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))
302 /* The check for setup is here because if ops
303 * does not have that filled up, it is not possible
304 * to use the ops for creating device. So do not
305 * fill up dellink as well. That disables rtnl_dellink.
307 if (ops->setup && !ops->dellink)
308 ops->dellink = unregister_netdevice_queue;
310 list_add_tail(&ops->list, &link_ops);
313 EXPORT_SYMBOL_GPL(__rtnl_link_register);
316 * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
317 * @ops: struct rtnl_link_ops * to register
319 * Returns 0 on success or a negative error code.
321 int rtnl_link_register(struct rtnl_link_ops *ops)
326 err = __rtnl_link_register(ops);
330 EXPORT_SYMBOL_GPL(rtnl_link_register);
332 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
334 struct net_device *dev;
335 LIST_HEAD(list_kill);
337 for_each_netdev(net, dev) {
338 if (dev->rtnl_link_ops == ops)
339 ops->dellink(dev, &list_kill);
341 unregister_netdevice_many(&list_kill);
345 * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
346 * @ops: struct rtnl_link_ops * to unregister
348 * The caller must hold the rtnl_mutex.
350 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
355 __rtnl_kill_links(net, ops);
357 list_del(&ops->list);
359 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
361 /* Return with the rtnl_lock held when there are no network
362 * devices unregistering in any network namespace.
364 static void rtnl_lock_unregistering_all(void)
371 prepare_to_wait(&netdev_unregistering_wq, &wait,
372 TASK_UNINTERRUPTIBLE);
373 unregistering = false;
376 if (net->dev_unreg_count > 0) {
377 unregistering = true;
386 finish_wait(&netdev_unregistering_wq, &wait);
390 * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
391 * @ops: struct rtnl_link_ops * to unregister
393 void rtnl_link_unregister(struct rtnl_link_ops *ops)
395 /* Close the race with cleanup_net() */
396 mutex_lock(&net_mutex);
397 rtnl_lock_unregistering_all();
398 __rtnl_link_unregister(ops);
400 mutex_unlock(&net_mutex);
402 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
404 static size_t rtnl_link_get_slave_info_data_size(const struct net_device *dev)
406 struct net_device *master_dev;
407 const struct rtnl_link_ops *ops;
409 master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
412 ops = master_dev->rtnl_link_ops;
413 if (!ops || !ops->get_slave_size)
415 /* IFLA_INFO_SLAVE_DATA + nested data */
416 return nla_total_size(sizeof(struct nlattr)) +
417 ops->get_slave_size(master_dev, dev);
420 static size_t rtnl_link_get_size(const struct net_device *dev)
422 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
428 size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
429 nla_total_size(strlen(ops->kind) + 1); /* IFLA_INFO_KIND */
432 /* IFLA_INFO_DATA + nested data */
433 size += nla_total_size(sizeof(struct nlattr)) +
436 if (ops->get_xstats_size)
437 /* IFLA_INFO_XSTATS */
438 size += nla_total_size(ops->get_xstats_size(dev));
440 size += rtnl_link_get_slave_info_data_size(dev);
445 static LIST_HEAD(rtnl_af_ops);
447 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
449 const struct rtnl_af_ops *ops;
451 list_for_each_entry(ops, &rtnl_af_ops, list) {
452 if (ops->family == family)
460 * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
461 * @ops: struct rtnl_af_ops * to register
463 * Returns 0 on success or a negative error code.
465 void rtnl_af_register(struct rtnl_af_ops *ops)
468 list_add_tail(&ops->list, &rtnl_af_ops);
471 EXPORT_SYMBOL_GPL(rtnl_af_register);
474 * __rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
475 * @ops: struct rtnl_af_ops * to unregister
477 * The caller must hold the rtnl_mutex.
479 void __rtnl_af_unregister(struct rtnl_af_ops *ops)
481 list_del(&ops->list);
483 EXPORT_SYMBOL_GPL(__rtnl_af_unregister);
486 * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
487 * @ops: struct rtnl_af_ops * to unregister
489 void rtnl_af_unregister(struct rtnl_af_ops *ops)
492 __rtnl_af_unregister(ops);
495 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
497 static size_t rtnl_link_get_af_size(const struct net_device *dev)
499 struct rtnl_af_ops *af_ops;
503 size = nla_total_size(sizeof(struct nlattr));
505 list_for_each_entry(af_ops, &rtnl_af_ops, list) {
506 if (af_ops->get_link_af_size) {
507 /* AF_* + nested data */
508 size += nla_total_size(sizeof(struct nlattr)) +
509 af_ops->get_link_af_size(dev);
516 static bool rtnl_have_link_slave_info(const struct net_device *dev)
518 struct net_device *master_dev;
520 master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
521 if (master_dev && master_dev->rtnl_link_ops)
526 static int rtnl_link_slave_info_fill(struct sk_buff *skb,
527 const struct net_device *dev)
529 struct net_device *master_dev;
530 const struct rtnl_link_ops *ops;
531 struct nlattr *slave_data;
534 master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
537 ops = master_dev->rtnl_link_ops;
540 if (nla_put_string(skb, IFLA_INFO_SLAVE_KIND, ops->kind) < 0)
542 if (ops->fill_slave_info) {
543 slave_data = nla_nest_start(skb, IFLA_INFO_SLAVE_DATA);
546 err = ops->fill_slave_info(skb, master_dev, dev);
548 goto err_cancel_slave_data;
549 nla_nest_end(skb, slave_data);
553 err_cancel_slave_data:
554 nla_nest_cancel(skb, slave_data);
558 static int rtnl_link_info_fill(struct sk_buff *skb,
559 const struct net_device *dev)
561 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
567 if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
569 if (ops->fill_xstats) {
570 err = ops->fill_xstats(skb, dev);
574 if (ops->fill_info) {
575 data = nla_nest_start(skb, IFLA_INFO_DATA);
578 err = ops->fill_info(skb, dev);
580 goto err_cancel_data;
581 nla_nest_end(skb, data);
586 nla_nest_cancel(skb, data);
590 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
592 struct nlattr *linkinfo;
595 linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
596 if (linkinfo == NULL)
599 err = rtnl_link_info_fill(skb, dev);
601 goto err_cancel_link;
603 err = rtnl_link_slave_info_fill(skb, dev);
605 goto err_cancel_link;
607 nla_nest_end(skb, linkinfo);
611 nla_nest_cancel(skb, linkinfo);
616 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
618 struct sock *rtnl = net->rtnl;
621 NETLINK_CB(skb).dst_group = group;
623 atomic_inc(&skb->users);
624 netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
626 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
630 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
632 struct sock *rtnl = net->rtnl;
634 return nlmsg_unicast(rtnl, skb, pid);
636 EXPORT_SYMBOL(rtnl_unicast);
638 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
639 struct nlmsghdr *nlh, gfp_t flags)
641 struct sock *rtnl = net->rtnl;
645 report = nlmsg_report(nlh);
647 nlmsg_notify(rtnl, skb, pid, group, report, flags);
649 EXPORT_SYMBOL(rtnl_notify);
651 void rtnl_set_sk_err(struct net *net, u32 group, int error)
653 struct sock *rtnl = net->rtnl;
655 netlink_set_err(rtnl, 0, group, error);
657 EXPORT_SYMBOL(rtnl_set_sk_err);
659 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
664 mx = nla_nest_start(skb, RTA_METRICS);
668 for (i = 0; i < RTAX_MAX; i++) {
671 if (nla_put_u32(skb, i+1, metrics[i]))
672 goto nla_put_failure;
677 nla_nest_cancel(skb, mx);
681 return nla_nest_end(skb, mx);
684 nla_nest_cancel(skb, mx);
687 EXPORT_SYMBOL(rtnetlink_put_metrics);
689 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
690 long expires, u32 error)
692 struct rta_cacheinfo ci = {
693 .rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse),
694 .rta_used = dst->__use,
695 .rta_clntref = atomic_read(&(dst->__refcnt)),
703 clock = jiffies_to_clock_t(abs(expires));
704 clock = min_t(unsigned long, clock, INT_MAX);
705 ci.rta_expires = (expires > 0) ? clock : -clock;
707 return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
709 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
711 static void set_operstate(struct net_device *dev, unsigned char transition)
713 unsigned char operstate = dev->operstate;
715 switch (transition) {
717 if ((operstate == IF_OPER_DORMANT ||
718 operstate == IF_OPER_UNKNOWN) &&
720 operstate = IF_OPER_UP;
723 case IF_OPER_DORMANT:
724 if (operstate == IF_OPER_UP ||
725 operstate == IF_OPER_UNKNOWN)
726 operstate = IF_OPER_DORMANT;
730 if (dev->operstate != operstate) {
731 write_lock_bh(&dev_base_lock);
732 dev->operstate = operstate;
733 write_unlock_bh(&dev_base_lock);
734 netdev_state_change(dev);
738 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
740 return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
741 (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
744 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
745 const struct ifinfomsg *ifm)
747 unsigned int flags = ifm->ifi_flags;
749 /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
751 flags = (flags & ifm->ifi_change) |
752 (rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
757 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
758 const struct rtnl_link_stats64 *b)
760 a->rx_packets = b->rx_packets;
761 a->tx_packets = b->tx_packets;
762 a->rx_bytes = b->rx_bytes;
763 a->tx_bytes = b->tx_bytes;
764 a->rx_errors = b->rx_errors;
765 a->tx_errors = b->tx_errors;
766 a->rx_dropped = b->rx_dropped;
767 a->tx_dropped = b->tx_dropped;
769 a->multicast = b->multicast;
770 a->collisions = b->collisions;
772 a->rx_length_errors = b->rx_length_errors;
773 a->rx_over_errors = b->rx_over_errors;
774 a->rx_crc_errors = b->rx_crc_errors;
775 a->rx_frame_errors = b->rx_frame_errors;
776 a->rx_fifo_errors = b->rx_fifo_errors;
777 a->rx_missed_errors = b->rx_missed_errors;
779 a->tx_aborted_errors = b->tx_aborted_errors;
780 a->tx_carrier_errors = b->tx_carrier_errors;
781 a->tx_fifo_errors = b->tx_fifo_errors;
782 a->tx_heartbeat_errors = b->tx_heartbeat_errors;
783 a->tx_window_errors = b->tx_window_errors;
785 a->rx_compressed = b->rx_compressed;
786 a->tx_compressed = b->tx_compressed;
789 static void copy_rtnl_link_stats64(void *v, const struct rtnl_link_stats64 *b)
791 memcpy(v, b, sizeof(*b));
795 static inline int rtnl_vfinfo_size(const struct net_device *dev,
798 if (dev->dev.parent && dev_is_pci(dev->dev.parent) &&
799 (ext_filter_mask & RTEXT_FILTER_VF)) {
800 int num_vfs = dev_num_vf(dev->dev.parent);
801 size_t size = nla_total_size(sizeof(struct nlattr));
802 size += nla_total_size(num_vfs * sizeof(struct nlattr));
804 (nla_total_size(sizeof(struct ifla_vf_mac)) +
805 nla_total_size(sizeof(struct ifla_vf_vlan)) +
806 nla_total_size(sizeof(struct ifla_vf_spoofchk)) +
807 nla_total_size(sizeof(struct ifla_vf_rate)));
813 static size_t rtnl_port_size(const struct net_device *dev,
816 size_t port_size = nla_total_size(4) /* PORT_VF */
817 + nla_total_size(PORT_PROFILE_MAX) /* PORT_PROFILE */
818 + nla_total_size(sizeof(struct ifla_port_vsi))
820 + nla_total_size(PORT_UUID_MAX) /* PORT_INSTANCE_UUID */
821 + nla_total_size(PORT_UUID_MAX) /* PORT_HOST_UUID */
822 + nla_total_size(1) /* PROT_VDP_REQUEST */
823 + nla_total_size(2); /* PORT_VDP_RESPONSE */
824 size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
825 size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
827 size_t port_self_size = nla_total_size(sizeof(struct nlattr))
830 if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
831 !(ext_filter_mask & RTEXT_FILTER_VF))
833 if (dev_num_vf(dev->dev.parent))
834 return port_self_size + vf_ports_size +
835 vf_port_size * dev_num_vf(dev->dev.parent);
837 return port_self_size;
840 static noinline size_t if_nlmsg_size(const struct net_device *dev,
843 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
844 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
845 + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
846 + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
847 + nla_total_size(sizeof(struct rtnl_link_ifmap))
848 + nla_total_size(sizeof(struct rtnl_link_stats))
849 + nla_total_size(sizeof(struct rtnl_link_stats64))
850 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
851 + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
852 + nla_total_size(4) /* IFLA_TXQLEN */
853 + nla_total_size(4) /* IFLA_WEIGHT */
854 + nla_total_size(4) /* IFLA_MTU */
855 + nla_total_size(4) /* IFLA_LINK */
856 + nla_total_size(4) /* IFLA_MASTER */
857 + nla_total_size(1) /* IFLA_CARRIER */
858 + nla_total_size(4) /* IFLA_PROMISCUITY */
859 + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
860 + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
861 + nla_total_size(1) /* IFLA_OPERSTATE */
862 + nla_total_size(1) /* IFLA_LINKMODE */
863 + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
864 + nla_total_size(ext_filter_mask
865 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
866 + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
867 + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
868 + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
869 + rtnl_link_get_af_size(dev) /* IFLA_AF_SPEC */
870 + nla_total_size(MAX_PHYS_PORT_ID_LEN); /* IFLA_PHYS_PORT_ID */
873 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
875 struct nlattr *vf_ports;
876 struct nlattr *vf_port;
880 vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
884 for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
885 vf_port = nla_nest_start(skb, IFLA_VF_PORT);
887 goto nla_put_failure;
888 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
889 goto nla_put_failure;
890 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
891 if (err == -EMSGSIZE)
892 goto nla_put_failure;
894 nla_nest_cancel(skb, vf_port);
897 nla_nest_end(skb, vf_port);
900 nla_nest_end(skb, vf_ports);
905 nla_nest_cancel(skb, vf_ports);
909 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
911 struct nlattr *port_self;
914 port_self = nla_nest_start(skb, IFLA_PORT_SELF);
918 err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
920 nla_nest_cancel(skb, port_self);
921 return (err == -EMSGSIZE) ? err : 0;
924 nla_nest_end(skb, port_self);
929 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
934 if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
935 !(ext_filter_mask & RTEXT_FILTER_VF))
938 err = rtnl_port_self_fill(skb, dev);
942 if (dev_num_vf(dev->dev.parent)) {
943 err = rtnl_vf_ports_fill(skb, dev);
951 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
954 struct netdev_phys_port_id ppid;
956 err = dev_get_phys_port_id(dev, &ppid);
958 if (err == -EOPNOTSUPP)
963 if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
969 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
970 int type, u32 pid, u32 seq, u32 change,
971 unsigned int flags, u32 ext_filter_mask)
973 struct ifinfomsg *ifm;
974 struct nlmsghdr *nlh;
975 struct rtnl_link_stats64 temp;
976 const struct rtnl_link_stats64 *stats;
977 struct nlattr *attr, *af_spec;
978 struct rtnl_af_ops *af_ops;
979 struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
982 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
986 ifm = nlmsg_data(nlh);
987 ifm->ifi_family = AF_UNSPEC;
989 ifm->ifi_type = dev->type;
990 ifm->ifi_index = dev->ifindex;
991 ifm->ifi_flags = dev_get_flags(dev);
992 ifm->ifi_change = change;
994 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
995 nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
996 nla_put_u8(skb, IFLA_OPERSTATE,
997 netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
998 nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
999 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
1000 nla_put_u32(skb, IFLA_GROUP, dev->group) ||
1001 nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
1002 nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
1004 nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1006 (dev->ifindex != dev->iflink &&
1007 nla_put_u32(skb, IFLA_LINK, dev->iflink)) ||
1009 nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex)) ||
1010 nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1012 nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
1014 nla_put_string(skb, IFLA_IFALIAS, dev->ifalias)) ||
1015 nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1016 atomic_read(&dev->carrier_changes)))
1017 goto nla_put_failure;
1020 struct rtnl_link_ifmap map = {
1021 .mem_start = dev->mem_start,
1022 .mem_end = dev->mem_end,
1023 .base_addr = dev->base_addr,
1026 .port = dev->if_port,
1028 if (nla_put(skb, IFLA_MAP, sizeof(map), &map))
1029 goto nla_put_failure;
1032 if (dev->addr_len) {
1033 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1034 nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1035 goto nla_put_failure;
1038 if (rtnl_phys_port_id_fill(skb, dev))
1039 goto nla_put_failure;
1041 attr = nla_reserve(skb, IFLA_STATS,
1042 sizeof(struct rtnl_link_stats));
1044 goto nla_put_failure;
1046 stats = dev_get_stats(dev, &temp);
1047 copy_rtnl_link_stats(nla_data(attr), stats);
1049 attr = nla_reserve(skb, IFLA_STATS64,
1050 sizeof(struct rtnl_link_stats64));
1052 goto nla_put_failure;
1053 copy_rtnl_link_stats64(nla_data(attr), stats);
1055 if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF) &&
1056 nla_put_u32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent)))
1057 goto nla_put_failure;
1059 if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent
1060 && (ext_filter_mask & RTEXT_FILTER_VF)) {
1063 struct nlattr *vfinfo, *vf;
1064 int num_vfs = dev_num_vf(dev->dev.parent);
1066 vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
1068 goto nla_put_failure;
1069 for (i = 0; i < num_vfs; i++) {
1070 struct ifla_vf_info ivi;
1071 struct ifla_vf_mac vf_mac;
1072 struct ifla_vf_vlan vf_vlan;
1073 struct ifla_vf_rate vf_rate;
1074 struct ifla_vf_tx_rate vf_tx_rate;
1075 struct ifla_vf_spoofchk vf_spoofchk;
1076 struct ifla_vf_link_state vf_linkstate;
1079 * Not all SR-IOV capable drivers support the
1080 * spoofcheck query. Preset to -1 so the user
1081 * space tool can detect that the driver didn't
1085 memset(ivi.mac, 0, sizeof(ivi.mac));
1086 /* The default value for VF link state is "auto"
1087 * IFLA_VF_LINK_STATE_AUTO which equals zero
1090 if (dev->netdev_ops->ndo_get_vf_config(dev, i, &ivi))
1097 vf_linkstate.vf = ivi.vf;
1099 memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1100 vf_vlan.vlan = ivi.vlan;
1101 vf_vlan.qos = ivi.qos;
1102 vf_tx_rate.rate = ivi.max_tx_rate;
1103 vf_rate.min_tx_rate = ivi.min_tx_rate;
1104 vf_rate.max_tx_rate = ivi.max_tx_rate;
1105 vf_spoofchk.setting = ivi.spoofchk;
1106 vf_linkstate.link_state = ivi.linkstate;
1107 vf = nla_nest_start(skb, IFLA_VF_INFO);
1109 nla_nest_cancel(skb, vfinfo);
1110 goto nla_put_failure;
1112 if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1113 nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1114 nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1116 nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1118 nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1120 nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1122 goto nla_put_failure;
1123 nla_nest_end(skb, vf);
1125 nla_nest_end(skb, vfinfo);
1128 if (rtnl_port_fill(skb, dev, ext_filter_mask))
1129 goto nla_put_failure;
1131 if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1132 if (rtnl_link_fill(skb, dev) < 0)
1133 goto nla_put_failure;
1136 if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1137 goto nla_put_failure;
1139 list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1140 if (af_ops->fill_link_af) {
1144 if (!(af = nla_nest_start(skb, af_ops->family)))
1145 goto nla_put_failure;
1147 err = af_ops->fill_link_af(skb, dev);
1150 * Caller may return ENODATA to indicate that there
1151 * was no data to be dumped. This is not an error, it
1152 * means we should trim the attribute header and
1155 if (err == -ENODATA)
1156 nla_nest_cancel(skb, af);
1158 goto nla_put_failure;
1160 nla_nest_end(skb, af);
1164 nla_nest_end(skb, af_spec);
1166 return nlmsg_end(skb, nlh);
1169 nlmsg_cancel(skb, nlh);
1173 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1174 [IFLA_IFNAME] = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1175 [IFLA_ADDRESS] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1176 [IFLA_BROADCAST] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1177 [IFLA_MAP] = { .len = sizeof(struct rtnl_link_ifmap) },
1178 [IFLA_MTU] = { .type = NLA_U32 },
1179 [IFLA_LINK] = { .type = NLA_U32 },
1180 [IFLA_MASTER] = { .type = NLA_U32 },
1181 [IFLA_CARRIER] = { .type = NLA_U8 },
1182 [IFLA_TXQLEN] = { .type = NLA_U32 },
1183 [IFLA_WEIGHT] = { .type = NLA_U32 },
1184 [IFLA_OPERSTATE] = { .type = NLA_U8 },
1185 [IFLA_LINKMODE] = { .type = NLA_U8 },
1186 [IFLA_LINKINFO] = { .type = NLA_NESTED },
1187 [IFLA_NET_NS_PID] = { .type = NLA_U32 },
1188 [IFLA_NET_NS_FD] = { .type = NLA_U32 },
1189 [IFLA_IFALIAS] = { .type = NLA_STRING, .len = IFALIASZ-1 },
1190 [IFLA_VFINFO_LIST] = {. type = NLA_NESTED },
1191 [IFLA_VF_PORTS] = { .type = NLA_NESTED },
1192 [IFLA_PORT_SELF] = { .type = NLA_NESTED },
1193 [IFLA_AF_SPEC] = { .type = NLA_NESTED },
1194 [IFLA_EXT_MASK] = { .type = NLA_U32 },
1195 [IFLA_PROMISCUITY] = { .type = NLA_U32 },
1196 [IFLA_NUM_TX_QUEUES] = { .type = NLA_U32 },
1197 [IFLA_NUM_RX_QUEUES] = { .type = NLA_U32 },
1198 [IFLA_PHYS_PORT_ID] = { .type = NLA_BINARY, .len = MAX_PHYS_PORT_ID_LEN },
1199 [IFLA_CARRIER_CHANGES] = { .type = NLA_U32 }, /* ignored */
1202 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1203 [IFLA_INFO_KIND] = { .type = NLA_STRING },
1204 [IFLA_INFO_DATA] = { .type = NLA_NESTED },
1205 [IFLA_INFO_SLAVE_KIND] = { .type = NLA_STRING },
1206 [IFLA_INFO_SLAVE_DATA] = { .type = NLA_NESTED },
1209 static const struct nla_policy ifla_vfinfo_policy[IFLA_VF_INFO_MAX+1] = {
1210 [IFLA_VF_INFO] = { .type = NLA_NESTED },
1213 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1214 [IFLA_VF_MAC] = { .type = NLA_BINARY,
1215 .len = sizeof(struct ifla_vf_mac) },
1216 [IFLA_VF_VLAN] = { .type = NLA_BINARY,
1217 .len = sizeof(struct ifla_vf_vlan) },
1218 [IFLA_VF_TX_RATE] = { .type = NLA_BINARY,
1219 .len = sizeof(struct ifla_vf_tx_rate) },
1220 [IFLA_VF_SPOOFCHK] = { .type = NLA_BINARY,
1221 .len = sizeof(struct ifla_vf_spoofchk) },
1222 [IFLA_VF_RATE] = { .type = NLA_BINARY,
1223 .len = sizeof(struct ifla_vf_rate) },
1224 [IFLA_VF_LINK_STATE] = { .type = NLA_BINARY,
1225 .len = sizeof(struct ifla_vf_link_state) },
1228 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1229 [IFLA_PORT_VF] = { .type = NLA_U32 },
1230 [IFLA_PORT_PROFILE] = { .type = NLA_STRING,
1231 .len = PORT_PROFILE_MAX },
1232 [IFLA_PORT_VSI_TYPE] = { .type = NLA_BINARY,
1233 .len = sizeof(struct ifla_port_vsi)},
1234 [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1235 .len = PORT_UUID_MAX },
1236 [IFLA_PORT_HOST_UUID] = { .type = NLA_STRING,
1237 .len = PORT_UUID_MAX },
1238 [IFLA_PORT_REQUEST] = { .type = NLA_U8, },
1239 [IFLA_PORT_RESPONSE] = { .type = NLA_U16, },
1242 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1244 struct net *net = sock_net(skb->sk);
1247 struct net_device *dev;
1248 struct hlist_head *head;
1249 struct nlattr *tb[IFLA_MAX+1];
1250 u32 ext_filter_mask = 0;
1255 s_idx = cb->args[1];
1258 cb->seq = net->dev_base_seq;
1260 /* A hack to preserve kernel<->userspace interface.
1261 * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
1262 * However, before Linux v3.9 the code here assumed rtgenmsg and that's
1263 * what iproute2 < v3.9.0 used.
1264 * We can detect the old iproute2. Even including the IFLA_EXT_MASK
1265 * attribute, its netlink message is shorter than struct ifinfomsg.
1267 hdrlen = nlmsg_len(cb->nlh) < sizeof(struct ifinfomsg) ?
1268 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
1270 if (nlmsg_parse(cb->nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
1272 if (tb[IFLA_EXT_MASK])
1273 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1276 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1278 head = &net->dev_index_head[h];
1279 hlist_for_each_entry_rcu(dev, head, index_hlist) {
1282 err = rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1283 NETLINK_CB(cb->skb).portid,
1284 cb->nlh->nlmsg_seq, 0,
1287 /* If we ran out of room on the first message,
1290 WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
1295 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1308 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len)
1310 return nla_parse(tb, IFLA_MAX, head, len, ifla_policy);
1312 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
1314 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1317 /* Examine the link attributes and figure out which
1318 * network namespace we are talking about.
1320 if (tb[IFLA_NET_NS_PID])
1321 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1322 else if (tb[IFLA_NET_NS_FD])
1323 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1325 net = get_net(src_net);
1328 EXPORT_SYMBOL(rtnl_link_get_net);
1330 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1333 if (tb[IFLA_ADDRESS] &&
1334 nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1337 if (tb[IFLA_BROADCAST] &&
1338 nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1342 if (tb[IFLA_AF_SPEC]) {
1346 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1347 const struct rtnl_af_ops *af_ops;
1349 if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1350 return -EAFNOSUPPORT;
1352 if (!af_ops->set_link_af)
1355 if (af_ops->validate_link_af) {
1356 err = af_ops->validate_link_af(dev, af);
1366 static int do_setvfinfo(struct net_device *dev, struct nlattr *attr)
1368 int rem, err = -EINVAL;
1370 const struct net_device_ops *ops = dev->netdev_ops;
1372 nla_for_each_nested(vf, attr, rem) {
1373 switch (nla_type(vf)) {
1375 struct ifla_vf_mac *ivm;
1378 if (ops->ndo_set_vf_mac)
1379 err = ops->ndo_set_vf_mac(dev, ivm->vf,
1383 case IFLA_VF_VLAN: {
1384 struct ifla_vf_vlan *ivv;
1387 if (ops->ndo_set_vf_vlan)
1388 err = ops->ndo_set_vf_vlan(dev, ivv->vf,
1393 case IFLA_VF_TX_RATE: {
1394 struct ifla_vf_tx_rate *ivt;
1395 struct ifla_vf_info ivf;
1398 if (ops->ndo_get_vf_config)
1399 err = ops->ndo_get_vf_config(dev, ivt->vf,
1404 if (ops->ndo_set_vf_rate)
1405 err = ops->ndo_set_vf_rate(dev, ivt->vf,
1410 case IFLA_VF_RATE: {
1411 struct ifla_vf_rate *ivt;
1414 if (ops->ndo_set_vf_rate)
1415 err = ops->ndo_set_vf_rate(dev, ivt->vf,
1420 case IFLA_VF_SPOOFCHK: {
1421 struct ifla_vf_spoofchk *ivs;
1424 if (ops->ndo_set_vf_spoofchk)
1425 err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1429 case IFLA_VF_LINK_STATE: {
1430 struct ifla_vf_link_state *ivl;
1433 if (ops->ndo_set_vf_link_state)
1434 err = ops->ndo_set_vf_link_state(dev, ivl->vf,
1448 static int do_set_master(struct net_device *dev, int ifindex)
1450 struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1451 const struct net_device_ops *ops;
1455 if (upper_dev->ifindex == ifindex)
1457 ops = upper_dev->netdev_ops;
1458 if (ops->ndo_del_slave) {
1459 err = ops->ndo_del_slave(upper_dev, dev);
1468 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
1471 ops = upper_dev->netdev_ops;
1472 if (ops->ndo_add_slave) {
1473 err = ops->ndo_add_slave(upper_dev, dev);
1483 static int do_setlink(const struct sk_buff *skb,
1484 struct net_device *dev, struct ifinfomsg *ifm,
1485 struct nlattr **tb, char *ifname, int modified)
1487 const struct net_device_ops *ops = dev->netdev_ops;
1490 if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1491 struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1496 if (!netlink_ns_capable(skb, net->user_ns, CAP_NET_ADMIN)) {
1500 err = dev_change_net_namespace(dev, net, ifname);
1508 struct rtnl_link_ifmap *u_map;
1511 if (!ops->ndo_set_config) {
1516 if (!netif_device_present(dev)) {
1521 u_map = nla_data(tb[IFLA_MAP]);
1522 k_map.mem_start = (unsigned long) u_map->mem_start;
1523 k_map.mem_end = (unsigned long) u_map->mem_end;
1524 k_map.base_addr = (unsigned short) u_map->base_addr;
1525 k_map.irq = (unsigned char) u_map->irq;
1526 k_map.dma = (unsigned char) u_map->dma;
1527 k_map.port = (unsigned char) u_map->port;
1529 err = ops->ndo_set_config(dev, &k_map);
1536 if (tb[IFLA_ADDRESS]) {
1537 struct sockaddr *sa;
1540 len = sizeof(sa_family_t) + dev->addr_len;
1541 sa = kmalloc(len, GFP_KERNEL);
1546 sa->sa_family = dev->type;
1547 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1549 err = dev_set_mac_address(dev, sa);
1557 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1563 if (tb[IFLA_GROUP]) {
1564 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1569 * Interface selected by interface index but interface
1570 * name provided implies that a name change has been
1573 if (ifm->ifi_index > 0 && ifname[0]) {
1574 err = dev_change_name(dev, ifname);
1580 if (tb[IFLA_IFALIAS]) {
1581 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
1582 nla_len(tb[IFLA_IFALIAS]));
1588 if (tb[IFLA_BROADCAST]) {
1589 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
1590 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
1593 if (ifm->ifi_flags || ifm->ifi_change) {
1594 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1599 if (tb[IFLA_MASTER]) {
1600 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
1606 if (tb[IFLA_CARRIER]) {
1607 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
1613 if (tb[IFLA_TXQLEN])
1614 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1616 if (tb[IFLA_OPERSTATE])
1617 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1619 if (tb[IFLA_LINKMODE]) {
1620 write_lock_bh(&dev_base_lock);
1621 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1622 write_unlock_bh(&dev_base_lock);
1625 if (tb[IFLA_VFINFO_LIST]) {
1626 struct nlattr *attr;
1628 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
1629 if (nla_type(attr) != IFLA_VF_INFO) {
1633 err = do_setvfinfo(dev, attr);
1641 if (tb[IFLA_VF_PORTS]) {
1642 struct nlattr *port[IFLA_PORT_MAX+1];
1643 struct nlattr *attr;
1648 if (!ops->ndo_set_vf_port)
1651 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
1652 if (nla_type(attr) != IFLA_VF_PORT)
1654 err = nla_parse_nested(port, IFLA_PORT_MAX,
1655 attr, ifla_port_policy);
1658 if (!port[IFLA_PORT_VF]) {
1662 vf = nla_get_u32(port[IFLA_PORT_VF]);
1663 err = ops->ndo_set_vf_port(dev, vf, port);
1671 if (tb[IFLA_PORT_SELF]) {
1672 struct nlattr *port[IFLA_PORT_MAX+1];
1674 err = nla_parse_nested(port, IFLA_PORT_MAX,
1675 tb[IFLA_PORT_SELF], ifla_port_policy);
1680 if (ops->ndo_set_vf_port)
1681 err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
1687 if (tb[IFLA_AF_SPEC]) {
1691 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1692 const struct rtnl_af_ops *af_ops;
1694 if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1697 err = af_ops->set_link_af(dev, af);
1707 if (err < 0 && modified)
1708 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",
1714 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
1716 struct net *net = sock_net(skb->sk);
1717 struct ifinfomsg *ifm;
1718 struct net_device *dev;
1720 struct nlattr *tb[IFLA_MAX+1];
1721 char ifname[IFNAMSIZ];
1723 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1727 if (tb[IFLA_IFNAME])
1728 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1733 ifm = nlmsg_data(nlh);
1734 if (ifm->ifi_index > 0)
1735 dev = __dev_get_by_index(net, ifm->ifi_index);
1736 else if (tb[IFLA_IFNAME])
1737 dev = __dev_get_by_name(net, ifname);
1746 err = validate_linkmsg(dev, tb);
1750 err = do_setlink(skb, dev, ifm, tb, ifname, 0);
1755 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
1757 struct net *net = sock_net(skb->sk);
1758 const struct rtnl_link_ops *ops;
1759 struct net_device *dev;
1760 struct ifinfomsg *ifm;
1761 char ifname[IFNAMSIZ];
1762 struct nlattr *tb[IFLA_MAX+1];
1764 LIST_HEAD(list_kill);
1766 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1770 if (tb[IFLA_IFNAME])
1771 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1773 ifm = nlmsg_data(nlh);
1774 if (ifm->ifi_index > 0)
1775 dev = __dev_get_by_index(net, ifm->ifi_index);
1776 else if (tb[IFLA_IFNAME])
1777 dev = __dev_get_by_name(net, ifname);
1784 ops = dev->rtnl_link_ops;
1785 if (!ops || !ops->dellink)
1788 ops->dellink(dev, &list_kill);
1789 unregister_netdevice_many(&list_kill);
1793 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
1795 unsigned int old_flags;
1798 old_flags = dev->flags;
1799 if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
1800 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1805 dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
1807 __dev_notify_flags(dev, old_flags, ~0U);
1810 EXPORT_SYMBOL(rtnl_configure_link);
1812 struct net_device *rtnl_create_link(struct net *net,
1813 char *ifname, unsigned char name_assign_type,
1814 const struct rtnl_link_ops *ops, struct nlattr *tb[])
1817 struct net_device *dev;
1818 unsigned int num_tx_queues = 1;
1819 unsigned int num_rx_queues = 1;
1821 if (tb[IFLA_NUM_TX_QUEUES])
1822 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
1823 else if (ops->get_num_tx_queues)
1824 num_tx_queues = ops->get_num_tx_queues();
1826 if (tb[IFLA_NUM_RX_QUEUES])
1827 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
1828 else if (ops->get_num_rx_queues)
1829 num_rx_queues = ops->get_num_rx_queues();
1832 dev = alloc_netdev_mqs(ops->priv_size, ifname, name_assign_type,
1833 ops->setup, num_tx_queues, num_rx_queues);
1837 dev_net_set(dev, net);
1838 dev->rtnl_link_ops = ops;
1839 dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
1842 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
1843 if (tb[IFLA_ADDRESS]) {
1844 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
1845 nla_len(tb[IFLA_ADDRESS]));
1846 dev->addr_assign_type = NET_ADDR_SET;
1848 if (tb[IFLA_BROADCAST])
1849 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
1850 nla_len(tb[IFLA_BROADCAST]));
1851 if (tb[IFLA_TXQLEN])
1852 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1853 if (tb[IFLA_OPERSTATE])
1854 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1855 if (tb[IFLA_LINKMODE])
1856 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1858 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1863 return ERR_PTR(err);
1865 EXPORT_SYMBOL(rtnl_create_link);
1867 static int rtnl_group_changelink(const struct sk_buff *skb,
1868 struct net *net, int group,
1869 struct ifinfomsg *ifm,
1872 struct net_device *dev;
1875 for_each_netdev(net, dev) {
1876 if (dev->group == group) {
1877 err = do_setlink(skb, dev, ifm, tb, NULL, 0);
1886 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh)
1888 struct net *net = sock_net(skb->sk);
1889 const struct rtnl_link_ops *ops;
1890 const struct rtnl_link_ops *m_ops = NULL;
1891 struct net_device *dev;
1892 struct net_device *master_dev = NULL;
1893 struct ifinfomsg *ifm;
1894 char kind[MODULE_NAME_LEN];
1895 char ifname[IFNAMSIZ];
1896 struct nlattr *tb[IFLA_MAX+1];
1897 struct nlattr *linkinfo[IFLA_INFO_MAX+1];
1898 unsigned char name_assign_type = NET_NAME_USER;
1901 #ifdef CONFIG_MODULES
1904 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1908 if (tb[IFLA_IFNAME])
1909 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1913 ifm = nlmsg_data(nlh);
1914 if (ifm->ifi_index > 0)
1915 dev = __dev_get_by_index(net, ifm->ifi_index);
1918 dev = __dev_get_by_name(net, ifname);
1924 master_dev = netdev_master_upper_dev_get(dev);
1926 m_ops = master_dev->rtnl_link_ops;
1929 err = validate_linkmsg(dev, tb);
1933 if (tb[IFLA_LINKINFO]) {
1934 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
1935 tb[IFLA_LINKINFO], ifla_info_policy);
1939 memset(linkinfo, 0, sizeof(linkinfo));
1941 if (linkinfo[IFLA_INFO_KIND]) {
1942 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
1943 ops = rtnl_link_ops_get(kind);
1950 struct nlattr *attr[ops ? ops->maxtype + 1 : 0];
1951 struct nlattr *slave_attr[m_ops ? m_ops->slave_maxtype + 1 : 0];
1952 struct nlattr **data = NULL;
1953 struct nlattr **slave_data = NULL;
1954 struct net *dest_net;
1957 if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
1958 err = nla_parse_nested(attr, ops->maxtype,
1959 linkinfo[IFLA_INFO_DATA],
1965 if (ops->validate) {
1966 err = ops->validate(tb, data);
1973 if (m_ops->slave_maxtype &&
1974 linkinfo[IFLA_INFO_SLAVE_DATA]) {
1975 err = nla_parse_nested(slave_attr,
1976 m_ops->slave_maxtype,
1977 linkinfo[IFLA_INFO_SLAVE_DATA],
1978 m_ops->slave_policy);
1981 slave_data = slave_attr;
1983 if (m_ops->slave_validate) {
1984 err = m_ops->slave_validate(tb, slave_data);
1993 if (nlh->nlmsg_flags & NLM_F_EXCL)
1995 if (nlh->nlmsg_flags & NLM_F_REPLACE)
1998 if (linkinfo[IFLA_INFO_DATA]) {
1999 if (!ops || ops != dev->rtnl_link_ops ||
2003 err = ops->changelink(dev, tb, data);
2009 if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
2010 if (!m_ops || !m_ops->slave_changelink)
2013 err = m_ops->slave_changelink(master_dev, dev,
2020 return do_setlink(skb, dev, ifm, tb, ifname, modified);
2023 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
2024 if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
2025 return rtnl_group_changelink(skb, net,
2026 nla_get_u32(tb[IFLA_GROUP]),
2031 if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
2035 #ifdef CONFIG_MODULES
2038 request_module("rtnl-link-%s", kind);
2040 ops = rtnl_link_ops_get(kind);
2052 snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
2053 name_assign_type = NET_NAME_ENUM;
2056 dest_net = rtnl_link_get_net(net, tb);
2057 if (IS_ERR(dest_net))
2058 return PTR_ERR(dest_net);
2060 dev = rtnl_create_link(dest_net, ifname, name_assign_type, ops, tb);
2066 dev->ifindex = ifm->ifi_index;
2069 err = ops->newlink(net, dev, tb, data);
2070 /* Drivers should call free_netdev() in ->destructor
2071 * and unregister it on failure after registration
2072 * so that device could be finally freed in rtnl_unlock.
2075 /* If device is not registered at all, free it now */
2076 if (dev->reg_state == NETREG_UNINITIALIZED)
2081 err = register_netdevice(dev);
2087 err = rtnl_configure_link(dev, ifm);
2089 unregister_netdevice(dev);
2096 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh)
2098 struct net *net = sock_net(skb->sk);
2099 struct ifinfomsg *ifm;
2100 char ifname[IFNAMSIZ];
2101 struct nlattr *tb[IFLA_MAX+1];
2102 struct net_device *dev = NULL;
2103 struct sk_buff *nskb;
2105 u32 ext_filter_mask = 0;
2107 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2111 if (tb[IFLA_IFNAME])
2112 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2114 if (tb[IFLA_EXT_MASK])
2115 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2117 ifm = nlmsg_data(nlh);
2118 if (ifm->ifi_index > 0)
2119 dev = __dev_get_by_index(net, ifm->ifi_index);
2120 else if (tb[IFLA_IFNAME])
2121 dev = __dev_get_by_name(net, ifname);
2128 nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
2132 err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).portid,
2133 nlh->nlmsg_seq, 0, 0, ext_filter_mask);
2135 /* -EMSGSIZE implies BUG in if_nlmsg_size */
2136 WARN_ON(err == -EMSGSIZE);
2139 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
2144 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
2146 struct net *net = sock_net(skb->sk);
2147 struct net_device *dev;
2148 struct nlattr *tb[IFLA_MAX+1];
2149 u32 ext_filter_mask = 0;
2150 u16 min_ifinfo_dump_size = 0;
2153 /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
2154 hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
2155 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
2157 if (nlmsg_parse(nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
2158 if (tb[IFLA_EXT_MASK])
2159 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2162 if (!ext_filter_mask)
2163 return NLMSG_GOODSIZE;
2165 * traverse the list of net devices and compute the minimum
2166 * buffer size based upon the filter mask.
2168 list_for_each_entry(dev, &net->dev_base_head, dev_list) {
2169 min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
2174 return min_ifinfo_dump_size;
2177 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
2180 int s_idx = cb->family;
2184 for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
2185 int type = cb->nlh->nlmsg_type-RTM_BASE;
2186 if (idx < s_idx || idx == PF_PACKET)
2188 if (rtnl_msg_handlers[idx] == NULL ||
2189 rtnl_msg_handlers[idx][type].dumpit == NULL)
2192 memset(&cb->args[0], 0, sizeof(cb->args));
2196 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
2204 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
2207 struct net *net = dev_net(dev);
2208 struct sk_buff *skb;
2210 size_t if_info_size;
2212 skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
2216 err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0);
2218 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
2219 WARN_ON(err == -EMSGSIZE);
2223 rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
2227 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2229 EXPORT_SYMBOL(rtmsg_ifinfo);
2231 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
2232 struct net_device *dev,
2233 u8 *addr, u32 pid, u32 seq,
2234 int type, unsigned int flags,
2237 struct nlmsghdr *nlh;
2240 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
2244 ndm = nlmsg_data(nlh);
2245 ndm->ndm_family = AF_BRIDGE;
2248 ndm->ndm_flags = flags;
2250 ndm->ndm_ifindex = dev->ifindex;
2251 ndm->ndm_state = NUD_PERMANENT;
2253 if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
2254 goto nla_put_failure;
2256 return nlmsg_end(skb, nlh);
2259 nlmsg_cancel(skb, nlh);
2263 static inline size_t rtnl_fdb_nlmsg_size(void)
2265 return NLMSG_ALIGN(sizeof(struct ndmsg)) + nla_total_size(ETH_ALEN);
2268 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, int type)
2270 struct net *net = dev_net(dev);
2271 struct sk_buff *skb;
2274 skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
2278 err = nlmsg_populate_fdb_fill(skb, dev, addr, 0, 0, type, NTF_SELF, 0);
2284 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2287 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
2291 * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
2293 int ndo_dflt_fdb_add(struct ndmsg *ndm,
2294 struct nlattr *tb[],
2295 struct net_device *dev,
2296 const unsigned char *addr,
2301 /* If aging addresses are supported device will need to
2302 * implement its own handler for this.
2304 if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
2305 pr_info("%s: FDB only supports static addresses\n", dev->name);
2309 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2310 err = dev_uc_add_excl(dev, addr);
2311 else if (is_multicast_ether_addr(addr))
2312 err = dev_mc_add_excl(dev, addr);
2314 /* Only return duplicate errors if NLM_F_EXCL is set */
2315 if (err == -EEXIST && !(flags & NLM_F_EXCL))
2320 EXPORT_SYMBOL(ndo_dflt_fdb_add);
2322 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh)
2324 struct net *net = sock_net(skb->sk);
2326 struct nlattr *tb[NDA_MAX+1];
2327 struct net_device *dev;
2331 err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2335 ndm = nlmsg_data(nlh);
2336 if (ndm->ndm_ifindex == 0) {
2337 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n");
2341 dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2343 pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n");
2347 if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2348 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n");
2352 addr = nla_data(tb[NDA_LLADDR]);
2356 /* Support fdb on master device the net/bridge default case */
2357 if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2358 (dev->priv_flags & IFF_BRIDGE_PORT)) {
2359 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2360 const struct net_device_ops *ops = br_dev->netdev_ops;
2362 err = ops->ndo_fdb_add(ndm, tb, dev, addr, nlh->nlmsg_flags);
2366 ndm->ndm_flags &= ~NTF_MASTER;
2369 /* Embedded bridge, macvlan, and any other device support */
2370 if ((ndm->ndm_flags & NTF_SELF)) {
2371 if (dev->netdev_ops->ndo_fdb_add)
2372 err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
2375 err = ndo_dflt_fdb_add(ndm, tb, dev, addr,
2379 rtnl_fdb_notify(dev, addr, RTM_NEWNEIGH);
2380 ndm->ndm_flags &= ~NTF_SELF;
2388 * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
2390 int ndo_dflt_fdb_del(struct ndmsg *ndm,
2391 struct nlattr *tb[],
2392 struct net_device *dev,
2393 const unsigned char *addr)
2397 /* If aging addresses are supported device will need to
2398 * implement its own handler for this.
2400 if (!(ndm->ndm_state & NUD_PERMANENT)) {
2401 pr_info("%s: FDB only supports static addresses\n", dev->name);
2405 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2406 err = dev_uc_del(dev, addr);
2407 else if (is_multicast_ether_addr(addr))
2408 err = dev_mc_del(dev, addr);
2412 EXPORT_SYMBOL(ndo_dflt_fdb_del);
2414 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh)
2416 struct net *net = sock_net(skb->sk);
2418 struct nlattr *tb[NDA_MAX+1];
2419 struct net_device *dev;
2423 if (!netlink_capable(skb, CAP_NET_ADMIN))
2426 err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2430 ndm = nlmsg_data(nlh);
2431 if (ndm->ndm_ifindex == 0) {
2432 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n");
2436 dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2438 pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n");
2442 if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2443 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid address\n");
2447 addr = nla_data(tb[NDA_LLADDR]);
2451 /* Support fdb on master device the net/bridge default case */
2452 if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2453 (dev->priv_flags & IFF_BRIDGE_PORT)) {
2454 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2455 const struct net_device_ops *ops = br_dev->netdev_ops;
2457 if (ops->ndo_fdb_del)
2458 err = ops->ndo_fdb_del(ndm, tb, dev, addr);
2463 ndm->ndm_flags &= ~NTF_MASTER;
2466 /* Embedded bridge, macvlan, and any other device support */
2467 if (ndm->ndm_flags & NTF_SELF) {
2468 if (dev->netdev_ops->ndo_fdb_del)
2469 err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr);
2471 err = ndo_dflt_fdb_del(ndm, tb, dev, addr);
2474 rtnl_fdb_notify(dev, addr, RTM_DELNEIGH);
2475 ndm->ndm_flags &= ~NTF_SELF;
2482 static int nlmsg_populate_fdb(struct sk_buff *skb,
2483 struct netlink_callback *cb,
2484 struct net_device *dev,
2486 struct netdev_hw_addr_list *list)
2488 struct netdev_hw_addr *ha;
2492 portid = NETLINK_CB(cb->skb).portid;
2493 seq = cb->nlh->nlmsg_seq;
2495 list_for_each_entry(ha, &list->list, list) {
2496 if (*idx < cb->args[0])
2499 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr,
2501 RTM_NEWNEIGH, NTF_SELF,
2512 * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
2513 * @nlh: netlink message header
2516 * Default netdevice operation to dump the existing unicast address list.
2517 * Returns number of addresses from list put in skb.
2519 int ndo_dflt_fdb_dump(struct sk_buff *skb,
2520 struct netlink_callback *cb,
2521 struct net_device *dev,
2522 struct net_device *filter_dev,
2527 netif_addr_lock_bh(dev);
2528 err = nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->uc);
2531 nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->mc);
2533 netif_addr_unlock_bh(dev);
2536 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
2538 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
2540 struct net_device *dev;
2541 struct nlattr *tb[IFLA_MAX+1];
2542 struct net_device *bdev = NULL;
2543 struct net_device *br_dev = NULL;
2544 const struct net_device_ops *ops = NULL;
2545 const struct net_device_ops *cops = NULL;
2546 struct ifinfomsg *ifm = nlmsg_data(cb->nlh);
2547 struct net *net = sock_net(skb->sk);
2552 if (nlmsg_parse(cb->nlh, sizeof(struct ifinfomsg), tb, IFLA_MAX,
2553 ifla_policy) == 0) {
2554 if (tb[IFLA_MASTER])
2555 br_idx = nla_get_u32(tb[IFLA_MASTER]);
2558 brport_idx = ifm->ifi_index;
2561 br_dev = __dev_get_by_index(net, br_idx);
2565 ops = br_dev->netdev_ops;
2569 for_each_netdev(net, dev) {
2570 if (brport_idx && (dev->ifindex != brport_idx))
2573 if (!br_idx) { /* user did not specify a specific bridge */
2574 if (dev->priv_flags & IFF_BRIDGE_PORT) {
2575 br_dev = netdev_master_upper_dev_get(dev);
2576 cops = br_dev->netdev_ops;
2581 if (dev != br_dev &&
2582 !(dev->priv_flags & IFF_BRIDGE_PORT))
2585 if (br_dev != netdev_master_upper_dev_get(dev) &&
2586 !(dev->priv_flags & IFF_EBRIDGE))
2593 if (dev->priv_flags & IFF_BRIDGE_PORT) {
2594 if (cops && cops->ndo_fdb_dump)
2595 idx = cops->ndo_fdb_dump(skb, cb, br_dev, dev,
2599 idx = ndo_dflt_fdb_dump(skb, cb, dev, NULL, idx);
2600 if (dev->netdev_ops->ndo_fdb_dump)
2601 idx = dev->netdev_ops->ndo_fdb_dump(skb, cb, bdev, dev,
2611 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
2612 struct net_device *dev, u16 mode)
2614 struct nlmsghdr *nlh;
2615 struct ifinfomsg *ifm;
2616 struct nlattr *br_afspec;
2617 u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
2618 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2620 nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), NLM_F_MULTI);
2624 ifm = nlmsg_data(nlh);
2625 ifm->ifi_family = AF_BRIDGE;
2627 ifm->ifi_type = dev->type;
2628 ifm->ifi_index = dev->ifindex;
2629 ifm->ifi_flags = dev_get_flags(dev);
2630 ifm->ifi_change = 0;
2633 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
2634 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
2635 nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
2637 nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
2639 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
2640 (dev->ifindex != dev->iflink &&
2641 nla_put_u32(skb, IFLA_LINK, dev->iflink)))
2642 goto nla_put_failure;
2644 br_afspec = nla_nest_start(skb, IFLA_AF_SPEC);
2646 goto nla_put_failure;
2648 if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF) ||
2649 nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
2650 nla_nest_cancel(skb, br_afspec);
2651 goto nla_put_failure;
2653 nla_nest_end(skb, br_afspec);
2655 return nlmsg_end(skb, nlh);
2657 nlmsg_cancel(skb, nlh);
2660 EXPORT_SYMBOL(ndo_dflt_bridge_getlink);
2662 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
2664 struct net *net = sock_net(skb->sk);
2665 struct net_device *dev;
2667 u32 portid = NETLINK_CB(cb->skb).portid;
2668 u32 seq = cb->nlh->nlmsg_seq;
2669 struct nlattr *extfilt;
2670 u32 filter_mask = 0;
2672 extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct ifinfomsg),
2675 filter_mask = nla_get_u32(extfilt);
2678 for_each_netdev_rcu(net, dev) {
2679 const struct net_device_ops *ops = dev->netdev_ops;
2680 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2682 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
2683 if (idx >= cb->args[0] &&
2684 br_dev->netdev_ops->ndo_bridge_getlink(
2685 skb, portid, seq, dev, filter_mask) < 0)
2690 if (ops->ndo_bridge_getlink) {
2691 if (idx >= cb->args[0] &&
2692 ops->ndo_bridge_getlink(skb, portid, seq, dev,
2704 static inline size_t bridge_nlmsg_size(void)
2706 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
2707 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
2708 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
2709 + nla_total_size(sizeof(u32)) /* IFLA_MASTER */
2710 + nla_total_size(sizeof(u32)) /* IFLA_MTU */
2711 + nla_total_size(sizeof(u32)) /* IFLA_LINK */
2712 + nla_total_size(sizeof(u32)) /* IFLA_OPERSTATE */
2713 + nla_total_size(sizeof(u8)) /* IFLA_PROTINFO */
2714 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
2715 + nla_total_size(sizeof(u16)) /* IFLA_BRIDGE_FLAGS */
2716 + nla_total_size(sizeof(u16)); /* IFLA_BRIDGE_MODE */
2719 static int rtnl_bridge_notify(struct net_device *dev, u16 flags)
2721 struct net *net = dev_net(dev);
2722 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2723 struct sk_buff *skb;
2724 int err = -EOPNOTSUPP;
2726 skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
2732 if ((!flags || (flags & BRIDGE_FLAGS_MASTER)) &&
2733 br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
2734 err = br_dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0);
2739 if ((flags & BRIDGE_FLAGS_SELF) &&
2740 dev->netdev_ops->ndo_bridge_getlink) {
2741 err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0);
2746 rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
2749 WARN_ON(err == -EMSGSIZE);
2751 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2755 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
2757 struct net *net = sock_net(skb->sk);
2758 struct ifinfomsg *ifm;
2759 struct net_device *dev;
2760 struct nlattr *br_spec, *attr = NULL;
2761 int rem, err = -EOPNOTSUPP;
2762 u16 oflags, flags = 0;
2763 bool have_flags = false;
2765 if (nlmsg_len(nlh) < sizeof(*ifm))
2768 ifm = nlmsg_data(nlh);
2769 if (ifm->ifi_family != AF_BRIDGE)
2770 return -EPFNOSUPPORT;
2772 dev = __dev_get_by_index(net, ifm->ifi_index);
2774 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
2778 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
2780 nla_for_each_nested(attr, br_spec, rem) {
2781 if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
2783 flags = nla_get_u16(attr);
2791 if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
2792 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2794 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
2799 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh);
2803 flags &= ~BRIDGE_FLAGS_MASTER;
2806 if ((flags & BRIDGE_FLAGS_SELF)) {
2807 if (!dev->netdev_ops->ndo_bridge_setlink)
2810 err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh);
2813 flags &= ~BRIDGE_FLAGS_SELF;
2817 memcpy(nla_data(attr), &flags, sizeof(flags));
2818 /* Generate event to notify upper layer of bridge change */
2820 err = rtnl_bridge_notify(dev, oflags);
2825 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
2827 struct net *net = sock_net(skb->sk);
2828 struct ifinfomsg *ifm;
2829 struct net_device *dev;
2830 struct nlattr *br_spec, *attr = NULL;
2831 int rem, err = -EOPNOTSUPP;
2832 u16 oflags, flags = 0;
2833 bool have_flags = false;
2835 if (nlmsg_len(nlh) < sizeof(*ifm))
2838 ifm = nlmsg_data(nlh);
2839 if (ifm->ifi_family != AF_BRIDGE)
2840 return -EPFNOSUPPORT;
2842 dev = __dev_get_by_index(net, ifm->ifi_index);
2844 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
2848 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
2850 nla_for_each_nested(attr, br_spec, rem) {
2851 if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
2853 flags = nla_get_u16(attr);
2861 if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
2862 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2864 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
2869 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh);
2873 flags &= ~BRIDGE_FLAGS_MASTER;
2876 if ((flags & BRIDGE_FLAGS_SELF)) {
2877 if (!dev->netdev_ops->ndo_bridge_dellink)
2880 err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh);
2883 flags &= ~BRIDGE_FLAGS_SELF;
2887 memcpy(nla_data(attr), &flags, sizeof(flags));
2888 /* Generate event to notify upper layer of bridge change */
2890 err = rtnl_bridge_notify(dev, oflags);
2895 /* Process one rtnetlink message. */
2897 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
2899 struct net *net = sock_net(skb->sk);
2900 rtnl_doit_func doit;
2906 type = nlh->nlmsg_type;
2912 /* All the messages must have at least 1 byte length */
2913 if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
2916 family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
2920 if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
2923 if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
2925 rtnl_dumpit_func dumpit;
2926 rtnl_calcit_func calcit;
2927 u16 min_dump_alloc = 0;
2929 dumpit = rtnl_get_dumpit(family, type);
2932 calcit = rtnl_get_calcit(family, type);
2934 min_dump_alloc = calcit(skb, nlh);
2939 struct netlink_dump_control c = {
2941 .min_dump_alloc = min_dump_alloc,
2943 err = netlink_dump_start(rtnl, skb, nlh, &c);
2949 doit = rtnl_get_doit(family, type);
2953 return doit(skb, nlh);
2956 static void rtnetlink_rcv(struct sk_buff *skb)
2959 netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
2963 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
2965 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2971 case NETDEV_POST_INIT:
2972 case NETDEV_REGISTER:
2974 case NETDEV_PRE_TYPE_CHANGE:
2975 case NETDEV_GOING_DOWN:
2976 case NETDEV_UNREGISTER:
2977 case NETDEV_UNREGISTER_FINAL:
2978 case NETDEV_RELEASE:
2982 rtmsg_ifinfo(RTM_NEWLINK, dev, 0, GFP_KERNEL);
2988 static struct notifier_block rtnetlink_dev_notifier = {
2989 .notifier_call = rtnetlink_event,
2993 static int __net_init rtnetlink_net_init(struct net *net)
2996 struct netlink_kernel_cfg cfg = {
2997 .groups = RTNLGRP_MAX,
2998 .input = rtnetlink_rcv,
2999 .cb_mutex = &rtnl_mutex,
3000 .flags = NL_CFG_F_NONROOT_RECV,
3003 sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
3010 static void __net_exit rtnetlink_net_exit(struct net *net)
3012 netlink_kernel_release(net->rtnl);
3016 static struct pernet_operations rtnetlink_net_ops = {
3017 .init = rtnetlink_net_init,
3018 .exit = rtnetlink_net_exit,
3021 void __init rtnetlink_init(void)
3023 if (register_pernet_subsys(&rtnetlink_net_ops))
3024 panic("rtnetlink_init: cannot initialize rtnetlink\n");
3026 register_netdevice_notifier(&rtnetlink_dev_notifier);
3028 rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
3029 rtnl_dump_ifinfo, rtnl_calcit);
3030 rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
3031 rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
3032 rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
3034 rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
3035 rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
3037 rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, NULL);
3038 rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, NULL);
3039 rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, NULL);
3041 rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, NULL);
3042 rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, NULL);
3043 rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, NULL);