtnl_af_ops *af_ops;
1652 if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1655 err = af_ops->set_link_af(dev, af);
1665 if (err < 0 && modified)
1666 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",
1672 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
1674 struct net *net = sock_net(skb->sk);
1675 struct ifinfomsg *ifm;
1676 struct net_device *dev;
1678 struct nlattr *tb[IFLA_MAX+1];
1679 char ifname[IFNAMSIZ];
1681 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1685 if (tb[IFLA_IFNAME])
1686 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1691 ifm = nlmsg_data(nlh);
1692 if (ifm->ifi_index > 0)
1693 dev = __dev_get_by_index(net, ifm->ifi_index);
1694 else if (tb[IFLA_IFNAME])
1695 dev = __dev_get_by_name(net, ifname);
1704 err = validate_linkmsg(dev, tb);
1708 err = do_setlink(skb, dev, ifm, tb, ifname, 0);
1713 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
1715 struct net *net = sock_net(skb->sk);
1716 const struct rtnl_link_ops *ops;
1717 struct net_device *dev;
1718 struct ifinfomsg *ifm;
1719 char ifname[IFNAMSIZ];
1720 struct nlattr *tb[IFLA_MAX+1];
1722 LIST_HEAD(list_kill);
1724 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1728 if (tb[IFLA_IFNAME])
1729 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1731 ifm = nlmsg_data(nlh);
1732 if (ifm->ifi_index > 0)
1733 dev = __dev_get_by_index(net, ifm->ifi_index);
1734 else if (tb[IFLA_IFNAME])
1735 dev = __dev_get_by_name(net, ifname);
1742 ops = dev->rtnl_link_ops;
1746 ops->dellink(dev, &list_kill);
1747 unregister_netdevice_many(&list_kill);
1751 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
1753 unsigned int old_flags;
1756 old_flags = dev->flags;
1757 if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
1758 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1763 dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
1765 __dev_notify_flags(dev, old_flags, ~0U);
1768 EXPORT_SYMBOL(rtnl_configure_link);
1770 struct net_device *rtnl_create_link(struct net *net,
1771 char *ifname, const struct rtnl_link_ops *ops, struct nlattr *tb[])
1774 struct net_device *dev;
1775 unsigned int num_tx_queues = 1;
1776 unsigned int num_rx_queues = 1;
1778 if (tb[IFLA_NUM_TX_QUEUES])
1779 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
1780 else if (ops->get_num_tx_queues)
1781 num_tx_queues = ops->get_num_tx_queues();
1783 if (tb[IFLA_NUM_RX_QUEUES])
1784 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
1785 else if (ops->get_num_rx_queues)
1786 num_rx_queues = ops->get_num_rx_queues();
1789 dev = alloc_netdev_mqs(ops->priv_size, ifname, ops->setup,
1790 num_tx_queues, num_rx_queues);
1794 dev_net_set(dev, net);
1795 dev->rtnl_link_ops = ops;
1796 dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
1799 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
1800 if (tb[IFLA_ADDRESS]) {
1801 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
1802 nla_len(tb[IFLA_ADDRESS]));
1803 dev->addr_assign_type = NET_ADDR_SET;
1805 if (tb[IFLA_BROADCAST])
1806 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
1807 nla_len(tb[IFLA_BROADCAST]));
1808 if (tb[IFLA_TXQLEN])
1809 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1810 if (tb[IFLA_OPERSTATE])
1811 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1812 if (tb[IFLA_LINKMODE])
1813 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1815 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1820 return ERR_PTR(err);
1822 EXPORT_SYMBOL(rtnl_create_link);
1824 static int rtnl_group_changelink(const struct sk_buff *skb,
1825 struct net *net, int group,
1826 struct ifinfomsg *ifm,
1829 struct net_device *dev;
1832 for_each_netdev(net, dev) {
1833 if (dev->group == group) {
1834 err = do_setlink(skb, dev, ifm, tb, NULL, 0);
1843 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh)
1845 struct net *net = sock_net(skb->sk);
1846 const struct rtnl_link_ops *ops;
1847 const struct rtnl_link_ops *m_ops = NULL;
1848 struct net_device *dev;
1849 struct net_device *master_dev = NULL;
1850 struct ifinfomsg *ifm;
1851 char kind[MODULE_NAME_LEN];
1852 char ifname[IFNAMSIZ];
1853 struct nlattr *tb[IFLA_MAX+1];
1854 struct nlattr *linkinfo[IFLA_INFO_MAX+1];
1857 #ifdef CONFIG_MODULES
1860 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1864 if (tb[IFLA_IFNAME])
1865 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1869 ifm = nlmsg_data(nlh);
1870 if (ifm->ifi_index > 0)
1871 dev = __dev_get_by_index(net, ifm->ifi_index);
1874 dev = __dev_get_by_name(net, ifname);
1880 master_dev = netdev_master_upper_dev_get(dev);
1882 m_ops = master_dev->rtnl_link_ops;
1885 err = validate_linkmsg(dev, tb);
1889 if (tb[IFLA_LINKINFO]) {
1890 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
1891 tb[IFLA_LINKINFO], ifla_info_policy);
1895 memset(linkinfo, 0, sizeof(linkinfo));
1897 if (linkinfo[IFLA_INFO_KIND]) {
1898 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
1899 ops = rtnl_link_ops_get(kind);
1906 struct nlattr *attr[ops ? ops->maxtype + 1 : 0];
1907 struct nlattr *slave_attr[m_ops ? m_ops->slave_maxtype + 1 : 0];
1908 struct nlattr **data = NULL;
1909 struct nlattr **slave_data = NULL;
1910 struct net *dest_net;
1913 if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
1914 err = nla_parse_nested(attr, ops->maxtype,
1915 linkinfo[IFLA_INFO_DATA],
1921 if (ops->validate) {
1922 err = ops->validate(tb, data);
1929 if (m_ops->slave_maxtype &&
1930 linkinfo[IFLA_INFO_SLAVE_DATA]) {
1931 err = nla_parse_nested(slave_attr,
1932 m_ops->slave_maxtype,
1933 linkinfo[IFLA_INFO_SLAVE_DATA],
1934 m_ops->slave_policy);
1937 slave_data = slave_attr;
1939 if (m_ops->slave_validate) {
1940 err = m_ops->slave_validate(tb, slave_data);
1949 if (nlh->nlmsg_flags & NLM_F_EXCL)
1951 if (nlh->nlmsg_flags & NLM_F_REPLACE)
1954 if (linkinfo[IFLA_INFO_DATA]) {
1955 if (!ops || ops != dev->rtnl_link_ops ||
1959 err = ops->changelink(dev, tb, data);
1965 if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
1966 if (!m_ops || !m_ops->slave_changelink)
1969 err = m_ops->slave_changelink(master_dev, dev,
1976 return do_setlink(skb, dev, ifm, tb, ifname, modified);
1979 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
1980 if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
1981 return rtnl_group_changelink(skb, net,
1982 nla_get_u32(tb[IFLA_GROUP]),
1987 if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
1991 #ifdef CONFIG_MODULES
1994 request_module("rtnl-link-%s", kind);
1996 ops = rtnl_link_ops_get(kind);
2005 snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
2007 dest_net = rtnl_link_get_net(net, tb);
2008 if (IS_ERR(dest_net))
2009 return PTR_ERR(dest_net);
2011 dev = rtnl_create_link(dest_net, ifname, ops, tb);
2017 dev->ifindex = ifm->ifi_index;
2020 err = ops->newlink(net, dev, tb, data);
2021 /* Drivers should call free_netdev() in ->destructor
2022 * and unregister it on failure so that device could be
2023 * finally freed in rtnl_unlock.
2028 err = register_netdevice(dev);
2034 err = rtnl_configure_link(dev, ifm);
2036 unregister_netdevice(dev);
2043 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh)
2045 struct net *net = sock_net(skb->sk);
2046 struct ifinfomsg *ifm;
2047 char ifname[IFNAMSIZ];
2048 struct nlattr *tb[IFLA_MAX+1];
2049 struct net_device *dev = NULL;
2050 struct sk_buff *nskb;
2052 u32 ext_filter_mask = 0;
2054 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2058 if (tb[IFLA_IFNAME])
2059 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2061 if (tb[IFLA_EXT_MASK])
2062 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2064 ifm = nlmsg_data(nlh);
2065 if (ifm->ifi_index > 0)
2066 dev = __dev_get_by_index(net, ifm->ifi_index);
2067 else if (tb[IFLA_IFNAME])
2068 dev = __dev_get_by_name(net, ifname);
2075 nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
2079 err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).portid,
2080 nlh->nlmsg_seq, 0, 0, ext_filter_mask);
2082 /* -EMSGSIZE implies BUG in if_nlmsg_size */
2083 WARN_ON(err == -EMSGSIZE);
2086 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
2091 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
2093 struct net *net = sock_net(skb->sk);
2094 struct net_device *dev;
2095 struct nlattr *tb[IFLA_MAX+1];
2096 u32 ext_filter_mask = 0;
2097 u16 min_ifinfo_dump_size = 0;
2100 /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
2101 hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
2102 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
2104 if (nlmsg_parse(nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
2105 if (tb[IFLA_EXT_MASK])
2106 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2109 if (!ext_filter_mask)
2110 return NLMSG_GOODSIZE;
2112 * traverse the list of net devices and compute the minimum
2113 * buffer size based upon the filter mask.
2115 list_for_each_entry(dev, &net->dev_base_head, dev_list) {
2116 min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
2121 return min_ifinfo_dump_size;
2124 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
2127 int s_idx = cb->family;
2131 for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
2132 int type = cb->nlh->nlmsg_type-RTM_BASE;
2133 if (idx < s_idx || idx == PF_PACKET)
2135 if (rtnl_msg_handlers[idx] == NULL ||
2136 rtnl_msg_handlers[idx][type].dumpit == NULL)
2139 memset(&cb->args[0], 0, sizeof(cb->args));
2143 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
2151 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
2154 struct net *net = dev_net(dev);
2155 struct sk_buff *skb;
2157 size_t if_info_size;
2159 skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
2163 err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0);
2165 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
2166 WARN_ON(err == -EMSGSIZE);
2170 rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
2174 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2176 EXPORT_SYMBOL(rtmsg_ifinfo);
2178 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
2179 struct net_device *dev,
2180 u8 *addr, u32 pid, u32 seq,
2181 int type, unsigned int flags,
2184 struct nlmsghdr *nlh;
2187 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
2191 ndm = nlmsg_data(nlh);
2192 ndm->ndm_family = AF_BRIDGE;
2195 ndm->ndm_flags = flags;
2197 ndm->ndm_ifindex = dev->ifindex;
2198 ndm->ndm_state = NUD_PERMANENT;
2200 if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
2201 goto nla_put_failure;
2203 return nlmsg_end(skb, nlh);
2206 nlmsg_cancel(skb, nlh);
2210 static inline size_t rtnl_fdb_nlmsg_size(void)
2212 return NLMSG_ALIGN(sizeof(struct ndmsg)) + nla_total_size(ETH_ALEN);
2215 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, int type)
2217 struct net *net = dev_net(dev);
2218 struct sk_buff *skb;
2221 skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
2225 err = nlmsg_populate_fdb_fill(skb, dev, addr, 0, 0, type, NTF_SELF, 0);
2231 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2234 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
2238 * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
2240 int ndo_dflt_fdb_add(struct ndmsg *ndm,
2241 struct nlattr *tb[],
2242 struct net_device *dev,
2243 const unsigned char *addr,
2248 /* If aging addresses are supported device will need to
2249 * implement its own handler for this.
2251 if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
2252 pr_info("%s: FDB only supports static addresses\n", dev->name);
2256 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2257 err = dev_uc_add_excl(dev, addr);
2258 else if (is_multicast_ether_addr(addr))
2259 err = dev_mc_add_excl(dev, addr);
2261 /* Only return duplicate errors if NLM_F_EXCL is set */
2262 if (err == -EEXIST && !(flags & NLM_F_EXCL))
2267 EXPORT_SYMBOL(ndo_dflt_fdb_add);
2269 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh)
2271 struct net *net = sock_net(skb->sk);
2273 struct nlattr *tb[NDA_MAX+1];
2274 struct net_device *dev;
2278 err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2282 ndm = nlmsg_data(nlh);
2283 if (ndm->ndm_ifindex == 0) {
2284 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n");
2288 dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2290 pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n");
2294 if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2295 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n");
2299 addr = nla_data(tb[NDA_LLADDR]);
2303 /* Support fdb on master device the net/bridge default case */
2304 if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2305 (dev->priv_flags & IFF_BRIDGE_PORT)) {
2306 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2307 const struct net_device_ops *ops = br_dev->netdev_ops;
2309 err = ops->ndo_fdb_add(ndm, tb, dev, addr, nlh->nlmsg_flags);
2313 ndm->ndm_flags &= ~NTF_MASTER;
2316 /* Embedded bridge, macvlan, and any other device support */
2317 if ((ndm->ndm_flags & NTF_SELF)) {
2318 if (dev->netdev_ops->ndo_fdb_add)
2319 err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
2322 err = ndo_dflt_fdb_add(ndm, tb, dev, addr,
2326 rtnl_fdb_notify(dev, addr, RTM_NEWNEIGH);
2327 ndm->ndm_flags &= ~NTF_SELF;
2335 * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
2337 int ndo_dflt_fdb_del(struct ndmsg *ndm,
2338 struct nlattr *tb[],
2339 struct net_device *dev,
2340 const unsigned char *addr)
2342 int err = -EOPNOTSUPP;
2344 /* If aging addresses are supported device will need to
2345 * implement its own handler for this.
2347 if (!(ndm->ndm_state & NUD_PERMANENT)) {
2348 pr_info("%s: FDB only supports static addresses\n", dev->name);
2352 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2353 err = dev_uc_del(dev, addr);
2354 else if (is_multicast_ether_addr(addr))
2355 err = dev_mc_del(dev, addr);
2361 EXPORT_SYMBOL(ndo_dflt_fdb_del);
2363 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh)
2365 struct net *net = sock_net(skb->sk);
2367 struct nlattr *tb[NDA_MAX+1];
2368 struct net_device *dev;
2372 if (!netlink_capable(skb, CAP_NET_ADMIN))
2375 err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2379 ndm = nlmsg_data(nlh);
2380 if (ndm->ndm_ifindex == 0) {
2381 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n");
2385 dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2387 pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n");
2391 if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2392 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid address\n");
2396 addr = nla_data(tb[NDA_LLADDR]);
2400 /* Support fdb on master device the net/bridge default case */
2401 if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2402 (dev->priv_flags & IFF_BRIDGE_PORT)) {
2403 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2404 const struct net_device_ops *ops = br_dev->netdev_ops;
2406 if (ops->ndo_fdb_del)
2407 err = ops->ndo_fdb_del(ndm, tb, dev, addr);
2412 ndm->ndm_flags &= ~NTF_MASTER;
2415 /* Embedded bridge, macvlan, and any other device support */
2416 if (ndm->ndm_flags & NTF_SELF) {
2417 if (dev->netdev_ops->ndo_fdb_del)
2418 err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr);
2420 err = ndo_dflt_fdb_del(ndm, tb, dev, addr);
2423 rtnl_fdb_notify(dev, addr, RTM_DELNEIGH);
2424 ndm->ndm_flags &= ~NTF_SELF;
2431 static int nlmsg_populate_fdb(struct sk_buff *skb,
2432 struct netlink_callback *cb,
2433 struct net_device *dev,
2435 struct netdev_hw_addr_list *list)
2437 struct netdev_hw_addr *ha;
2441 portid = NETLINK_CB(cb->skb).portid;
2442 seq = cb->nlh->nlmsg_seq;
2444 list_for_each_entry(ha, &list->list, list) {
2445 if (*idx < cb->args[0])
2448 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr,
2450 RTM_NEWNEIGH, NTF_SELF,
2461 * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
2462 * @nlh: netlink message header
2465 * Default netdevice operation to dump the existing unicast address list.
2466 * Returns number of addresses from list put in skb.
2468 int ndo_dflt_fdb_dump(struct sk_buff *skb,
2469 struct netlink_callback *cb,
2470 struct net_device *dev,
2475 netif_addr_lock_bh(dev);
2476 err = nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->uc);
2479 nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->mc);
2481 netif_addr_unlock_bh(dev);
2484 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
2486 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
2489 struct net *net = sock_net(skb->sk);
2490 struct net_device *dev;
2493 for_each_netdev_rcu(net, dev) {
2494 if (dev->priv_flags & IFF_BRIDGE_PORT) {
2495 struct net_device *br_dev;
2496 const struct net_device_ops *ops;
2498 br_dev = netdev_master_upper_dev_get(dev);
2499 ops = br_dev->netdev_ops;
2500 if (ops->ndo_fdb_dump)
2501 idx = ops->ndo_fdb_dump(skb, cb, dev, idx);
2504 if (dev->netdev_ops->ndo_fdb_dump)
2505 idx = dev->netdev_ops->ndo_fdb_dump(skb, cb, dev, idx);
2507 idx = ndo_dflt_fdb_dump(skb, cb, dev, idx);
2515 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
2516 struct net_device *dev, u16 mode)
2518 struct nlmsghdr *nlh;
2519 struct ifinfomsg *ifm;
2520 struct nlattr *br_afspec;
2521 u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
2522 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2524 nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), NLM_F_MULTI);
2528 ifm = nlmsg_data(nlh);
2529 ifm->ifi_family = AF_BRIDGE;
2531 ifm->ifi_type = dev->type;
2532 ifm->ifi_index = dev->ifindex;
2533 ifm->ifi_flags = dev_get_flags(dev);
2534 ifm->ifi_change = 0;
2537 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
2538 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
2539 nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
2541 nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
2543 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
2544 (dev->ifindex != dev->iflink &&
2545 nla_put_u32(skb, IFLA_LINK, dev->iflink)))
2546 goto nla_put_failure;
2548 br_afspec = nla_nest_start(skb, IFLA_AF_SPEC);
2550 goto nla_put_failure;
2552 if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF) ||
2553 nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
2554 nla_nest_cancel(skb, br_afspec);
2555 goto nla_put_failure;
2557 nla_nest_end(skb, br_afspec);
2559 return nlmsg_end(skb, nlh);
2561 nlmsg_cancel(skb, nlh);
2564 EXPORT_SYMBOL(ndo_dflt_bridge_getlink);
2566 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
2568 struct net *net = sock_net(skb->sk);
2569 struct net_device *dev;
2571 u32 portid = NETLINK_CB(cb->skb).portid;
2572 u32 seq = cb->nlh->nlmsg_seq;
2573 struct nlattr *extfilt;
2574 u32 filter_mask = 0;
2576 extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct ifinfomsg),
2579 filter_mask = nla_get_u32(extfilt);
2582 for_each_netdev_rcu(net, dev) {
2583 const struct net_device_ops *ops = dev->netdev_ops;
2584 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2586 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
2587 if (idx >= cb->args[0] &&
2588 br_dev->netdev_ops->ndo_bridge_getlink(
2589 skb, portid, seq, dev, filter_mask) < 0)
2594 if (ops->ndo_bridge_getlink) {
2595 if (idx >= cb->args[0] &&
2596 ops->ndo_bridge_getlink(skb, portid, seq, dev,
2608 static inline size_t bridge_nlmsg_size(void)
2610 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
2611 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
2612 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
2613 + nla_total_size(sizeof(u32)) /* IFLA_MASTER */
2614 + nla_total_size(sizeof(u32)) /* IFLA_MTU */
2615 + nla_total_size(sizeof(u32)) /* IFLA_LINK */
2616 + nla_total_size(sizeof(u32)) /* IFLA_OPERSTATE */
2617 + nla_total_size(sizeof(u8)) /* IFLA_PROTINFO */
2618 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
2619 + nla_total_size(sizeof(u16)) /* IFLA_BRIDGE_FLAGS */
2620 + nla_total_size(sizeof(u16)); /* IFLA_BRIDGE_MODE */
2623 static int rtnl_bridge_notify(struct net_device *dev, u16 flags)
2625 struct net *net = dev_net(dev);
2626 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2627 struct sk_buff *skb;
2628 int err = -EOPNOTSUPP;
2630 skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
2636 if ((!flags || (flags & BRIDGE_FLAGS_MASTER)) &&
2637 br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
2638 err = br_dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0);
2643 if ((flags & BRIDGE_FLAGS_SELF) &&
2644 dev->netdev_ops->ndo_bridge_getlink) {
2645 err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0);
2650 rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
2653 WARN_ON(err == -EMSGSIZE);
2655 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2659 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
2661 struct net *net = sock_net(skb->sk);
2662 struct ifinfomsg *ifm;
2663 struct net_device *dev;
2664 struct nlattr *br_spec, *attr = NULL;
2665 int rem, err = -EOPNOTSUPP;
2666 u16 oflags, flags = 0;
2667 bool have_flags = false;
2669 if (nlmsg_len(nlh) < sizeof(*ifm))
2672 ifm = nlmsg_data(nlh);
2673 if (ifm->ifi_family != AF_BRIDGE)
2674 return -EPFNOSUPPORT;
2676 dev = __dev_get_by_index(net, ifm->ifi_index);
2678 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
2682 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
2684 nla_for_each_nested(attr, br_spec, rem) {
2685 if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
2687 flags = nla_get_u16(attr);
2695 if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
2696 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2698 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
2703 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh);
2707 flags &= ~BRIDGE_FLAGS_MASTER;
2710 if ((flags & BRIDGE_FLAGS_SELF)) {
2711 if (!dev->netdev_ops->ndo_bridge_setlink)
2714 err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh);
2717 flags &= ~BRIDGE_FLAGS_SELF;
2721 memcpy(nla_data(attr), &flags, sizeof(flags));
2722 /* Generate event to notify upper layer of bridge change */
2724 err = rtnl_bridge_notify(dev, oflags);
2729 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
2731 struct net *net = sock_net(skb->sk);
2732 struct ifinfomsg *ifm;
2733 struct net_device *dev;
2734 struct nlattr *br_spec, *attr = NULL;
2735 int rem, err = -EOPNOTSUPP;
2736 u16 oflags, flags = 0;
2737 bool have_flags = false;
2739 if (nlmsg_len(nlh) < sizeof(*ifm))
2742 ifm = nlmsg_data(nlh);
2743 if (ifm->ifi_family != AF_BRIDGE)
2744 return -EPFNOSUPPORT;
2746 dev = __dev_get_by_index(net, ifm->ifi_index);
2748 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
2752 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
2754 nla_for_each_nested(attr, br_spec, rem) {
2755 if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
2757 flags = nla_get_u16(attr);
2765 if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
2766 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2768 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
2773 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh);
2777 flags &= ~BRIDGE_FLAGS_MASTER;
2780 if ((flags & BRIDGE_FLAGS_SELF)) {
2781 if (!dev->netdev_ops->ndo_bridge_dellink)
2784 err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh);
2787 flags &= ~BRIDGE_FLAGS_SELF;
2791 memcpy(nla_data(attr), &flags, sizeof(flags));
2792 /* Generate event to notify upper layer of bridge change */
2794 err = rtnl_bridge_notify(dev, oflags);
2799 /* Process one rtnetlink message. */
2801 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
2803 struct net *net = sock_net(skb->sk);
2804 rtnl_doit_func doit;
2810 type = nlh->nlmsg_type;
2816 /* All the messages must have at least 1 byte length */
2817 if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
2820 family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
2824 if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
2827 if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
2829 rtnl_dumpit_func dumpit;
2830 rtnl_calcit_func calcit;
2831 u16 min_dump_alloc = 0;
2833 dumpit = rtnl_get_dumpit(family, type);
2836 calcit = rtnl_get_calcit(family, type);
2838 min_dump_alloc = calcit(skb, nlh);
2843 struct netlink_dump_control c = {
2845 .min_dump_alloc = min_dump_alloc,
2847 err = netlink_dump_start(rtnl, skb, nlh, &c);
2853 doit = rtnl_get_doit(family, type);
2857 return doit(skb, nlh);
2860 static void rtnetlink_rcv(struct sk_buff *skb)
2863 netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
2867 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
2869 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2875 case NETDEV_POST_INIT:
2876 case NETDEV_REGISTER:
2878 case NETDEV_PRE_TYPE_CHANGE:
2879 case NETDEV_GOING_DOWN:
2880 case NETDEV_UNREGISTER:
2881 case NETDEV_UNREGISTER_FINAL:
2882 case NETDEV_RELEASE:
2886 rtmsg_ifinfo(RTM_NEWLINK, dev, 0, GFP_KERNEL);
2892 static struct notifier_block rtnetlink_dev_notifier = {
2893 .notifier_call = rtnetlink_event,
2897 static int __net_init rtnetlink_net_init(struct net *net)
2900 struct netlink_kernel_cfg cfg = {
2901 .groups = RTNLGRP_MAX,
2902 .input = rtnetlink_rcv,
2903 .cb_mutex = &rtnl_mutex,
2904 .flags = NL_CFG_F_NONROOT_RECV,
2907 sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
2914 static void __net_exit rtnetlink_net_exit(struct net *net)
2916 netlink_kernel_release(net->rtnl);
2920 static struct pernet_operations rtnetlink_net_ops = {
2921 .init = rtnetlink_net_init,
2922 .exit = rtnetlink_net_exit,
2925 void __init rtnetlink_init(void)
2927 if (register_pernet_subsys(&rtnetlink_net_ops))
2928 panic("rtnetlink_init: cannot initialize rtnetlink\n");
2930 register_netdevice_notifier(&rtnetlink_dev_notifier);
2932 rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
2933 rtnl_dump_ifinfo, rtnl_calcit);
2934 rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
2935 rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
2936 rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
2938 rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
2939 rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
2941 rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, NULL);
2942 rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, NULL);
2943 rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, NULL);
2945 rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, NULL);
2946 rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, NULL);
2947 rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, NULL);