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_spoofchk)) +
802 nla_total_size(sizeof(struct ifla_vf_rate)));
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(4) /* IFLA_CARRIER_CHANGES */
859 + nla_total_size(ext_filter_mask
860 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
861 + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
862 + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
863 + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
864 + rtnl_link_get_af_size(dev) /* IFLA_AF_SPEC */
865 + nla_total_size(MAX_PHYS_PORT_ID_LEN); /* IFLA_PHYS_PORT_ID */
868 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
870 struct nlattr *vf_ports;
871 struct nlattr *vf_port;
875 vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
879 for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
880 vf_port = nla_nest_start(skb, IFLA_VF_PORT);
882 goto nla_put_failure;
883 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
884 goto nla_put_failure;
885 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
886 if (err == -EMSGSIZE)
887 goto nla_put_failure;
889 nla_nest_cancel(skb, vf_port);
892 nla_nest_end(skb, vf_port);
895 nla_nest_end(skb, vf_ports);
900 nla_nest_cancel(skb, vf_ports);
904 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
906 struct nlattr *port_self;
909 port_self = nla_nest_start(skb, IFLA_PORT_SELF);
913 err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
915 nla_nest_cancel(skb, port_self);
916 return (err == -EMSGSIZE) ? err : 0;
919 nla_nest_end(skb, port_self);
924 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
929 if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
930 !(ext_filter_mask & RTEXT_FILTER_VF))
933 err = rtnl_port_self_fill(skb, dev);
937 if (dev_num_vf(dev->dev.parent)) {
938 err = rtnl_vf_ports_fill(skb, dev);
946 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
949 struct netdev_phys_port_id ppid;
951 err = dev_get_phys_port_id(dev, &ppid);
953 if (err == -EOPNOTSUPP)
958 if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
964 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
965 int type, u32 pid, u32 seq, u32 change,
966 unsigned int flags, u32 ext_filter_mask)
968 struct ifinfomsg *ifm;
969 struct nlmsghdr *nlh;
970 struct rtnl_link_stats64 temp;
971 const struct rtnl_link_stats64 *stats;
972 struct nlattr *attr, *af_spec;
973 struct rtnl_af_ops *af_ops;
974 struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
977 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
981 ifm = nlmsg_data(nlh);
982 ifm->ifi_family = AF_UNSPEC;
984 ifm->ifi_type = dev->type;
985 ifm->ifi_index = dev->ifindex;
986 ifm->ifi_flags = dev_get_flags(dev);
987 ifm->ifi_change = change;
989 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
990 nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
991 nla_put_u8(skb, IFLA_OPERSTATE,
992 netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
993 nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
994 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
995 nla_put_u32(skb, IFLA_GROUP, dev->group) ||
996 nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
997 nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
999 nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1001 (dev->ifindex != dev->iflink &&
1002 nla_put_u32(skb, IFLA_LINK, dev->iflink)) ||
1004 nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex)) ||
1005 nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1007 nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
1009 nla_put_string(skb, IFLA_IFALIAS, dev->ifalias)) ||
1010 nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1011 atomic_read(&dev->carrier_changes)))
1012 goto nla_put_failure;
1015 struct rtnl_link_ifmap map = {
1016 .mem_start = dev->mem_start,
1017 .mem_end = dev->mem_end,
1018 .base_addr = dev->base_addr,
1021 .port = dev->if_port,
1023 if (nla_put(skb, IFLA_MAP, sizeof(map), &map))
1024 goto nla_put_failure;
1027 if (dev->addr_len) {
1028 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1029 nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1030 goto nla_put_failure;
1033 if (rtnl_phys_port_id_fill(skb, dev))
1034 goto nla_put_failure;
1036 attr = nla_reserve(skb, IFLA_STATS,
1037 sizeof(struct rtnl_link_stats));
1039 goto nla_put_failure;
1041 stats = dev_get_stats(dev, &temp);
1042 copy_rtnl_link_stats(nla_data(attr), stats);
1044 attr = nla_reserve(skb, IFLA_STATS64,
1045 sizeof(struct rtnl_link_stats64));
1047 goto nla_put_failure;
1048 copy_rtnl_link_stats64(nla_data(attr), stats);
1050 if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF) &&
1051 nla_put_u32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent)))
1052 goto nla_put_failure;
1054 if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent
1055 && (ext_filter_mask & RTEXT_FILTER_VF)) {
1058 struct nlattr *vfinfo, *vf;
1059 int num_vfs = dev_num_vf(dev->dev.parent);
1061 vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
1063 goto nla_put_failure;
1064 for (i = 0; i < num_vfs; i++) {
1065 struct ifla_vf_info ivi;
1066 struct ifla_vf_mac vf_mac;
1067 struct ifla_vf_vlan vf_vlan;
1068 struct ifla_vf_rate vf_rate;
1069 struct ifla_vf_tx_rate vf_tx_rate;
1070 struct ifla_vf_spoofchk vf_spoofchk;
1071 struct ifla_vf_link_state vf_linkstate;
1074 * Not all SR-IOV capable drivers support the
1075 * spoofcheck query. Preset to -1 so the user
1076 * space tool can detect that the driver didn't
1080 memset(ivi.mac, 0, sizeof(ivi.mac));
1081 /* The default value for VF link state is "auto"
1082 * IFLA_VF_LINK_STATE_AUTO which equals zero
1085 if (dev->netdev_ops->ndo_get_vf_config(dev, i, &ivi))
1092 vf_linkstate.vf = ivi.vf;
1094 memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1095 vf_vlan.vlan = ivi.vlan;
1096 vf_vlan.qos = ivi.qos;
1097 vf_tx_rate.rate = ivi.max_tx_rate;
1098 vf_rate.min_tx_rate = ivi.min_tx_rate;
1099 vf_rate.max_tx_rate = ivi.max_tx_rate;
1100 vf_spoofchk.setting = ivi.spoofchk;
1101 vf_linkstate.link_state = ivi.linkstate;
1102 vf = nla_nest_start(skb, IFLA_VF_INFO);
1104 nla_nest_cancel(skb, vfinfo);
1105 goto nla_put_failure;
1107 if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1108 nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1109 nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1111 nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1113 nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1115 nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1117 goto nla_put_failure;
1118 nla_nest_end(skb, vf);
1120 nla_nest_end(skb, vfinfo);
1123 if (rtnl_port_fill(skb, dev, ext_filter_mask))
1124 goto nla_put_failure;
1126 if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1127 if (rtnl_link_fill(skb, dev) < 0)
1128 goto nla_put_failure;
1131 if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1132 goto nla_put_failure;
1134 list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1135 if (af_ops->fill_link_af) {
1139 if (!(af = nla_nest_start(skb, af_ops->family)))
1140 goto nla_put_failure;
1142 err = af_ops->fill_link_af(skb, dev);
1145 * Caller may return ENODATA to indicate that there
1146 * was no data to be dumped. This is not an error, it
1147 * means we should trim the attribute header and
1150 if (err == -ENODATA)
1151 nla_nest_cancel(skb, af);
1153 goto nla_put_failure;
1155 nla_nest_end(skb, af);
1159 nla_nest_end(skb, af_spec);
1161 return nlmsg_end(skb, nlh);
1164 nlmsg_cancel(skb, nlh);
1168 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1169 [IFLA_IFNAME] = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1170 [IFLA_ADDRESS] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1171 [IFLA_BROADCAST] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1172 [IFLA_MAP] = { .len = sizeof(struct rtnl_link_ifmap) },
1173 [IFLA_MTU] = { .type = NLA_U32 },
1174 [IFLA_LINK] = { .type = NLA_U32 },
1175 [IFLA_MASTER] = { .type = NLA_U32 },
1176 [IFLA_CARRIER] = { .type = NLA_U8 },
1177 [IFLA_TXQLEN] = { .type = NLA_U32 },
1178 [IFLA_WEIGHT] = { .type = NLA_U32 },
1179 [IFLA_OPERSTATE] = { .type = NLA_U8 },
1180 [IFLA_LINKMODE] = { .type = NLA_U8 },
1181 [IFLA_LINKINFO] = { .type = NLA_NESTED },
1182 [IFLA_NET_NS_PID] = { .type = NLA_U32 },
1183 [IFLA_NET_NS_FD] = { .type = NLA_U32 },
1184 [IFLA_IFALIAS] = { .type = NLA_STRING, .len = IFALIASZ-1 },
1185 [IFLA_VFINFO_LIST] = {. type = NLA_NESTED },
1186 [IFLA_VF_PORTS] = { .type = NLA_NESTED },
1187 [IFLA_PORT_SELF] = { .type = NLA_NESTED },
1188 [IFLA_AF_SPEC] = { .type = NLA_NESTED },
1189 [IFLA_EXT_MASK] = { .type = NLA_U32 },
1190 [IFLA_PROMISCUITY] = { .type = NLA_U32 },
1191 [IFLA_NUM_TX_QUEUES] = { .type = NLA_U32 },
1192 [IFLA_NUM_RX_QUEUES] = { .type = NLA_U32 },
1193 [IFLA_PHYS_PORT_ID] = { .type = NLA_BINARY, .len = MAX_PHYS_PORT_ID_LEN },
1194 [IFLA_CARRIER_CHANGES] = { .type = NLA_U32 }, /* ignored */
1197 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1198 [IFLA_INFO_KIND] = { .type = NLA_STRING },
1199 [IFLA_INFO_DATA] = { .type = NLA_NESTED },
1200 [IFLA_INFO_SLAVE_KIND] = { .type = NLA_STRING },
1201 [IFLA_INFO_SLAVE_DATA] = { .type = NLA_NESTED },
1204 static const struct nla_policy ifla_vfinfo_policy[IFLA_VF_INFO_MAX+1] = {
1205 [IFLA_VF_INFO] = { .type = NLA_NESTED },
1208 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1209 [IFLA_VF_MAC] = { .type = NLA_BINARY,
1210 .len = sizeof(struct ifla_vf_mac) },
1211 [IFLA_VF_VLAN] = { .type = NLA_BINARY,
1212 .len = sizeof(struct ifla_vf_vlan) },
1213 [IFLA_VF_TX_RATE] = { .type = NLA_BINARY,
1214 .len = sizeof(struct ifla_vf_tx_rate) },
1215 [IFLA_VF_SPOOFCHK] = { .type = NLA_BINARY,
1216 .len = sizeof(struct ifla_vf_spoofchk) },
1217 [IFLA_VF_RATE] = { .type = NLA_BINARY,
1218 .len = sizeof(struct ifla_vf_rate) },
1221 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1222 [IFLA_PORT_VF] = { .type = NLA_U32 },
1223 [IFLA_PORT_PROFILE] = { .type = NLA_STRING,
1224 .len = PORT_PROFILE_MAX },
1225 [IFLA_PORT_VSI_TYPE] = { .type = NLA_BINARY,
1226 .len = sizeof(struct ifla_port_vsi)},
1227 [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1228 .len = PORT_UUID_MAX },
1229 [IFLA_PORT_HOST_UUID] = { .type = NLA_STRING,
1230 .len = PORT_UUID_MAX },
1231 [IFLA_PORT_REQUEST] = { .type = NLA_U8, },
1232 [IFLA_PORT_RESPONSE] = { .type = NLA_U16, },
1235 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1237 struct net *net = sock_net(skb->sk);
1240 struct net_device *dev;
1241 struct hlist_head *head;
1242 struct nlattr *tb[IFLA_MAX+1];
1243 u32 ext_filter_mask = 0;
1247 s_idx = cb->args[1];
1250 cb->seq = net->dev_base_seq;
1252 if (nlmsg_parse(cb->nlh, sizeof(struct ifinfomsg), tb, IFLA_MAX,
1253 ifla_policy) >= 0) {
1255 if (tb[IFLA_EXT_MASK])
1256 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1259 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1261 head = &net->dev_index_head[h];
1262 hlist_for_each_entry_rcu(dev, head, index_hlist) {
1265 err = rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1266 NETLINK_CB(cb->skb).portid,
1267 cb->nlh->nlmsg_seq, 0,
1270 /* If we ran out of room on the first message,
1273 WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
1278 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1291 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len)
1293 return nla_parse(tb, IFLA_MAX, head, len, ifla_policy);
1295 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
1297 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1300 /* Examine the link attributes and figure out which
1301 * network namespace we are talking about.
1303 if (tb[IFLA_NET_NS_PID])
1304 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1305 else if (tb[IFLA_NET_NS_FD])
1306 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1308 net = get_net(src_net);
1311 EXPORT_SYMBOL(rtnl_link_get_net);
1313 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1316 if (tb[IFLA_ADDRESS] &&
1317 nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1320 if (tb[IFLA_BROADCAST] &&
1321 nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1325 if (tb[IFLA_AF_SPEC]) {
1329 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1330 const struct rtnl_af_ops *af_ops;
1332 if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1333 return -EAFNOSUPPORT;
1335 if (!af_ops->set_link_af)
1338 if (af_ops->validate_link_af) {
1339 err = af_ops->validate_link_af(dev, af);
1349 static int do_setvfinfo(struct net_device *dev, struct nlattr *attr)
1351 int rem, err = -EINVAL;
1353 const struct net_device_ops *ops = dev->netdev_ops;
1355 nla_for_each_nested(vf, attr, rem) {
1356 switch (nla_type(vf)) {
1358 struct ifla_vf_mac *ivm;
1361 if (ops->ndo_set_vf_mac)
1362 err = ops->ndo_set_vf_mac(dev, ivm->vf,
1366 case IFLA_VF_VLAN: {
1367 struct ifla_vf_vlan *ivv;
1370 if (ops->ndo_set_vf_vlan)
1371 err = ops->ndo_set_vf_vlan(dev, ivv->vf,
1376 case IFLA_VF_TX_RATE: {
1377 struct ifla_vf_tx_rate *ivt;
1378 struct ifla_vf_info ivf;
1381 if (ops->ndo_get_vf_config)
1382 err = ops->ndo_get_vf_config(dev, ivt->vf,
1387 if (ops->ndo_set_vf_rate)
1388 err = ops->ndo_set_vf_rate(dev, ivt->vf,
1393 case IFLA_VF_RATE: {
1394 struct ifla_vf_rate *ivt;
1397 if (ops->ndo_set_vf_rate)
1398 err = ops->ndo_set_vf_rate(dev, ivt->vf,
1403 case IFLA_VF_SPOOFCHK: {
1404 struct ifla_vf_spoofchk *ivs;
1407 if (ops->ndo_set_vf_spoofchk)
1408 err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1412 case IFLA_VF_LINK_STATE: {
1413 struct ifla_vf_link_state *ivl;
1416 if (ops->ndo_set_vf_link_state)
1417 err = ops->ndo_set_vf_link_state(dev, ivl->vf,
1431 static int do_set_master(struct net_device *dev, int ifindex)
1433 struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1434 const struct net_device_ops *ops;
1438 if (upper_dev->ifindex == ifindex)
1440 ops = upper_dev->netdev_ops;
1441 if (ops->ndo_del_slave) {
1442 err = ops->ndo_del_slave(upper_dev, dev);
1451 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
1454 ops = upper_dev->netdev_ops;
1455 if (ops->ndo_add_slave) {
1456 err = ops->ndo_add_slave(upper_dev, dev);
1466 static int do_setlink(const struct sk_buff *skb,
1467 struct net_device *dev, struct ifinfomsg *ifm,
1468 struct nlattr **tb, char *ifname, int modified)
1470 const struct net_device_ops *ops = dev->netdev_ops;
1473 if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1474 struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1479 if (!netlink_ns_capable(skb, net->user_ns, CAP_NET_ADMIN)) {
1483 err = dev_change_net_namespace(dev, net, ifname);
1491 struct rtnl_link_ifmap *u_map;
1494 if (!ops->ndo_set_config) {
1499 if (!netif_device_present(dev)) {
1504 u_map = nla_data(tb[IFLA_MAP]);
1505 k_map.mem_start = (unsigned long) u_map->mem_start;
1506 k_map.mem_end = (unsigned long) u_map->mem_end;
1507 k_map.base_addr = (unsigned short) u_map->base_addr;
1508 k_map.irq = (unsigned char) u_map->irq;
1509 k_map.dma = (unsigned char) u_map->dma;
1510 k_map.port = (unsigned char) u_map->port;
1512 err = ops->ndo_set_config(dev, &k_map);
1519 if (tb[IFLA_ADDRESS]) {
1520 struct sockaddr *sa;
1523 len = sizeof(sa_family_t) + dev->addr_len;
1524 sa = kmalloc(len, GFP_KERNEL);
1529 sa->sa_family = dev->type;
1530 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1532 err = dev_set_mac_address(dev, sa);
1540 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1546 if (tb[IFLA_GROUP]) {
1547 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1552 * Interface selected by interface index but interface
1553 * name provided implies that a name change has been
1556 if (ifm->ifi_index > 0 && ifname[0]) {
1557 err = dev_change_name(dev, ifname);
1563 if (tb[IFLA_IFALIAS]) {
1564 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
1565 nla_len(tb[IFLA_IFALIAS]));
1571 if (tb[IFLA_BROADCAST]) {
1572 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
1573 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
1576 if (ifm->ifi_flags || ifm->ifi_change) {
1577 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1582 if (tb[IFLA_MASTER]) {
1583 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
1589 if (tb[IFLA_CARRIER]) {
1590 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
1596 if (tb[IFLA_TXQLEN])
1597 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1599 if (tb[IFLA_OPERSTATE])
1600 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1602 if (tb[IFLA_LINKMODE]) {
1603 write_lock_bh(&dev_base_lock);
1604 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1605 write_unlock_bh(&dev_base_lock);
1608 if (tb[IFLA_VFINFO_LIST]) {
1609 struct nlattr *attr;
1611 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
1612 if (nla_type(attr) != IFLA_VF_INFO) {
1616 err = do_setvfinfo(dev, attr);
1624 if (tb[IFLA_VF_PORTS]) {
1625 struct nlattr *port[IFLA_PORT_MAX+1];
1626 struct nlattr *attr;
1631 if (!ops->ndo_set_vf_port)
1634 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
1635 if (nla_type(attr) != IFLA_VF_PORT)
1637 err = nla_parse_nested(port, IFLA_PORT_MAX,
1638 attr, ifla_port_policy);
1641 if (!port[IFLA_PORT_VF]) {
1645 vf = nla_get_u32(port[IFLA_PORT_VF]);
1646 err = ops->ndo_set_vf_port(dev, vf, port);
1654 if (tb[IFLA_PORT_SELF]) {
1655 struct nlattr *port[IFLA_PORT_MAX+1];
1657 err = nla_parse_nested(port, IFLA_PORT_MAX,
1658 tb[IFLA_PORT_SELF], ifla_port_policy);
1663 if (ops->ndo_set_vf_port)
1664 err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
1670 if (tb[IFLA_AF_SPEC]) {
1674 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1675 const struct rtnl_af_ops *af_ops;
1677 if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1680 err = af_ops->set_link_af(dev, af);
1690 if (err < 0 && modified)
1691 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",
1697 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
1699 struct net *net = sock_net(skb->sk);
1700 struct ifinfomsg *ifm;
1701 struct net_device *dev;
1703 struct nlattr *tb[IFLA_MAX+1];
1704 char ifname[IFNAMSIZ];
1706 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1710 if (tb[IFLA_IFNAME])
1711 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1716 ifm = nlmsg_data(nlh);
1717 if (ifm->ifi_index > 0)
1718 dev = __dev_get_by_index(net, ifm->ifi_index);
1719 else if (tb[IFLA_IFNAME])
1720 dev = __dev_get_by_name(net, ifname);
1729 err = validate_linkmsg(dev, tb);
1733 err = do_setlink(skb, dev, ifm, tb, ifname, 0);
1738 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
1740 struct net *net = sock_net(skb->sk);
1741 const struct rtnl_link_ops *ops;
1742 struct net_device *dev;
1743 struct ifinfomsg *ifm;
1744 char ifname[IFNAMSIZ];
1745 struct nlattr *tb[IFLA_MAX+1];
1747 LIST_HEAD(list_kill);
1749 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1753 if (tb[IFLA_IFNAME])
1754 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1756 ifm = nlmsg_data(nlh);
1757 if (ifm->ifi_index > 0)
1758 dev = __dev_get_by_index(net, ifm->ifi_index);
1759 else if (tb[IFLA_IFNAME])
1760 dev = __dev_get_by_name(net, ifname);
1767 ops = dev->rtnl_link_ops;
1771 ops->dellink(dev, &list_kill);
1772 unregister_netdevice_many(&list_kill);
1773 list_del(&list_kill);
1777 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
1779 unsigned int old_flags;
1782 old_flags = dev->flags;
1783 if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
1784 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1789 dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
1791 __dev_notify_flags(dev, old_flags, ~0U);
1794 EXPORT_SYMBOL(rtnl_configure_link);
1796 struct net_device *rtnl_create_link(struct net *net,
1797 char *ifname, const struct rtnl_link_ops *ops, struct nlattr *tb[])
1800 struct net_device *dev;
1801 unsigned int num_tx_queues = 1;
1802 unsigned int num_rx_queues = 1;
1804 if (tb[IFLA_NUM_TX_QUEUES])
1805 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
1806 else if (ops->get_num_tx_queues)
1807 num_tx_queues = ops->get_num_tx_queues();
1809 if (tb[IFLA_NUM_RX_QUEUES])
1810 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
1811 else if (ops->get_num_rx_queues)
1812 num_rx_queues = ops->get_num_rx_queues();
1815 dev = alloc_netdev_mqs(ops->priv_size, ifname, ops->setup,
1816 num_tx_queues, num_rx_queues);
1820 dev_net_set(dev, net);
1821 dev->rtnl_link_ops = ops;
1822 dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
1825 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
1826 if (tb[IFLA_ADDRESS]) {
1827 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
1828 nla_len(tb[IFLA_ADDRESS]));
1829 dev->addr_assign_type = NET_ADDR_SET;
1831 if (tb[IFLA_BROADCAST])
1832 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
1833 nla_len(tb[IFLA_BROADCAST]));
1834 if (tb[IFLA_TXQLEN])
1835 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1836 if (tb[IFLA_OPERSTATE])
1837 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1838 if (tb[IFLA_LINKMODE])
1839 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1841 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1846 return ERR_PTR(err);
1848 EXPORT_SYMBOL(rtnl_create_link);
1850 static int rtnl_group_changelink(const struct sk_buff *skb,
1851 struct net *net, int group,
1852 struct ifinfomsg *ifm,
1855 struct net_device *dev;
1858 for_each_netdev(net, dev) {
1859 if (dev->group == group) {
1860 err = do_setlink(skb, dev, ifm, tb, NULL, 0);
1869 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh)
1871 struct net *net = sock_net(skb->sk);
1872 const struct rtnl_link_ops *ops;
1873 const struct rtnl_link_ops *m_ops = NULL;
1874 struct net_device *dev;
1875 struct net_device *master_dev = NULL;
1876 struct ifinfomsg *ifm;
1877 char kind[MODULE_NAME_LEN];
1878 char ifname[IFNAMSIZ];
1879 struct nlattr *tb[IFLA_MAX+1];
1880 struct nlattr *linkinfo[IFLA_INFO_MAX+1];
1883 #ifdef CONFIG_MODULES
1886 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1890 if (tb[IFLA_IFNAME])
1891 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1895 ifm = nlmsg_data(nlh);
1896 if (ifm->ifi_index > 0)
1897 dev = __dev_get_by_index(net, ifm->ifi_index);
1900 dev = __dev_get_by_name(net, ifname);
1906 master_dev = netdev_master_upper_dev_get(dev);
1908 m_ops = master_dev->rtnl_link_ops;
1911 err = validate_linkmsg(dev, tb);
1915 if (tb[IFLA_LINKINFO]) {
1916 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
1917 tb[IFLA_LINKINFO], ifla_info_policy);
1921 memset(linkinfo, 0, sizeof(linkinfo));
1923 if (linkinfo[IFLA_INFO_KIND]) {
1924 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
1925 ops = rtnl_link_ops_get(kind);
1932 struct nlattr *attr[ops ? ops->maxtype + 1 : 0];
1933 struct nlattr *slave_attr[m_ops ? m_ops->slave_maxtype + 1 : 0];
1934 struct nlattr **data = NULL;
1935 struct nlattr **slave_data = NULL;
1936 struct net *dest_net;
1939 if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
1940 err = nla_parse_nested(attr, ops->maxtype,
1941 linkinfo[IFLA_INFO_DATA],
1947 if (ops->validate) {
1948 err = ops->validate(tb, data);
1955 if (m_ops->slave_maxtype &&
1956 linkinfo[IFLA_INFO_SLAVE_DATA]) {
1957 err = nla_parse_nested(slave_attr,
1958 m_ops->slave_maxtype,
1959 linkinfo[IFLA_INFO_SLAVE_DATA],
1960 m_ops->slave_policy);
1963 slave_data = slave_attr;
1965 if (m_ops->slave_validate) {
1966 err = m_ops->slave_validate(tb, slave_data);
1975 if (nlh->nlmsg_flags & NLM_F_EXCL)
1977 if (nlh->nlmsg_flags & NLM_F_REPLACE)
1980 if (linkinfo[IFLA_INFO_DATA]) {
1981 if (!ops || ops != dev->rtnl_link_ops ||
1985 err = ops->changelink(dev, tb, data);
1991 if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
1992 if (!m_ops || !m_ops->slave_changelink)
1995 err = m_ops->slave_changelink(master_dev, dev,
2002 return do_setlink(skb, dev, ifm, tb, ifname, modified);
2005 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
2006 if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
2007 return rtnl_group_changelink(skb, net,
2008 nla_get_u32(tb[IFLA_GROUP]),
2013 if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
2017 #ifdef CONFIG_MODULES
2020 request_module("rtnl-link-%s", kind);
2022 ops = rtnl_link_ops_get(kind);
2031 snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
2033 dest_net = rtnl_link_get_net(net, tb);
2034 if (IS_ERR(dest_net))
2035 return PTR_ERR(dest_net);
2037 dev = rtnl_create_link(dest_net, ifname, ops, tb);
2043 dev->ifindex = ifm->ifi_index;
2046 err = ops->newlink(net, dev, tb, data);
2047 /* Drivers should call free_netdev() in ->destructor
2048 * and unregister it on failure so that device could be
2049 * finally freed in rtnl_unlock.
2054 err = register_netdevice(dev);
2060 err = rtnl_configure_link(dev, ifm);
2062 unregister_netdevice(dev);
2069 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh)
2071 struct net *net = sock_net(skb->sk);
2072 struct ifinfomsg *ifm;
2073 char ifname[IFNAMSIZ];
2074 struct nlattr *tb[IFLA_MAX+1];
2075 struct net_device *dev = NULL;
2076 struct sk_buff *nskb;
2078 u32 ext_filter_mask = 0;
2080 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2084 if (tb[IFLA_IFNAME])
2085 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2087 if (tb[IFLA_EXT_MASK])
2088 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2090 ifm = nlmsg_data(nlh);
2091 if (ifm->ifi_index > 0)
2092 dev = __dev_get_by_index(net, ifm->ifi_index);
2093 else if (tb[IFLA_IFNAME])
2094 dev = __dev_get_by_name(net, ifname);
2101 nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
2105 err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).portid,
2106 nlh->nlmsg_seq, 0, 0, ext_filter_mask);
2108 /* -EMSGSIZE implies BUG in if_nlmsg_size */
2109 WARN_ON(err == -EMSGSIZE);
2112 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
2117 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
2119 struct net *net = sock_net(skb->sk);
2120 struct net_device *dev;
2121 struct nlattr *tb[IFLA_MAX+1];
2122 u32 ext_filter_mask = 0;
2123 u16 min_ifinfo_dump_size = 0;
2125 if (nlmsg_parse(nlh, sizeof(struct ifinfomsg), tb, IFLA_MAX,
2126 ifla_policy) >= 0) {
2127 if (tb[IFLA_EXT_MASK])
2128 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2131 if (!ext_filter_mask)
2132 return NLMSG_GOODSIZE;
2134 * traverse the list of net devices and compute the minimum
2135 * buffer size based upon the filter mask.
2137 list_for_each_entry(dev, &net->dev_base_head, dev_list) {
2138 min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
2143 return min_ifinfo_dump_size;
2146 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
2149 int s_idx = cb->family;
2153 for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
2154 int type = cb->nlh->nlmsg_type-RTM_BASE;
2155 if (idx < s_idx || idx == PF_PACKET)
2157 if (rtnl_msg_handlers[idx] == NULL ||
2158 rtnl_msg_handlers[idx][type].dumpit == NULL)
2161 memset(&cb->args[0], 0, sizeof(cb->args));
2165 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
2173 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
2176 struct net *net = dev_net(dev);
2177 struct sk_buff *skb;
2179 size_t if_info_size;
2181 skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
2185 err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0);
2187 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
2188 WARN_ON(err == -EMSGSIZE);
2192 rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
2196 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2198 EXPORT_SYMBOL(rtmsg_ifinfo);
2200 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
2201 struct net_device *dev,
2202 u8 *addr, u32 pid, u32 seq,
2203 int type, unsigned int flags,
2206 struct nlmsghdr *nlh;
2209 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
2213 ndm = nlmsg_data(nlh);
2214 ndm->ndm_family = AF_BRIDGE;
2217 ndm->ndm_flags = flags;
2219 ndm->ndm_ifindex = dev->ifindex;
2220 ndm->ndm_state = NUD_PERMANENT;
2222 if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
2223 goto nla_put_failure;
2225 return nlmsg_end(skb, nlh);
2228 nlmsg_cancel(skb, nlh);
2232 static inline size_t rtnl_fdb_nlmsg_size(void)
2234 return NLMSG_ALIGN(sizeof(struct ndmsg)) + nla_total_size(ETH_ALEN);
2237 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, int type)
2239 struct net *net = dev_net(dev);
2240 struct sk_buff *skb;
2243 skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
2247 err = nlmsg_populate_fdb_fill(skb, dev, addr, 0, 0, type, NTF_SELF, 0);
2253 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2256 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
2260 * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
2262 int ndo_dflt_fdb_add(struct ndmsg *ndm,
2263 struct nlattr *tb[],
2264 struct net_device *dev,
2265 const unsigned char *addr,
2270 /* If aging addresses are supported device will need to
2271 * implement its own handler for this.
2273 if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
2274 pr_info("%s: FDB only supports static addresses\n", dev->name);
2278 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2279 err = dev_uc_add_excl(dev, addr);
2280 else if (is_multicast_ether_addr(addr))
2281 err = dev_mc_add_excl(dev, addr);
2283 /* Only return duplicate errors if NLM_F_EXCL is set */
2284 if (err == -EEXIST && !(flags & NLM_F_EXCL))
2289 EXPORT_SYMBOL(ndo_dflt_fdb_add);
2291 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh)
2293 struct net *net = sock_net(skb->sk);
2295 struct nlattr *tb[NDA_MAX+1];
2296 struct net_device *dev;
2300 err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2304 ndm = nlmsg_data(nlh);
2305 if (ndm->ndm_ifindex == 0) {
2306 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n");
2310 dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2312 pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n");
2316 if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2317 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n");
2321 addr = nla_data(tb[NDA_LLADDR]);
2325 /* Support fdb on master device the net/bridge default case */
2326 if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2327 (dev->priv_flags & IFF_BRIDGE_PORT)) {
2328 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2329 const struct net_device_ops *ops = br_dev->netdev_ops;
2331 err = ops->ndo_fdb_add(ndm, tb, dev, addr, nlh->nlmsg_flags);
2335 ndm->ndm_flags &= ~NTF_MASTER;
2338 /* Embedded bridge, macvlan, and any other device support */
2339 if ((ndm->ndm_flags & NTF_SELF)) {
2340 if (dev->netdev_ops->ndo_fdb_add)
2341 err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
2344 err = ndo_dflt_fdb_add(ndm, tb, dev, addr,
2348 rtnl_fdb_notify(dev, addr, RTM_NEWNEIGH);
2349 ndm->ndm_flags &= ~NTF_SELF;
2357 * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
2359 int ndo_dflt_fdb_del(struct ndmsg *ndm,
2360 struct nlattr *tb[],
2361 struct net_device *dev,
2362 const unsigned char *addr)
2364 int err = -EOPNOTSUPP;
2366 /* If aging addresses are supported device will need to
2367 * implement its own handler for this.
2369 if (!(ndm->ndm_state & NUD_PERMANENT)) {
2370 pr_info("%s: FDB only supports static addresses\n", dev->name);
2374 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2375 err = dev_uc_del(dev, addr);
2376 else if (is_multicast_ether_addr(addr))
2377 err = dev_mc_del(dev, addr);
2383 EXPORT_SYMBOL(ndo_dflt_fdb_del);
2385 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh)
2387 struct net *net = sock_net(skb->sk);
2389 struct nlattr *tb[NDA_MAX+1];
2390 struct net_device *dev;
2394 if (!netlink_capable(skb, CAP_NET_ADMIN))
2397 err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2401 ndm = nlmsg_data(nlh);
2402 if (ndm->ndm_ifindex == 0) {
2403 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n");
2407 dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2409 pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n");
2413 if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2414 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid address\n");
2418 addr = nla_data(tb[NDA_LLADDR]);
2422 /* Support fdb on master device the net/bridge default case */
2423 if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2424 (dev->priv_flags & IFF_BRIDGE_PORT)) {
2425 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2426 const struct net_device_ops *ops = br_dev->netdev_ops;
2428 if (ops->ndo_fdb_del)
2429 err = ops->ndo_fdb_del(ndm, tb, dev, addr);
2434 ndm->ndm_flags &= ~NTF_MASTER;
2437 /* Embedded bridge, macvlan, and any other device support */
2438 if (ndm->ndm_flags & NTF_SELF) {
2439 if (dev->netdev_ops->ndo_fdb_del)
2440 err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr);
2442 err = ndo_dflt_fdb_del(ndm, tb, dev, addr);
2445 rtnl_fdb_notify(dev, addr, RTM_DELNEIGH);
2446 ndm->ndm_flags &= ~NTF_SELF;
2453 static int nlmsg_populate_fdb(struct sk_buff *skb,
2454 struct netlink_callback *cb,
2455 struct net_device *dev,
2457 struct netdev_hw_addr_list *list)
2459 struct netdev_hw_addr *ha;
2463 portid = NETLINK_CB(cb->skb).portid;
2464 seq = cb->nlh->nlmsg_seq;
2466 list_for_each_entry(ha, &list->list, list) {
2467 if (*idx < cb->args[0])
2470 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr,
2472 RTM_NEWNEIGH, NTF_SELF,
2483 * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
2484 * @nlh: netlink message header
2487 * Default netdevice operation to dump the existing unicast address list.
2488 * Returns number of addresses from list put in skb.
2490 int ndo_dflt_fdb_dump(struct sk_buff *skb,
2491 struct netlink_callback *cb,
2492 struct net_device *dev,
2497 netif_addr_lock_bh(dev);
2498 err = nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->uc);
2501 nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->mc);
2503 netif_addr_unlock_bh(dev);
2506 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
2508 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
2511 struct net *net = sock_net(skb->sk);
2512 struct net_device *dev;
2515 for_each_netdev_rcu(net, dev) {
2516 if (dev->priv_flags & IFF_BRIDGE_PORT) {
2517 struct net_device *br_dev;
2518 const struct net_device_ops *ops;
2520 br_dev = netdev_master_upper_dev_get(dev);
2521 ops = br_dev->netdev_ops;
2522 if (ops->ndo_fdb_dump)
2523 idx = ops->ndo_fdb_dump(skb, cb, dev, idx);
2526 if (dev->netdev_ops->ndo_fdb_dump)
2527 idx = dev->netdev_ops->ndo_fdb_dump(skb, cb, dev, idx);
2529 idx = ndo_dflt_fdb_dump(skb, cb, dev, idx);
2537 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
2538 struct net_device *dev, u16 mode)
2540 struct nlmsghdr *nlh;
2541 struct ifinfomsg *ifm;
2542 struct nlattr *br_afspec;
2543 u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
2544 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2546 nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), NLM_F_MULTI);
2550 ifm = nlmsg_data(nlh);
2551 ifm->ifi_family = AF_BRIDGE;
2553 ifm->ifi_type = dev->type;
2554 ifm->ifi_index = dev->ifindex;
2555 ifm->ifi_flags = dev_get_flags(dev);
2556 ifm->ifi_change = 0;
2559 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
2560 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
2561 nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
2563 nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
2565 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
2566 (dev->ifindex != dev->iflink &&
2567 nla_put_u32(skb, IFLA_LINK, dev->iflink)))
2568 goto nla_put_failure;
2570 br_afspec = nla_nest_start(skb, IFLA_AF_SPEC);
2572 goto nla_put_failure;
2574 if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF) ||
2575 nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
2576 nla_nest_cancel(skb, br_afspec);
2577 goto nla_put_failure;
2579 nla_nest_end(skb, br_afspec);
2581 return nlmsg_end(skb, nlh);
2583 nlmsg_cancel(skb, nlh);
2586 EXPORT_SYMBOL(ndo_dflt_bridge_getlink);
2588 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
2590 struct net *net = sock_net(skb->sk);
2591 struct net_device *dev;
2593 u32 portid = NETLINK_CB(cb->skb).portid;
2594 u32 seq = cb->nlh->nlmsg_seq;
2595 struct nlattr *extfilt;
2596 u32 filter_mask = 0;
2598 extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct ifinfomsg),
2601 filter_mask = nla_get_u32(extfilt);
2604 for_each_netdev_rcu(net, dev) {
2605 const struct net_device_ops *ops = dev->netdev_ops;
2606 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2608 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
2609 if (idx >= cb->args[0] &&
2610 br_dev->netdev_ops->ndo_bridge_getlink(
2611 skb, portid, seq, dev, filter_mask) < 0)
2616 if (ops->ndo_bridge_getlink) {
2617 if (idx >= cb->args[0] &&
2618 ops->ndo_bridge_getlink(skb, portid, seq, dev,
2630 static inline size_t bridge_nlmsg_size(void)
2632 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
2633 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
2634 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
2635 + nla_total_size(sizeof(u32)) /* IFLA_MASTER */
2636 + nla_total_size(sizeof(u32)) /* IFLA_MTU */
2637 + nla_total_size(sizeof(u32)) /* IFLA_LINK */
2638 + nla_total_size(sizeof(u32)) /* IFLA_OPERSTATE */
2639 + nla_total_size(sizeof(u8)) /* IFLA_PROTINFO */
2640 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
2641 + nla_total_size(sizeof(u16)) /* IFLA_BRIDGE_FLAGS */
2642 + nla_total_size(sizeof(u16)); /* IFLA_BRIDGE_MODE */
2645 static int rtnl_bridge_notify(struct net_device *dev, u16 flags)
2647 struct net *net = dev_net(dev);
2648 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2649 struct sk_buff *skb;
2650 int err = -EOPNOTSUPP;
2652 skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
2658 if ((!flags || (flags & BRIDGE_FLAGS_MASTER)) &&
2659 br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
2660 err = br_dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0);
2665 if ((flags & BRIDGE_FLAGS_SELF) &&
2666 dev->netdev_ops->ndo_bridge_getlink) {
2667 err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0);
2672 rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
2675 WARN_ON(err == -EMSGSIZE);
2677 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2681 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
2683 struct net *net = sock_net(skb->sk);
2684 struct ifinfomsg *ifm;
2685 struct net_device *dev;
2686 struct nlattr *br_spec, *attr = NULL;
2687 int rem, err = -EOPNOTSUPP;
2688 u16 oflags, flags = 0;
2689 bool have_flags = false;
2691 if (nlmsg_len(nlh) < sizeof(*ifm))
2694 ifm = nlmsg_data(nlh);
2695 if (ifm->ifi_family != AF_BRIDGE)
2696 return -EPFNOSUPPORT;
2698 dev = __dev_get_by_index(net, ifm->ifi_index);
2700 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
2704 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
2706 nla_for_each_nested(attr, br_spec, rem) {
2707 if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
2709 flags = nla_get_u16(attr);
2717 if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
2718 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2720 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
2725 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh);
2729 flags &= ~BRIDGE_FLAGS_MASTER;
2732 if ((flags & BRIDGE_FLAGS_SELF)) {
2733 if (!dev->netdev_ops->ndo_bridge_setlink)
2736 err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh);
2739 flags &= ~BRIDGE_FLAGS_SELF;
2743 memcpy(nla_data(attr), &flags, sizeof(flags));
2744 /* Generate event to notify upper layer of bridge change */
2746 err = rtnl_bridge_notify(dev, oflags);
2751 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
2753 struct net *net = sock_net(skb->sk);
2754 struct ifinfomsg *ifm;
2755 struct net_device *dev;
2756 struct nlattr *br_spec, *attr = NULL;
2757 int rem, err = -EOPNOTSUPP;
2758 u16 oflags, flags = 0;
2759 bool have_flags = false;
2761 if (nlmsg_len(nlh) < sizeof(*ifm))
2764 ifm = nlmsg_data(nlh);
2765 if (ifm->ifi_family != AF_BRIDGE)
2766 return -EPFNOSUPPORT;
2768 dev = __dev_get_by_index(net, ifm->ifi_index);
2770 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
2774 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
2776 nla_for_each_nested(attr, br_spec, rem) {
2777 if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
2779 flags = nla_get_u16(attr);
2787 if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
2788 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2790 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
2795 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh);
2799 flags &= ~BRIDGE_FLAGS_MASTER;
2802 if ((flags & BRIDGE_FLAGS_SELF)) {
2803 if (!dev->netdev_ops->ndo_bridge_dellink)
2806 err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh);
2809 flags &= ~BRIDGE_FLAGS_SELF;
2813 memcpy(nla_data(attr), &flags, sizeof(flags));
2814 /* Generate event to notify upper layer of bridge change */
2816 err = rtnl_bridge_notify(dev, oflags);
2821 /* Process one rtnetlink message. */
2823 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
2825 struct net *net = sock_net(skb->sk);
2826 rtnl_doit_func doit;
2832 type = nlh->nlmsg_type;
2838 /* All the messages must have at least 1 byte length */
2839 if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
2842 family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
2846 if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
2849 if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
2851 rtnl_dumpit_func dumpit;
2852 rtnl_calcit_func calcit;
2853 u16 min_dump_alloc = 0;
2855 dumpit = rtnl_get_dumpit(family, type);
2858 calcit = rtnl_get_calcit(family, type);
2860 min_dump_alloc = calcit(skb, nlh);
2865 struct netlink_dump_control c = {
2867 .min_dump_alloc = min_dump_alloc,
2869 err = netlink_dump_start(rtnl, skb, nlh, &c);
2875 doit = rtnl_get_doit(family, type);
2879 return doit(skb, nlh);
2882 static void rtnetlink_rcv(struct sk_buff *skb)
2885 netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
2889 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
2891 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2897 case NETDEV_POST_INIT:
2898 case NETDEV_REGISTER:
2900 case NETDEV_PRE_TYPE_CHANGE:
2901 case NETDEV_GOING_DOWN:
2902 case NETDEV_UNREGISTER:
2903 case NETDEV_UNREGISTER_FINAL:
2904 case NETDEV_RELEASE:
2908 rtmsg_ifinfo(RTM_NEWLINK, dev, 0, GFP_KERNEL);
2914 static struct notifier_block rtnetlink_dev_notifier = {
2915 .notifier_call = rtnetlink_event,
2919 static int __net_init rtnetlink_net_init(struct net *net)
2922 struct netlink_kernel_cfg cfg = {
2923 .groups = RTNLGRP_MAX,
2924 .input = rtnetlink_rcv,
2925 .cb_mutex = &rtnl_mutex,
2926 .flags = NL_CFG_F_NONROOT_RECV,
2929 sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
2936 static void __net_exit rtnetlink_net_exit(struct net *net)
2938 netlink_kernel_release(net->rtnl);
2942 static struct pernet_operations rtnetlink_net_ops = {
2943 .init = rtnetlink_net_init,
2944 .exit = rtnetlink_net_exit,
2947 void __init rtnetlink_init(void)
2949 if (register_pernet_subsys(&rtnetlink_net_ops))
2950 panic("rtnetlink_init: cannot initialize rtnetlink\n");
2952 register_netdevice_notifier(&rtnetlink_dev_notifier);
2954 rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
2955 rtnl_dump_ifinfo, rtnl_calcit);
2956 rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
2957 rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
2958 rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
2960 rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
2961 rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
2963 rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, NULL);
2964 rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, NULL);
2965 rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, NULL);
2967 rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, NULL);
2968 rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, NULL);
2969 rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, NULL);