2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
6 * Routing netlink socket interface: protocol independent part.
8 * Authors: Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
10 * This program is free software; you can redistribute it and/or
11 * modify it under the terms of the GNU General Public License
12 * as published by the Free Software Foundation; either version
13 * 2 of the License, or (at your option) any later version.
16 * Vitaly E. Lavrov RTA_OK arithmetics was wrong.
19 #include <linux/errno.h>
20 #include <linux/module.h>
21 #include <linux/types.h>
22 #include <linux/socket.h>
23 #include <linux/kernel.h>
24 #include <linux/timer.h>
25 #include <linux/string.h>
26 #include <linux/sockios.h>
27 #include <linux/net.h>
28 #include <linux/fcntl.h>
30 #include <linux/slab.h>
31 #include <linux/interrupt.h>
32 #include <linux/capability.h>
33 #include <linux/skbuff.h>
34 #include <linux/init.h>
35 #include <linux/security.h>
36 #include <linux/mutex.h>
37 #include <linux/if_addr.h>
38 #include <linux/if_bridge.h>
39 #include <linux/if_vlan.h>
40 #include <linux/pci.h>
41 #include <linux/etherdevice.h>
43 #include <asm/uaccess.h>
45 #include <linux/inet.h>
46 #include <linux/netdevice.h>
47 #include <net/switchdev.h>
49 #include <net/protocol.h>
51 #include <net/route.h>
55 #include <net/pkt_sched.h>
56 #include <net/fib_rules.h>
57 #include <net/rtnetlink.h>
58 #include <net/net_namespace.h>
62 rtnl_dumpit_func dumpit;
63 rtnl_calcit_func calcit;
66 static DEFINE_MUTEX(rtnl_mutex);
70 mutex_lock(&rtnl_mutex);
72 EXPORT_SYMBOL(rtnl_lock);
74 void __rtnl_unlock(void)
76 mutex_unlock(&rtnl_mutex);
79 void rtnl_unlock(void)
81 /* This fellow will unlock it for us. */
84 EXPORT_SYMBOL(rtnl_unlock);
86 int rtnl_trylock(void)
88 return mutex_trylock(&rtnl_mutex);
90 EXPORT_SYMBOL(rtnl_trylock);
92 int rtnl_is_locked(void)
94 return mutex_is_locked(&rtnl_mutex);
96 EXPORT_SYMBOL(rtnl_is_locked);
98 #ifdef CONFIG_PROVE_LOCKING
99 bool lockdep_rtnl_is_held(void)
101 return lockdep_is_held(&rtnl_mutex);
103 EXPORT_SYMBOL(lockdep_rtnl_is_held);
104 #endif /* #ifdef CONFIG_PROVE_LOCKING */
106 static struct rtnl_link *rtnl_msg_handlers[RTNL_FAMILY_MAX + 1];
108 static inline int rtm_msgindex(int msgtype)
110 int msgindex = msgtype - RTM_BASE;
113 * msgindex < 0 implies someone tried to register a netlink
114 * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
115 * the message type has not been added to linux/rtnetlink.h
117 BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
122 static rtnl_doit_func rtnl_get_doit(int protocol, int msgindex)
124 struct rtnl_link *tab;
126 if (protocol <= RTNL_FAMILY_MAX)
127 tab = rtnl_msg_handlers[protocol];
131 if (tab == NULL || tab[msgindex].doit == NULL)
132 tab = rtnl_msg_handlers[PF_UNSPEC];
134 return tab[msgindex].doit;
137 static rtnl_dumpit_func rtnl_get_dumpit(int protocol, int msgindex)
139 struct rtnl_link *tab;
141 if (protocol <= RTNL_FAMILY_MAX)
142 tab = rtnl_msg_handlers[protocol];
146 if (tab == NULL || tab[msgindex].dumpit == NULL)
147 tab = rtnl_msg_handlers[PF_UNSPEC];
149 return tab[msgindex].dumpit;
152 static rtnl_calcit_func rtnl_get_calcit(int protocol, int msgindex)
154 struct rtnl_link *tab;
156 if (protocol <= RTNL_FAMILY_MAX)
157 tab = rtnl_msg_handlers[protocol];
161 if (tab == NULL || tab[msgindex].calcit == NULL)
162 tab = rtnl_msg_handlers[PF_UNSPEC];
164 return tab[msgindex].calcit;
168 * __rtnl_register - Register a rtnetlink message type
169 * @protocol: Protocol family or PF_UNSPEC
170 * @msgtype: rtnetlink message type
171 * @doit: Function pointer called for each request message
172 * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
173 * @calcit: Function pointer to calc size of dump message
175 * Registers the specified function pointers (at least one of them has
176 * to be non-NULL) to be called whenever a request message for the
177 * specified protocol family and message type is received.
179 * The special protocol family PF_UNSPEC may be used to define fallback
180 * function pointers for the case when no entry for the specific protocol
183 * Returns 0 on success or a negative error code.
185 int __rtnl_register(int protocol, int msgtype,
186 rtnl_doit_func doit, rtnl_dumpit_func dumpit,
187 rtnl_calcit_func calcit)
189 struct rtnl_link *tab;
192 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
193 msgindex = rtm_msgindex(msgtype);
195 tab = rtnl_msg_handlers[protocol];
197 tab = kcalloc(RTM_NR_MSGTYPES, sizeof(*tab), GFP_KERNEL);
201 rtnl_msg_handlers[protocol] = tab;
205 tab[msgindex].doit = doit;
208 tab[msgindex].dumpit = dumpit;
211 tab[msgindex].calcit = calcit;
215 EXPORT_SYMBOL_GPL(__rtnl_register);
218 * rtnl_register - Register a rtnetlink message type
220 * Identical to __rtnl_register() but panics on failure. This is useful
221 * as failure of this function is very unlikely, it can only happen due
222 * to lack of memory when allocating the chain to store all message
223 * handlers for a protocol. Meant for use in init functions where lack
224 * of memory implies no sense in continuing.
226 void rtnl_register(int protocol, int msgtype,
227 rtnl_doit_func doit, rtnl_dumpit_func dumpit,
228 rtnl_calcit_func calcit)
230 if (__rtnl_register(protocol, msgtype, doit, dumpit, calcit) < 0)
231 panic("Unable to register rtnetlink message handler, "
232 "protocol = %d, message type = %d\n",
235 EXPORT_SYMBOL_GPL(rtnl_register);
238 * rtnl_unregister - Unregister a rtnetlink message type
239 * @protocol: Protocol family or PF_UNSPEC
240 * @msgtype: rtnetlink message type
242 * Returns 0 on success or a negative error code.
244 int rtnl_unregister(int protocol, int msgtype)
248 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
249 msgindex = rtm_msgindex(msgtype);
251 if (rtnl_msg_handlers[protocol] == NULL)
254 rtnl_msg_handlers[protocol][msgindex].doit = NULL;
255 rtnl_msg_handlers[protocol][msgindex].dumpit = NULL;
256 rtnl_msg_handlers[protocol][msgindex].calcit = NULL;
260 EXPORT_SYMBOL_GPL(rtnl_unregister);
263 * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
264 * @protocol : Protocol family or PF_UNSPEC
266 * Identical to calling rtnl_unregster() for all registered message types
267 * of a certain protocol family.
269 void rtnl_unregister_all(int protocol)
271 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
273 kfree(rtnl_msg_handlers[protocol]);
274 rtnl_msg_handlers[protocol] = NULL;
276 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
278 static LIST_HEAD(link_ops);
280 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
282 const struct rtnl_link_ops *ops;
284 list_for_each_entry(ops, &link_ops, list) {
285 if (!strcmp(ops->kind, kind))
292 * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
293 * @ops: struct rtnl_link_ops * to register
295 * The caller must hold the rtnl_mutex. This function should be used
296 * by drivers that create devices during module initialization. It
297 * must be called before registering the devices.
299 * Returns 0 on success or a negative error code.
301 int __rtnl_link_register(struct rtnl_link_ops *ops)
303 if (rtnl_link_ops_get(ops->kind))
306 /* The check for setup is here because if ops
307 * does not have that filled up, it is not possible
308 * to use the ops for creating device. So do not
309 * fill up dellink as well. That disables rtnl_dellink.
311 if (ops->setup && !ops->dellink)
312 ops->dellink = unregister_netdevice_queue;
314 list_add_tail(&ops->list, &link_ops);
317 EXPORT_SYMBOL_GPL(__rtnl_link_register);
320 * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
321 * @ops: struct rtnl_link_ops * to register
323 * Returns 0 on success or a negative error code.
325 int rtnl_link_register(struct rtnl_link_ops *ops)
330 err = __rtnl_link_register(ops);
334 EXPORT_SYMBOL_GPL(rtnl_link_register);
336 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
338 struct net_device *dev;
339 LIST_HEAD(list_kill);
341 for_each_netdev(net, dev) {
342 if (dev->rtnl_link_ops == ops)
343 ops->dellink(dev, &list_kill);
345 unregister_netdevice_many(&list_kill);
349 * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
350 * @ops: struct rtnl_link_ops * to unregister
352 * The caller must hold the rtnl_mutex.
354 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
359 __rtnl_kill_links(net, ops);
361 list_del(&ops->list);
363 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
365 /* Return with the rtnl_lock held when there are no network
366 * devices unregistering in any network namespace.
368 static void rtnl_lock_unregistering_all(void)
372 DEFINE_WAIT_FUNC(wait, woken_wake_function);
374 add_wait_queue(&netdev_unregistering_wq, &wait);
376 unregistering = false;
379 if (net->dev_unreg_count > 0) {
380 unregistering = true;
388 wait_woken(&wait, TASK_UNINTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
390 remove_wait_queue(&netdev_unregistering_wq, &wait);
394 * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
395 * @ops: struct rtnl_link_ops * to unregister
397 void rtnl_link_unregister(struct rtnl_link_ops *ops)
399 /* Close the race with cleanup_net() */
400 mutex_lock(&net_mutex);
401 rtnl_lock_unregistering_all();
402 __rtnl_link_unregister(ops);
404 mutex_unlock(&net_mutex);
406 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
408 static size_t rtnl_link_get_slave_info_data_size(const struct net_device *dev)
410 struct net_device *master_dev;
411 const struct rtnl_link_ops *ops;
413 master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
416 ops = master_dev->rtnl_link_ops;
417 if (!ops || !ops->get_slave_size)
419 /* IFLA_INFO_SLAVE_DATA + nested data */
420 return nla_total_size(sizeof(struct nlattr)) +
421 ops->get_slave_size(master_dev, dev);
424 static size_t rtnl_link_get_size(const struct net_device *dev)
426 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
432 size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
433 nla_total_size(strlen(ops->kind) + 1); /* IFLA_INFO_KIND */
436 /* IFLA_INFO_DATA + nested data */
437 size += nla_total_size(sizeof(struct nlattr)) +
440 if (ops->get_xstats_size)
441 /* IFLA_INFO_XSTATS */
442 size += nla_total_size(ops->get_xstats_size(dev));
444 size += rtnl_link_get_slave_info_data_size(dev);
449 static LIST_HEAD(rtnl_af_ops);
451 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
453 const struct rtnl_af_ops *ops;
455 list_for_each_entry(ops, &rtnl_af_ops, list) {
456 if (ops->family == family)
464 * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
465 * @ops: struct rtnl_af_ops * to register
467 * Returns 0 on success or a negative error code.
469 void rtnl_af_register(struct rtnl_af_ops *ops)
472 list_add_tail(&ops->list, &rtnl_af_ops);
475 EXPORT_SYMBOL_GPL(rtnl_af_register);
478 * __rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
479 * @ops: struct rtnl_af_ops * to unregister
481 * The caller must hold the rtnl_mutex.
483 void __rtnl_af_unregister(struct rtnl_af_ops *ops)
485 list_del(&ops->list);
487 EXPORT_SYMBOL_GPL(__rtnl_af_unregister);
490 * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
491 * @ops: struct rtnl_af_ops * to unregister
493 void rtnl_af_unregister(struct rtnl_af_ops *ops)
496 __rtnl_af_unregister(ops);
499 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
501 static size_t rtnl_link_get_af_size(const struct net_device *dev,
504 struct rtnl_af_ops *af_ops;
508 size = nla_total_size(sizeof(struct nlattr));
510 list_for_each_entry(af_ops, &rtnl_af_ops, list) {
511 if (af_ops->get_link_af_size) {
512 /* AF_* + nested data */
513 size += nla_total_size(sizeof(struct nlattr)) +
514 af_ops->get_link_af_size(dev, ext_filter_mask);
521 static bool rtnl_have_link_slave_info(const struct net_device *dev)
523 struct net_device *master_dev;
525 master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
526 if (master_dev && master_dev->rtnl_link_ops)
531 static int rtnl_link_slave_info_fill(struct sk_buff *skb,
532 const struct net_device *dev)
534 struct net_device *master_dev;
535 const struct rtnl_link_ops *ops;
536 struct nlattr *slave_data;
539 master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
542 ops = master_dev->rtnl_link_ops;
545 if (nla_put_string(skb, IFLA_INFO_SLAVE_KIND, ops->kind) < 0)
547 if (ops->fill_slave_info) {
548 slave_data = nla_nest_start(skb, IFLA_INFO_SLAVE_DATA);
551 err = ops->fill_slave_info(skb, master_dev, dev);
553 goto err_cancel_slave_data;
554 nla_nest_end(skb, slave_data);
558 err_cancel_slave_data:
559 nla_nest_cancel(skb, slave_data);
563 static int rtnl_link_info_fill(struct sk_buff *skb,
564 const struct net_device *dev)
566 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
572 if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
574 if (ops->fill_xstats) {
575 err = ops->fill_xstats(skb, dev);
579 if (ops->fill_info) {
580 data = nla_nest_start(skb, IFLA_INFO_DATA);
583 err = ops->fill_info(skb, dev);
585 goto err_cancel_data;
586 nla_nest_end(skb, data);
591 nla_nest_cancel(skb, data);
595 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
597 struct nlattr *linkinfo;
600 linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
601 if (linkinfo == NULL)
604 err = rtnl_link_info_fill(skb, dev);
606 goto err_cancel_link;
608 err = rtnl_link_slave_info_fill(skb, dev);
610 goto err_cancel_link;
612 nla_nest_end(skb, linkinfo);
616 nla_nest_cancel(skb, linkinfo);
621 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
623 struct sock *rtnl = net->rtnl;
626 NETLINK_CB(skb).dst_group = group;
628 atomic_inc(&skb->users);
629 netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
631 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
635 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
637 struct sock *rtnl = net->rtnl;
639 return nlmsg_unicast(rtnl, skb, pid);
641 EXPORT_SYMBOL(rtnl_unicast);
643 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
644 struct nlmsghdr *nlh, gfp_t flags)
646 struct sock *rtnl = net->rtnl;
650 report = nlmsg_report(nlh);
652 nlmsg_notify(rtnl, skb, pid, group, report, flags);
654 EXPORT_SYMBOL(rtnl_notify);
656 void rtnl_set_sk_err(struct net *net, u32 group, int error)
658 struct sock *rtnl = net->rtnl;
660 netlink_set_err(rtnl, 0, group, error);
662 EXPORT_SYMBOL(rtnl_set_sk_err);
664 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
669 mx = nla_nest_start(skb, RTA_METRICS);
673 for (i = 0; i < RTAX_MAX; i++) {
675 if (i == RTAX_CC_ALGO - 1) {
676 char tmp[TCP_CA_NAME_MAX], *name;
678 name = tcp_ca_get_name_by_key(metrics[i], tmp);
681 if (nla_put_string(skb, i + 1, name))
682 goto nla_put_failure;
683 } else if (i == RTAX_FEATURES - 1) {
684 u32 user_features = metrics[i] & RTAX_FEATURE_MASK;
686 BUILD_BUG_ON(RTAX_FEATURE_MASK & DST_FEATURE_MASK);
687 if (nla_put_u32(skb, i + 1, user_features))
688 goto nla_put_failure;
690 if (nla_put_u32(skb, i + 1, metrics[i]))
691 goto nla_put_failure;
698 nla_nest_cancel(skb, mx);
702 return nla_nest_end(skb, mx);
705 nla_nest_cancel(skb, mx);
708 EXPORT_SYMBOL(rtnetlink_put_metrics);
710 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
711 long expires, u32 error)
713 struct rta_cacheinfo ci = {
714 .rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse),
715 .rta_used = dst->__use,
716 .rta_clntref = atomic_read(&(dst->__refcnt)),
724 clock = jiffies_to_clock_t(abs(expires));
725 clock = min_t(unsigned long, clock, INT_MAX);
726 ci.rta_expires = (expires > 0) ? clock : -clock;
728 return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
730 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
732 static void set_operstate(struct net_device *dev, unsigned char transition)
734 unsigned char operstate = dev->operstate;
736 switch (transition) {
738 if ((operstate == IF_OPER_DORMANT ||
739 operstate == IF_OPER_UNKNOWN) &&
741 operstate = IF_OPER_UP;
744 case IF_OPER_DORMANT:
745 if (operstate == IF_OPER_UP ||
746 operstate == IF_OPER_UNKNOWN)
747 operstate = IF_OPER_DORMANT;
751 if (dev->operstate != operstate) {
752 write_lock_bh(&dev_base_lock);
753 dev->operstate = operstate;
754 write_unlock_bh(&dev_base_lock);
755 netdev_state_change(dev);
759 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
761 return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
762 (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
765 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
766 const struct ifinfomsg *ifm)
768 unsigned int flags = ifm->ifi_flags;
770 /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
772 flags = (flags & ifm->ifi_change) |
773 (rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
778 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
779 const struct rtnl_link_stats64 *b)
781 a->rx_packets = b->rx_packets;
782 a->tx_packets = b->tx_packets;
783 a->rx_bytes = b->rx_bytes;
784 a->tx_bytes = b->tx_bytes;
785 a->rx_errors = b->rx_errors;
786 a->tx_errors = b->tx_errors;
787 a->rx_dropped = b->rx_dropped;
788 a->tx_dropped = b->tx_dropped;
790 a->multicast = b->multicast;
791 a->collisions = b->collisions;
793 a->rx_length_errors = b->rx_length_errors;
794 a->rx_over_errors = b->rx_over_errors;
795 a->rx_crc_errors = b->rx_crc_errors;
796 a->rx_frame_errors = b->rx_frame_errors;
797 a->rx_fifo_errors = b->rx_fifo_errors;
798 a->rx_missed_errors = b->rx_missed_errors;
800 a->tx_aborted_errors = b->tx_aborted_errors;
801 a->tx_carrier_errors = b->tx_carrier_errors;
802 a->tx_fifo_errors = b->tx_fifo_errors;
803 a->tx_heartbeat_errors = b->tx_heartbeat_errors;
804 a->tx_window_errors = b->tx_window_errors;
806 a->rx_compressed = b->rx_compressed;
807 a->tx_compressed = b->tx_compressed;
810 static void copy_rtnl_link_stats64(void *v, const struct rtnl_link_stats64 *b)
812 memcpy(v, b, sizeof(*b));
816 static inline int rtnl_vfinfo_size(const struct net_device *dev,
819 if (dev->dev.parent && dev_is_pci(dev->dev.parent) &&
820 (ext_filter_mask & RTEXT_FILTER_VF)) {
821 int num_vfs = dev_num_vf(dev->dev.parent);
822 size_t size = nla_total_size(sizeof(struct nlattr));
823 size += nla_total_size(num_vfs * sizeof(struct nlattr));
825 (nla_total_size(sizeof(struct ifla_vf_mac)) +
826 nla_total_size(sizeof(struct ifla_vf_vlan)) +
827 nla_total_size(sizeof(struct ifla_vf_spoofchk)) +
828 nla_total_size(sizeof(struct ifla_vf_rate)) +
829 nla_total_size(sizeof(struct ifla_vf_link_state)) +
830 nla_total_size(sizeof(struct ifla_vf_rss_query_en)) +
831 /* IFLA_VF_STATS_RX_PACKETS */
832 nla_total_size(sizeof(__u64)) +
833 /* IFLA_VF_STATS_TX_PACKETS */
834 nla_total_size(sizeof(__u64)) +
835 /* IFLA_VF_STATS_RX_BYTES */
836 nla_total_size(sizeof(__u64)) +
837 /* IFLA_VF_STATS_TX_BYTES */
838 nla_total_size(sizeof(__u64)) +
839 /* IFLA_VF_STATS_BROADCAST */
840 nla_total_size(sizeof(__u64)) +
841 /* IFLA_VF_STATS_MULTICAST */
842 nla_total_size(sizeof(__u64)) +
843 nla_total_size(sizeof(struct ifla_vf_trust)));
849 static size_t rtnl_port_size(const struct net_device *dev,
852 size_t port_size = nla_total_size(4) /* PORT_VF */
853 + nla_total_size(PORT_PROFILE_MAX) /* PORT_PROFILE */
854 + nla_total_size(sizeof(struct ifla_port_vsi))
856 + nla_total_size(PORT_UUID_MAX) /* PORT_INSTANCE_UUID */
857 + nla_total_size(PORT_UUID_MAX) /* PORT_HOST_UUID */
858 + nla_total_size(1) /* PROT_VDP_REQUEST */
859 + nla_total_size(2); /* PORT_VDP_RESPONSE */
860 size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
861 size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
863 size_t port_self_size = nla_total_size(sizeof(struct nlattr))
866 if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
867 !(ext_filter_mask & RTEXT_FILTER_VF))
869 if (dev_num_vf(dev->dev.parent))
870 return port_self_size + vf_ports_size +
871 vf_port_size * dev_num_vf(dev->dev.parent);
873 return port_self_size;
876 static noinline size_t if_nlmsg_size(const struct net_device *dev,
879 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
880 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
881 + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
882 + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
883 + nla_total_size(sizeof(struct rtnl_link_ifmap))
884 + nla_total_size(sizeof(struct rtnl_link_stats))
885 + nla_total_size(sizeof(struct rtnl_link_stats64))
886 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
887 + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
888 + nla_total_size(4) /* IFLA_TXQLEN */
889 + nla_total_size(4) /* IFLA_WEIGHT */
890 + nla_total_size(4) /* IFLA_MTU */
891 + nla_total_size(4) /* IFLA_LINK */
892 + nla_total_size(4) /* IFLA_MASTER */
893 + nla_total_size(1) /* IFLA_CARRIER */
894 + nla_total_size(4) /* IFLA_PROMISCUITY */
895 + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
896 + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
897 + nla_total_size(1) /* IFLA_OPERSTATE */
898 + nla_total_size(1) /* IFLA_LINKMODE */
899 + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
900 + nla_total_size(4) /* IFLA_LINK_NETNSID */
901 + nla_total_size(4) /* IFLA_GROUP */
902 + nla_total_size(ext_filter_mask
903 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
904 + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
905 + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
906 + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
907 + rtnl_link_get_af_size(dev, ext_filter_mask) /* IFLA_AF_SPEC */
908 + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */
909 + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_SWITCH_ID */
910 + nla_total_size(IFNAMSIZ) /* IFLA_PHYS_PORT_NAME */
911 + nla_total_size(1); /* IFLA_PROTO_DOWN */
915 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
917 struct nlattr *vf_ports;
918 struct nlattr *vf_port;
922 vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
926 for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
927 vf_port = nla_nest_start(skb, IFLA_VF_PORT);
929 goto nla_put_failure;
930 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
931 goto nla_put_failure;
932 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
933 if (err == -EMSGSIZE)
934 goto nla_put_failure;
936 nla_nest_cancel(skb, vf_port);
939 nla_nest_end(skb, vf_port);
942 nla_nest_end(skb, vf_ports);
947 nla_nest_cancel(skb, vf_ports);
951 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
953 struct nlattr *port_self;
956 port_self = nla_nest_start(skb, IFLA_PORT_SELF);
960 err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
962 nla_nest_cancel(skb, port_self);
963 return (err == -EMSGSIZE) ? err : 0;
966 nla_nest_end(skb, port_self);
971 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
976 if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
977 !(ext_filter_mask & RTEXT_FILTER_VF))
980 err = rtnl_port_self_fill(skb, dev);
984 if (dev_num_vf(dev->dev.parent)) {
985 err = rtnl_vf_ports_fill(skb, dev);
993 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
996 struct netdev_phys_item_id ppid;
998 err = dev_get_phys_port_id(dev, &ppid);
1000 if (err == -EOPNOTSUPP)
1005 if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
1011 static int rtnl_phys_port_name_fill(struct sk_buff *skb, struct net_device *dev)
1013 char name[IFNAMSIZ];
1016 err = dev_get_phys_port_name(dev, name, sizeof(name));
1018 if (err == -EOPNOTSUPP)
1023 if (nla_put_string(skb, IFLA_PHYS_PORT_NAME, name))
1029 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev)
1032 struct switchdev_attr attr = {
1033 .id = SWITCHDEV_ATTR_ID_PORT_PARENT_ID,
1034 .flags = SWITCHDEV_F_NO_RECURSE,
1037 err = switchdev_port_attr_get(dev, &attr);
1039 if (err == -EOPNOTSUPP)
1044 if (nla_put(skb, IFLA_PHYS_SWITCH_ID, attr.u.ppid.id_len,
1051 static noinline_for_stack int rtnl_fill_stats(struct sk_buff *skb,
1052 struct net_device *dev)
1054 const struct rtnl_link_stats64 *stats;
1055 struct rtnl_link_stats64 temp;
1056 struct nlattr *attr;
1058 stats = dev_get_stats(dev, &temp);
1060 attr = nla_reserve(skb, IFLA_STATS,
1061 sizeof(struct rtnl_link_stats));
1065 copy_rtnl_link_stats(nla_data(attr), stats);
1067 attr = nla_reserve(skb, IFLA_STATS64,
1068 sizeof(struct rtnl_link_stats64));
1072 copy_rtnl_link_stats64(nla_data(attr), stats);
1077 static noinline_for_stack int rtnl_fill_vfinfo(struct sk_buff *skb,
1078 struct net_device *dev,
1080 struct nlattr *vfinfo)
1082 struct ifla_vf_rss_query_en vf_rss_query_en;
1083 struct ifla_vf_link_state vf_linkstate;
1084 struct ifla_vf_spoofchk vf_spoofchk;
1085 struct ifla_vf_tx_rate vf_tx_rate;
1086 struct ifla_vf_stats vf_stats;
1087 struct ifla_vf_trust vf_trust;
1088 struct ifla_vf_vlan vf_vlan;
1089 struct ifla_vf_rate vf_rate;
1090 struct nlattr *vf, *vfstats;
1091 struct ifla_vf_mac vf_mac;
1092 struct ifla_vf_info ivi;
1094 memset(&ivi, 0, sizeof(ivi));
1096 /* Not all SR-IOV capable drivers support the
1097 * spoofcheck and "RSS query enable" query. Preset to
1098 * -1 so the user space tool can detect that the driver
1099 * didn't report anything.
1102 ivi.rss_query_en = -1;
1104 /* The default value for VF link state is "auto"
1105 * IFLA_VF_LINK_STATE_AUTO which equals zero
1108 if (dev->netdev_ops->ndo_get_vf_config(dev, vfs_num, &ivi))
1117 vf_rss_query_en.vf =
1118 vf_trust.vf = ivi.vf;
1120 memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1121 vf_vlan.vlan = ivi.vlan;
1122 vf_vlan.qos = ivi.qos;
1123 vf_tx_rate.rate = ivi.max_tx_rate;
1124 vf_rate.min_tx_rate = ivi.min_tx_rate;
1125 vf_rate.max_tx_rate = ivi.max_tx_rate;
1126 vf_spoofchk.setting = ivi.spoofchk;
1127 vf_linkstate.link_state = ivi.linkstate;
1128 vf_rss_query_en.setting = ivi.rss_query_en;
1129 vf_trust.setting = ivi.trusted;
1130 vf = nla_nest_start(skb, IFLA_VF_INFO);
1132 nla_nest_cancel(skb, vfinfo);
1135 if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1136 nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1137 nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1139 nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1141 nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1143 nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1145 nla_put(skb, IFLA_VF_RSS_QUERY_EN,
1146 sizeof(vf_rss_query_en),
1147 &vf_rss_query_en) ||
1148 nla_put(skb, IFLA_VF_TRUST,
1149 sizeof(vf_trust), &vf_trust))
1151 memset(&vf_stats, 0, sizeof(vf_stats));
1152 if (dev->netdev_ops->ndo_get_vf_stats)
1153 dev->netdev_ops->ndo_get_vf_stats(dev, vfs_num,
1155 vfstats = nla_nest_start(skb, IFLA_VF_STATS);
1157 nla_nest_cancel(skb, vf);
1158 nla_nest_cancel(skb, vfinfo);
1161 if (nla_put_u64(skb, IFLA_VF_STATS_RX_PACKETS,
1162 vf_stats.rx_packets) ||
1163 nla_put_u64(skb, IFLA_VF_STATS_TX_PACKETS,
1164 vf_stats.tx_packets) ||
1165 nla_put_u64(skb, IFLA_VF_STATS_RX_BYTES,
1166 vf_stats.rx_bytes) ||
1167 nla_put_u64(skb, IFLA_VF_STATS_TX_BYTES,
1168 vf_stats.tx_bytes) ||
1169 nla_put_u64(skb, IFLA_VF_STATS_BROADCAST,
1170 vf_stats.broadcast) ||
1171 nla_put_u64(skb, IFLA_VF_STATS_MULTICAST,
1172 vf_stats.multicast))
1174 nla_nest_end(skb, vfstats);
1175 nla_nest_end(skb, vf);
1179 static int rtnl_fill_link_ifmap(struct sk_buff *skb, struct net_device *dev)
1181 struct rtnl_link_ifmap map;
1183 memset(&map, 0, sizeof(map));
1184 map.mem_start = dev->mem_start;
1185 map.mem_end = dev->mem_end;
1186 map.base_addr = dev->base_addr;
1189 map.port = dev->if_port;
1191 if (nla_put(skb, IFLA_MAP, sizeof(map), &map))
1197 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
1198 int type, u32 pid, u32 seq, u32 change,
1199 unsigned int flags, u32 ext_filter_mask)
1201 struct ifinfomsg *ifm;
1202 struct nlmsghdr *nlh;
1203 struct nlattr *af_spec;
1204 struct rtnl_af_ops *af_ops;
1205 struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1208 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
1212 ifm = nlmsg_data(nlh);
1213 ifm->ifi_family = AF_UNSPEC;
1215 ifm->ifi_type = dev->type;
1216 ifm->ifi_index = dev->ifindex;
1217 ifm->ifi_flags = dev_get_flags(dev);
1218 ifm->ifi_change = change;
1220 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
1221 nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
1222 nla_put_u8(skb, IFLA_OPERSTATE,
1223 netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
1224 nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
1225 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
1226 nla_put_u32(skb, IFLA_GROUP, dev->group) ||
1227 nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
1228 nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
1230 nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1232 (dev->ifindex != dev_get_iflink(dev) &&
1233 nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
1235 nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex)) ||
1236 nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1238 nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
1240 nla_put_string(skb, IFLA_IFALIAS, dev->ifalias)) ||
1241 nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1242 atomic_read(&dev->carrier_changes)) ||
1243 nla_put_u8(skb, IFLA_PROTO_DOWN, dev->proto_down))
1244 goto nla_put_failure;
1246 if (rtnl_fill_link_ifmap(skb, dev))
1247 goto nla_put_failure;
1249 if (dev->addr_len) {
1250 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1251 nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1252 goto nla_put_failure;
1255 if (rtnl_phys_port_id_fill(skb, dev))
1256 goto nla_put_failure;
1258 if (rtnl_phys_port_name_fill(skb, dev))
1259 goto nla_put_failure;
1261 if (rtnl_phys_switch_id_fill(skb, dev))
1262 goto nla_put_failure;
1264 if (rtnl_fill_stats(skb, dev))
1265 goto nla_put_failure;
1267 if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF) &&
1268 nla_put_u32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent)))
1269 goto nla_put_failure;
1271 if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent &&
1272 ext_filter_mask & RTEXT_FILTER_VF) {
1274 struct nlattr *vfinfo;
1275 int num_vfs = dev_num_vf(dev->dev.parent);
1277 vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
1279 goto nla_put_failure;
1280 for (i = 0; i < num_vfs; i++) {
1281 if (rtnl_fill_vfinfo(skb, dev, i, vfinfo))
1282 goto nla_put_failure;
1285 nla_nest_end(skb, vfinfo);
1288 if (rtnl_port_fill(skb, dev, ext_filter_mask))
1289 goto nla_put_failure;
1291 if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1292 if (rtnl_link_fill(skb, dev) < 0)
1293 goto nla_put_failure;
1296 if (dev->rtnl_link_ops &&
1297 dev->rtnl_link_ops->get_link_net) {
1298 struct net *link_net = dev->rtnl_link_ops->get_link_net(dev);
1300 if (!net_eq(dev_net(dev), link_net)) {
1301 int id = peernet2id_alloc(dev_net(dev), link_net);
1303 if (nla_put_s32(skb, IFLA_LINK_NETNSID, id))
1304 goto nla_put_failure;
1308 if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1309 goto nla_put_failure;
1311 list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1312 if (af_ops->fill_link_af) {
1316 if (!(af = nla_nest_start(skb, af_ops->family)))
1317 goto nla_put_failure;
1319 err = af_ops->fill_link_af(skb, dev, ext_filter_mask);
1322 * Caller may return ENODATA to indicate that there
1323 * was no data to be dumped. This is not an error, it
1324 * means we should trim the attribute header and
1327 if (err == -ENODATA)
1328 nla_nest_cancel(skb, af);
1330 goto nla_put_failure;
1332 nla_nest_end(skb, af);
1336 nla_nest_end(skb, af_spec);
1338 nlmsg_end(skb, nlh);
1342 nlmsg_cancel(skb, nlh);
1346 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1347 [IFLA_IFNAME] = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1348 [IFLA_ADDRESS] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1349 [IFLA_BROADCAST] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1350 [IFLA_MAP] = { .len = sizeof(struct rtnl_link_ifmap) },
1351 [IFLA_MTU] = { .type = NLA_U32 },
1352 [IFLA_LINK] = { .type = NLA_U32 },
1353 [IFLA_MASTER] = { .type = NLA_U32 },
1354 [IFLA_CARRIER] = { .type = NLA_U8 },
1355 [IFLA_TXQLEN] = { .type = NLA_U32 },
1356 [IFLA_WEIGHT] = { .type = NLA_U32 },
1357 [IFLA_OPERSTATE] = { .type = NLA_U8 },
1358 [IFLA_LINKMODE] = { .type = NLA_U8 },
1359 [IFLA_LINKINFO] = { .type = NLA_NESTED },
1360 [IFLA_NET_NS_PID] = { .type = NLA_U32 },
1361 [IFLA_NET_NS_FD] = { .type = NLA_U32 },
1362 [IFLA_IFALIAS] = { .type = NLA_STRING, .len = IFALIASZ-1 },
1363 [IFLA_VFINFO_LIST] = {. type = NLA_NESTED },
1364 [IFLA_VF_PORTS] = { .type = NLA_NESTED },
1365 [IFLA_PORT_SELF] = { .type = NLA_NESTED },
1366 [IFLA_AF_SPEC] = { .type = NLA_NESTED },
1367 [IFLA_EXT_MASK] = { .type = NLA_U32 },
1368 [IFLA_PROMISCUITY] = { .type = NLA_U32 },
1369 [IFLA_NUM_TX_QUEUES] = { .type = NLA_U32 },
1370 [IFLA_NUM_RX_QUEUES] = { .type = NLA_U32 },
1371 [IFLA_PHYS_PORT_ID] = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1372 [IFLA_CARRIER_CHANGES] = { .type = NLA_U32 }, /* ignored */
1373 [IFLA_PHYS_SWITCH_ID] = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1374 [IFLA_LINK_NETNSID] = { .type = NLA_S32 },
1375 [IFLA_PROTO_DOWN] = { .type = NLA_U8 },
1376 [IFLA_GROUP] = { .type = NLA_U32 },
1379 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1380 [IFLA_INFO_KIND] = { .type = NLA_STRING },
1381 [IFLA_INFO_DATA] = { .type = NLA_NESTED },
1382 [IFLA_INFO_SLAVE_KIND] = { .type = NLA_STRING },
1383 [IFLA_INFO_SLAVE_DATA] = { .type = NLA_NESTED },
1386 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1387 [IFLA_VF_MAC] = { .len = sizeof(struct ifla_vf_mac) },
1388 [IFLA_VF_VLAN] = { .len = sizeof(struct ifla_vf_vlan) },
1389 [IFLA_VF_TX_RATE] = { .len = sizeof(struct ifla_vf_tx_rate) },
1390 [IFLA_VF_SPOOFCHK] = { .len = sizeof(struct ifla_vf_spoofchk) },
1391 [IFLA_VF_RATE] = { .len = sizeof(struct ifla_vf_rate) },
1392 [IFLA_VF_LINK_STATE] = { .len = sizeof(struct ifla_vf_link_state) },
1393 [IFLA_VF_RSS_QUERY_EN] = { .len = sizeof(struct ifla_vf_rss_query_en) },
1394 [IFLA_VF_STATS] = { .type = NLA_NESTED },
1395 [IFLA_VF_TRUST] = { .len = sizeof(struct ifla_vf_trust) },
1398 static const struct nla_policy ifla_vf_stats_policy[IFLA_VF_STATS_MAX + 1] = {
1399 [IFLA_VF_STATS_RX_PACKETS] = { .type = NLA_U64 },
1400 [IFLA_VF_STATS_TX_PACKETS] = { .type = NLA_U64 },
1401 [IFLA_VF_STATS_RX_BYTES] = { .type = NLA_U64 },
1402 [IFLA_VF_STATS_TX_BYTES] = { .type = NLA_U64 },
1403 [IFLA_VF_STATS_BROADCAST] = { .type = NLA_U64 },
1404 [IFLA_VF_STATS_MULTICAST] = { .type = NLA_U64 },
1407 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1408 [IFLA_PORT_VF] = { .type = NLA_U32 },
1409 [IFLA_PORT_PROFILE] = { .type = NLA_STRING,
1410 .len = PORT_PROFILE_MAX },
1411 [IFLA_PORT_VSI_TYPE] = { .type = NLA_BINARY,
1412 .len = sizeof(struct ifla_port_vsi)},
1413 [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1414 .len = PORT_UUID_MAX },
1415 [IFLA_PORT_HOST_UUID] = { .type = NLA_STRING,
1416 .len = PORT_UUID_MAX },
1417 [IFLA_PORT_REQUEST] = { .type = NLA_U8, },
1418 [IFLA_PORT_RESPONSE] = { .type = NLA_U16, },
1421 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1423 struct net *net = sock_net(skb->sk);
1426 struct net_device *dev;
1427 struct hlist_head *head;
1428 struct nlattr *tb[IFLA_MAX+1];
1429 u32 ext_filter_mask = 0;
1434 s_idx = cb->args[1];
1436 cb->seq = net->dev_base_seq;
1438 /* A hack to preserve kernel<->userspace interface.
1439 * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
1440 * However, before Linux v3.9 the code here assumed rtgenmsg and that's
1441 * what iproute2 < v3.9.0 used.
1442 * We can detect the old iproute2. Even including the IFLA_EXT_MASK
1443 * attribute, its netlink message is shorter than struct ifinfomsg.
1445 hdrlen = nlmsg_len(cb->nlh) < sizeof(struct ifinfomsg) ?
1446 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
1448 if (nlmsg_parse(cb->nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
1450 if (tb[IFLA_EXT_MASK])
1451 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1454 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1456 head = &net->dev_index_head[h];
1457 hlist_for_each_entry(dev, head, index_hlist) {
1460 err = rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1461 NETLINK_CB(cb->skb).portid,
1462 cb->nlh->nlmsg_seq, 0,
1467 if (likely(skb->len))
1473 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1487 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len)
1489 return nla_parse(tb, IFLA_MAX, head, len, ifla_policy);
1491 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
1493 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1496 /* Examine the link attributes and figure out which
1497 * network namespace we are talking about.
1499 if (tb[IFLA_NET_NS_PID])
1500 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1501 else if (tb[IFLA_NET_NS_FD])
1502 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1504 net = get_net(src_net);
1507 EXPORT_SYMBOL(rtnl_link_get_net);
1509 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1512 if (tb[IFLA_ADDRESS] &&
1513 nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1516 if (tb[IFLA_BROADCAST] &&
1517 nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1521 if (tb[IFLA_AF_SPEC]) {
1525 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1526 const struct rtnl_af_ops *af_ops;
1528 if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1529 return -EAFNOSUPPORT;
1531 if (!af_ops->set_link_af)
1534 if (af_ops->validate_link_af) {
1535 err = af_ops->validate_link_af(dev, af);
1545 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb)
1547 const struct net_device_ops *ops = dev->netdev_ops;
1550 if (tb[IFLA_VF_MAC]) {
1551 struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]);
1553 if (ivm->vf >= INT_MAX)
1556 if (ops->ndo_set_vf_mac)
1557 err = ops->ndo_set_vf_mac(dev, ivm->vf,
1563 if (tb[IFLA_VF_VLAN]) {
1564 struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]);
1566 if (ivv->vf >= INT_MAX)
1569 if (ops->ndo_set_vf_vlan)
1570 err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan,
1576 if (tb[IFLA_VF_TX_RATE]) {
1577 struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]);
1578 struct ifla_vf_info ivf;
1580 if (ivt->vf >= INT_MAX)
1583 if (ops->ndo_get_vf_config)
1584 err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf);
1589 if (ops->ndo_set_vf_rate)
1590 err = ops->ndo_set_vf_rate(dev, ivt->vf,
1597 if (tb[IFLA_VF_RATE]) {
1598 struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]);
1600 if (ivt->vf >= INT_MAX)
1603 if (ops->ndo_set_vf_rate)
1604 err = ops->ndo_set_vf_rate(dev, ivt->vf,
1611 if (tb[IFLA_VF_SPOOFCHK]) {
1612 struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]);
1614 if (ivs->vf >= INT_MAX)
1617 if (ops->ndo_set_vf_spoofchk)
1618 err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1624 if (tb[IFLA_VF_LINK_STATE]) {
1625 struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]);
1627 if (ivl->vf >= INT_MAX)
1630 if (ops->ndo_set_vf_link_state)
1631 err = ops->ndo_set_vf_link_state(dev, ivl->vf,
1637 if (tb[IFLA_VF_RSS_QUERY_EN]) {
1638 struct ifla_vf_rss_query_en *ivrssq_en;
1641 ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]);
1642 if (ivrssq_en->vf >= INT_MAX)
1644 if (ops->ndo_set_vf_rss_query_en)
1645 err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf,
1646 ivrssq_en->setting);
1651 if (tb[IFLA_VF_TRUST]) {
1652 struct ifla_vf_trust *ivt = nla_data(tb[IFLA_VF_TRUST]);
1654 if (ivt->vf >= INT_MAX)
1657 if (ops->ndo_set_vf_trust)
1658 err = ops->ndo_set_vf_trust(dev, ivt->vf, ivt->setting);
1666 static int do_set_master(struct net_device *dev, int ifindex)
1668 struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1669 const struct net_device_ops *ops;
1673 if (upper_dev->ifindex == ifindex)
1675 ops = upper_dev->netdev_ops;
1676 if (ops->ndo_del_slave) {
1677 err = ops->ndo_del_slave(upper_dev, dev);
1686 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
1689 ops = upper_dev->netdev_ops;
1690 if (ops->ndo_add_slave) {
1691 err = ops->ndo_add_slave(upper_dev, dev);
1701 #define DO_SETLINK_MODIFIED 0x01
1702 /* notify flag means notify + modified. */
1703 #define DO_SETLINK_NOTIFY 0x03
1704 static int do_setlink(const struct sk_buff *skb,
1705 struct net_device *dev, struct ifinfomsg *ifm,
1706 struct nlattr **tb, char *ifname, int status)
1708 const struct net_device_ops *ops = dev->netdev_ops;
1711 err = validate_linkmsg(dev, tb);
1715 if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1716 struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1721 if (!netlink_ns_capable(skb, net->user_ns, CAP_NET_ADMIN)) {
1726 err = dev_change_net_namespace(dev, net, ifname);
1730 status |= DO_SETLINK_MODIFIED;
1734 struct rtnl_link_ifmap *u_map;
1737 if (!ops->ndo_set_config) {
1742 if (!netif_device_present(dev)) {
1747 u_map = nla_data(tb[IFLA_MAP]);
1748 k_map.mem_start = (unsigned long) u_map->mem_start;
1749 k_map.mem_end = (unsigned long) u_map->mem_end;
1750 k_map.base_addr = (unsigned short) u_map->base_addr;
1751 k_map.irq = (unsigned char) u_map->irq;
1752 k_map.dma = (unsigned char) u_map->dma;
1753 k_map.port = (unsigned char) u_map->port;
1755 err = ops->ndo_set_config(dev, &k_map);
1759 status |= DO_SETLINK_NOTIFY;
1762 if (tb[IFLA_ADDRESS]) {
1763 struct sockaddr *sa;
1766 len = sizeof(sa_family_t) + max_t(size_t, dev->addr_len,
1768 sa = kmalloc(len, GFP_KERNEL);
1773 sa->sa_family = dev->type;
1774 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1776 err = dev_set_mac_address(dev, sa);
1780 status |= DO_SETLINK_MODIFIED;
1784 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1787 status |= DO_SETLINK_MODIFIED;
1790 if (tb[IFLA_GROUP]) {
1791 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1792 status |= DO_SETLINK_NOTIFY;
1796 * Interface selected by interface index but interface
1797 * name provided implies that a name change has been
1800 if (ifm->ifi_index > 0 && ifname[0]) {
1801 err = dev_change_name(dev, ifname);
1804 status |= DO_SETLINK_MODIFIED;
1807 if (tb[IFLA_IFALIAS]) {
1808 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
1809 nla_len(tb[IFLA_IFALIAS]));
1812 status |= DO_SETLINK_NOTIFY;
1815 if (tb[IFLA_BROADCAST]) {
1816 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
1817 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
1820 if (ifm->ifi_flags || ifm->ifi_change) {
1821 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1826 if (tb[IFLA_MASTER]) {
1827 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
1830 status |= DO_SETLINK_MODIFIED;
1833 if (tb[IFLA_CARRIER]) {
1834 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
1837 status |= DO_SETLINK_MODIFIED;
1840 if (tb[IFLA_TXQLEN]) {
1841 unsigned long value = nla_get_u32(tb[IFLA_TXQLEN]);
1843 if (dev->tx_queue_len ^ value)
1844 status |= DO_SETLINK_NOTIFY;
1846 dev->tx_queue_len = value;
1849 if (tb[IFLA_OPERSTATE])
1850 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1852 if (tb[IFLA_LINKMODE]) {
1853 unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
1855 write_lock_bh(&dev_base_lock);
1856 if (dev->link_mode ^ value)
1857 status |= DO_SETLINK_NOTIFY;
1858 dev->link_mode = value;
1859 write_unlock_bh(&dev_base_lock);
1862 if (tb[IFLA_VFINFO_LIST]) {
1863 struct nlattr *vfinfo[IFLA_VF_MAX + 1];
1864 struct nlattr *attr;
1867 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
1868 if (nla_type(attr) != IFLA_VF_INFO ||
1869 nla_len(attr) < NLA_HDRLEN) {
1873 err = nla_parse_nested(vfinfo, IFLA_VF_MAX, attr,
1877 err = do_setvfinfo(dev, vfinfo);
1880 status |= DO_SETLINK_NOTIFY;
1885 if (tb[IFLA_VF_PORTS]) {
1886 struct nlattr *port[IFLA_PORT_MAX+1];
1887 struct nlattr *attr;
1892 if (!ops->ndo_set_vf_port)
1895 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
1896 if (nla_type(attr) != IFLA_VF_PORT ||
1897 nla_len(attr) < NLA_HDRLEN) {
1901 err = nla_parse_nested(port, IFLA_PORT_MAX, attr,
1905 if (!port[IFLA_PORT_VF]) {
1909 vf = nla_get_u32(port[IFLA_PORT_VF]);
1910 err = ops->ndo_set_vf_port(dev, vf, port);
1913 status |= DO_SETLINK_NOTIFY;
1918 if (tb[IFLA_PORT_SELF]) {
1919 struct nlattr *port[IFLA_PORT_MAX+1];
1921 err = nla_parse_nested(port, IFLA_PORT_MAX,
1922 tb[IFLA_PORT_SELF], ifla_port_policy);
1927 if (ops->ndo_set_vf_port)
1928 err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
1931 status |= DO_SETLINK_NOTIFY;
1934 if (tb[IFLA_AF_SPEC]) {
1938 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1939 const struct rtnl_af_ops *af_ops;
1941 if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1944 err = af_ops->set_link_af(dev, af);
1948 status |= DO_SETLINK_NOTIFY;
1953 if (tb[IFLA_PROTO_DOWN]) {
1954 err = dev_change_proto_down(dev,
1955 nla_get_u8(tb[IFLA_PROTO_DOWN]));
1958 status |= DO_SETLINK_NOTIFY;
1962 if (status & DO_SETLINK_MODIFIED) {
1963 if (status & DO_SETLINK_NOTIFY)
1964 netdev_state_change(dev);
1967 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",
1974 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
1976 struct net *net = sock_net(skb->sk);
1977 struct ifinfomsg *ifm;
1978 struct net_device *dev;
1980 struct nlattr *tb[IFLA_MAX+1];
1981 char ifname[IFNAMSIZ];
1983 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1987 if (tb[IFLA_IFNAME])
1988 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1993 ifm = nlmsg_data(nlh);
1994 if (ifm->ifi_index > 0)
1995 dev = __dev_get_by_index(net, ifm->ifi_index);
1996 else if (tb[IFLA_IFNAME])
1997 dev = __dev_get_by_name(net, ifname);
2006 err = do_setlink(skb, dev, ifm, tb, ifname, 0);
2011 static int rtnl_group_dellink(const struct net *net, int group)
2013 struct net_device *dev, *aux;
2014 LIST_HEAD(list_kill);
2020 for_each_netdev(net, dev) {
2021 if (dev->group == group) {
2022 const struct rtnl_link_ops *ops;
2025 ops = dev->rtnl_link_ops;
2026 if (!ops || !ops->dellink)
2034 for_each_netdev_safe(net, dev, aux) {
2035 if (dev->group == group) {
2036 const struct rtnl_link_ops *ops;
2038 ops = dev->rtnl_link_ops;
2039 ops->dellink(dev, &list_kill);
2042 unregister_netdevice_many(&list_kill);
2047 int rtnl_delete_link(struct net_device *dev)
2049 const struct rtnl_link_ops *ops;
2050 LIST_HEAD(list_kill);
2052 ops = dev->rtnl_link_ops;
2053 if (!ops || !ops->dellink)
2056 ops->dellink(dev, &list_kill);
2057 unregister_netdevice_many(&list_kill);
2061 EXPORT_SYMBOL_GPL(rtnl_delete_link);
2063 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
2065 struct net *net = sock_net(skb->sk);
2066 struct net_device *dev;
2067 struct ifinfomsg *ifm;
2068 char ifname[IFNAMSIZ];
2069 struct nlattr *tb[IFLA_MAX+1];
2072 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2076 if (tb[IFLA_IFNAME])
2077 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2079 ifm = nlmsg_data(nlh);
2080 if (ifm->ifi_index > 0)
2081 dev = __dev_get_by_index(net, ifm->ifi_index);
2082 else if (tb[IFLA_IFNAME])
2083 dev = __dev_get_by_name(net, ifname);
2084 else if (tb[IFLA_GROUP])
2085 return rtnl_group_dellink(net, nla_get_u32(tb[IFLA_GROUP]));
2092 return rtnl_delete_link(dev);
2095 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
2097 unsigned int old_flags;
2100 old_flags = dev->flags;
2101 if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
2102 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
2107 if (dev->rtnl_link_state == RTNL_LINK_INITIALIZED) {
2108 __dev_notify_flags(dev, old_flags, (old_flags ^ dev->flags));
2110 dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
2111 __dev_notify_flags(dev, old_flags, ~0U);
2115 EXPORT_SYMBOL(rtnl_configure_link);
2117 struct net_device *rtnl_create_link(struct net *net,
2118 const char *ifname, unsigned char name_assign_type,
2119 const struct rtnl_link_ops *ops, struct nlattr *tb[])
2122 struct net_device *dev;
2123 unsigned int num_tx_queues = 1;
2124 unsigned int num_rx_queues = 1;
2126 if (tb[IFLA_NUM_TX_QUEUES])
2127 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
2128 else if (ops->get_num_tx_queues)
2129 num_tx_queues = ops->get_num_tx_queues();
2131 if (tb[IFLA_NUM_RX_QUEUES])
2132 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
2133 else if (ops->get_num_rx_queues)
2134 num_rx_queues = ops->get_num_rx_queues();
2136 if (num_tx_queues < 1 || num_tx_queues > 4096)
2137 return ERR_PTR(-EINVAL);
2139 if (num_rx_queues < 1 || num_rx_queues > 4096)
2140 return ERR_PTR(-EINVAL);
2143 dev = alloc_netdev_mqs(ops->priv_size, ifname, name_assign_type,
2144 ops->setup, num_tx_queues, num_rx_queues);
2148 dev_net_set(dev, net);
2149 dev->rtnl_link_ops = ops;
2150 dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
2153 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
2154 if (tb[IFLA_ADDRESS]) {
2155 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
2156 nla_len(tb[IFLA_ADDRESS]));
2157 dev->addr_assign_type = NET_ADDR_SET;
2159 if (tb[IFLA_BROADCAST])
2160 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
2161 nla_len(tb[IFLA_BROADCAST]));
2162 if (tb[IFLA_TXQLEN])
2163 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
2164 if (tb[IFLA_OPERSTATE])
2165 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2166 if (tb[IFLA_LINKMODE])
2167 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
2169 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2174 return ERR_PTR(err);
2176 EXPORT_SYMBOL(rtnl_create_link);
2178 static int rtnl_group_changelink(const struct sk_buff *skb,
2179 struct net *net, int group,
2180 struct ifinfomsg *ifm,
2183 struct net_device *dev, *aux;
2186 for_each_netdev_safe(net, dev, aux) {
2187 if (dev->group == group) {
2188 err = do_setlink(skb, dev, ifm, tb, NULL, 0);
2197 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh)
2199 struct net *net = sock_net(skb->sk);
2200 const struct rtnl_link_ops *ops;
2201 const struct rtnl_link_ops *m_ops = NULL;
2202 struct net_device *dev;
2203 struct net_device *master_dev = NULL;
2204 struct ifinfomsg *ifm;
2205 char kind[MODULE_NAME_LEN];
2206 char ifname[IFNAMSIZ];
2207 struct nlattr *tb[IFLA_MAX+1];
2208 struct nlattr *linkinfo[IFLA_INFO_MAX+1];
2209 unsigned char name_assign_type = NET_NAME_USER;
2212 #ifdef CONFIG_MODULES
2215 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2219 if (tb[IFLA_IFNAME])
2220 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2224 ifm = nlmsg_data(nlh);
2225 if (ifm->ifi_index > 0)
2226 dev = __dev_get_by_index(net, ifm->ifi_index);
2229 dev = __dev_get_by_name(net, ifname);
2235 master_dev = netdev_master_upper_dev_get(dev);
2237 m_ops = master_dev->rtnl_link_ops;
2240 err = validate_linkmsg(dev, tb);
2244 if (tb[IFLA_LINKINFO]) {
2245 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
2246 tb[IFLA_LINKINFO], ifla_info_policy);
2250 memset(linkinfo, 0, sizeof(linkinfo));
2252 if (linkinfo[IFLA_INFO_KIND]) {
2253 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
2254 ops = rtnl_link_ops_get(kind);
2261 struct nlattr *attr[ops ? ops->maxtype + 1 : 1];
2262 struct nlattr *slave_attr[m_ops ? m_ops->slave_maxtype + 1 : 1];
2263 struct nlattr **data = NULL;
2264 struct nlattr **slave_data = NULL;
2265 struct net *dest_net, *link_net = NULL;
2268 if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
2269 err = nla_parse_nested(attr, ops->maxtype,
2270 linkinfo[IFLA_INFO_DATA],
2276 if (ops->validate) {
2277 err = ops->validate(tb, data);
2284 if (m_ops->slave_maxtype &&
2285 linkinfo[IFLA_INFO_SLAVE_DATA]) {
2286 err = nla_parse_nested(slave_attr,
2287 m_ops->slave_maxtype,
2288 linkinfo[IFLA_INFO_SLAVE_DATA],
2289 m_ops->slave_policy);
2292 slave_data = slave_attr;
2294 if (m_ops->slave_validate) {
2295 err = m_ops->slave_validate(tb, slave_data);
2304 if (nlh->nlmsg_flags & NLM_F_EXCL)
2306 if (nlh->nlmsg_flags & NLM_F_REPLACE)
2309 if (linkinfo[IFLA_INFO_DATA]) {
2310 if (!ops || ops != dev->rtnl_link_ops ||
2314 err = ops->changelink(dev, tb, data);
2317 status |= DO_SETLINK_NOTIFY;
2320 if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
2321 if (!m_ops || !m_ops->slave_changelink)
2324 err = m_ops->slave_changelink(master_dev, dev,
2328 status |= DO_SETLINK_NOTIFY;
2331 return do_setlink(skb, dev, ifm, tb, ifname, status);
2334 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
2335 if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
2336 return rtnl_group_changelink(skb, net,
2337 nla_get_u32(tb[IFLA_GROUP]),
2342 if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
2346 #ifdef CONFIG_MODULES
2349 request_module("rtnl-link-%s", kind);
2351 ops = rtnl_link_ops_get(kind);
2363 snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
2364 name_assign_type = NET_NAME_ENUM;
2367 dest_net = rtnl_link_get_net(net, tb);
2368 if (IS_ERR(dest_net))
2369 return PTR_ERR(dest_net);
2372 if (!netlink_ns_capable(skb, dest_net->user_ns, CAP_NET_ADMIN))
2375 if (tb[IFLA_LINK_NETNSID]) {
2376 int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
2378 link_net = get_net_ns_by_id(dest_net, id);
2384 if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
2388 dev = rtnl_create_link(link_net ? : dest_net, ifname,
2389 name_assign_type, ops, tb);
2395 dev->ifindex = ifm->ifi_index;
2398 err = ops->newlink(link_net ? : net, dev, tb, data);
2399 /* Drivers should call free_netdev() in ->destructor
2400 * and unregister it on failure after registration
2401 * so that device could be finally freed in rtnl_unlock.
2404 /* If device is not registered at all, free it now */
2405 if (dev->reg_state == NETREG_UNINITIALIZED)
2410 err = register_netdevice(dev);
2416 err = rtnl_configure_link(dev, ifm);
2418 goto out_unregister;
2420 err = dev_change_net_namespace(dev, dest_net, ifname);
2422 goto out_unregister;
2431 LIST_HEAD(list_kill);
2433 ops->dellink(dev, &list_kill);
2434 unregister_netdevice_many(&list_kill);
2436 unregister_netdevice(dev);
2442 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh)
2444 struct net *net = sock_net(skb->sk);
2445 struct ifinfomsg *ifm;
2446 char ifname[IFNAMSIZ];
2447 struct nlattr *tb[IFLA_MAX+1];
2448 struct net_device *dev = NULL;
2449 struct sk_buff *nskb;
2451 u32 ext_filter_mask = 0;
2453 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2457 if (tb[IFLA_IFNAME])
2458 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2460 if (tb[IFLA_EXT_MASK])
2461 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2463 ifm = nlmsg_data(nlh);
2464 if (ifm->ifi_index > 0)
2465 dev = __dev_get_by_index(net, ifm->ifi_index);
2466 else if (tb[IFLA_IFNAME])
2467 dev = __dev_get_by_name(net, ifname);
2474 nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
2478 err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).portid,
2479 nlh->nlmsg_seq, 0, 0, ext_filter_mask);
2481 /* -EMSGSIZE implies BUG in if_nlmsg_size */
2482 WARN_ON(err == -EMSGSIZE);
2485 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
2490 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
2492 struct net *net = sock_net(skb->sk);
2493 struct net_device *dev;
2494 struct nlattr *tb[IFLA_MAX+1];
2495 u32 ext_filter_mask = 0;
2496 u16 min_ifinfo_dump_size = 0;
2499 /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
2500 hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
2501 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
2503 if (nlmsg_parse(nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
2504 if (tb[IFLA_EXT_MASK])
2505 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2508 if (!ext_filter_mask)
2509 return NLMSG_GOODSIZE;
2511 * traverse the list of net devices and compute the minimum
2512 * buffer size based upon the filter mask.
2514 list_for_each_entry(dev, &net->dev_base_head, dev_list) {
2515 min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
2520 return min_ifinfo_dump_size;
2523 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
2526 int s_idx = cb->family;
2530 for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
2531 int type = cb->nlh->nlmsg_type-RTM_BASE;
2532 if (idx < s_idx || idx == PF_PACKET)
2534 if (rtnl_msg_handlers[idx] == NULL ||
2535 rtnl_msg_handlers[idx][type].dumpit == NULL)
2538 memset(&cb->args[0], 0, sizeof(cb->args));
2542 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
2550 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
2551 unsigned int change, gfp_t flags)
2553 struct net *net = dev_net(dev);
2554 struct sk_buff *skb;
2556 size_t if_info_size;
2558 skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
2562 err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0);
2564 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
2565 WARN_ON(err == -EMSGSIZE);
2572 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2576 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags)
2578 struct net *net = dev_net(dev);
2580 rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
2583 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
2586 struct sk_buff *skb;
2588 if (dev->reg_state != NETREG_REGISTERED)
2591 skb = rtmsg_ifinfo_build_skb(type, dev, change, flags);
2593 rtmsg_ifinfo_send(skb, dev, flags);
2595 EXPORT_SYMBOL(rtmsg_ifinfo);
2597 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
2598 struct net_device *dev,
2599 u8 *addr, u16 vid, u32 pid, u32 seq,
2600 int type, unsigned int flags,
2603 struct nlmsghdr *nlh;
2606 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
2610 ndm = nlmsg_data(nlh);
2611 ndm->ndm_family = AF_BRIDGE;
2614 ndm->ndm_flags = flags;
2616 ndm->ndm_ifindex = dev->ifindex;
2617 ndm->ndm_state = NUD_PERMANENT;
2619 if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
2620 goto nla_put_failure;
2622 if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
2623 goto nla_put_failure;
2625 nlmsg_end(skb, nlh);
2629 nlmsg_cancel(skb, nlh);
2633 static inline size_t rtnl_fdb_nlmsg_size(void)
2635 return NLMSG_ALIGN(sizeof(struct ndmsg)) +
2636 nla_total_size(ETH_ALEN) + /* NDA_LLADDR */
2637 nla_total_size(sizeof(u16)) + /* NDA_VLAN */
2641 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type)
2643 struct net *net = dev_net(dev);
2644 struct sk_buff *skb;
2647 skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
2651 err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
2652 0, 0, type, NTF_SELF, 0);
2658 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2661 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
2665 * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
2667 int ndo_dflt_fdb_add(struct ndmsg *ndm,
2668 struct nlattr *tb[],
2669 struct net_device *dev,
2670 const unsigned char *addr, u16 vid,
2675 /* If aging addresses are supported device will need to
2676 * implement its own handler for this.
2678 if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
2679 pr_info("%s: FDB only supports static addresses\n", dev->name);
2684 pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
2688 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2689 err = dev_uc_add_excl(dev, addr);
2690 else if (is_multicast_ether_addr(addr))
2691 err = dev_mc_add_excl(dev, addr);
2693 /* Only return duplicate errors if NLM_F_EXCL is set */
2694 if (err == -EEXIST && !(flags & NLM_F_EXCL))
2699 EXPORT_SYMBOL(ndo_dflt_fdb_add);
2701 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid)
2706 if (nla_len(vlan_attr) != sizeof(u16)) {
2707 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan\n");
2711 vid = nla_get_u16(vlan_attr);
2713 if (!vid || vid >= VLAN_VID_MASK) {
2714 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan id %d\n",
2723 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh)
2725 struct net *net = sock_net(skb->sk);
2727 struct nlattr *tb[NDA_MAX+1];
2728 struct net_device *dev;
2733 err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2737 ndm = nlmsg_data(nlh);
2738 if (ndm->ndm_ifindex == 0) {
2739 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n");
2743 dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2745 pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n");
2749 if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2750 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n");
2754 if (dev->type != ARPHRD_ETHER) {
2755 pr_info("PF_BRIDGE: FDB add only supported for Ethernet devices");
2759 addr = nla_data(tb[NDA_LLADDR]);
2761 err = fdb_vid_parse(tb[NDA_VLAN], &vid);
2767 /* Support fdb on master device the net/bridge default case */
2768 if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2769 (dev->priv_flags & IFF_BRIDGE_PORT)) {
2770 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2771 const struct net_device_ops *ops = br_dev->netdev_ops;
2773 err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
2778 ndm->ndm_flags &= ~NTF_MASTER;
2781 /* Embedded bridge, macvlan, and any other device support */
2782 if ((ndm->ndm_flags & NTF_SELF)) {
2783 if (dev->netdev_ops->ndo_fdb_add)
2784 err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
2788 err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
2792 rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH);
2793 ndm->ndm_flags &= ~NTF_SELF;
2801 * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
2803 int ndo_dflt_fdb_del(struct ndmsg *ndm,
2804 struct nlattr *tb[],
2805 struct net_device *dev,
2806 const unsigned char *addr, u16 vid)
2810 /* If aging addresses are supported device will need to
2811 * implement its own handler for this.
2813 if (!(ndm->ndm_state & NUD_PERMANENT)) {
2814 pr_info("%s: FDB only supports static addresses\n", dev->name);
2818 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2819 err = dev_uc_del(dev, addr);
2820 else if (is_multicast_ether_addr(addr))
2821 err = dev_mc_del(dev, addr);
2825 EXPORT_SYMBOL(ndo_dflt_fdb_del);
2827 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh)
2829 struct net *net = sock_net(skb->sk);
2831 struct nlattr *tb[NDA_MAX+1];
2832 struct net_device *dev;
2837 if (!netlink_capable(skb, CAP_NET_ADMIN))
2840 err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2844 ndm = nlmsg_data(nlh);
2845 if (ndm->ndm_ifindex == 0) {
2846 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n");
2850 dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2852 pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n");
2856 if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2857 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid address\n");
2861 if (dev->type != ARPHRD_ETHER) {
2862 pr_info("PF_BRIDGE: FDB delete only supported for Ethernet devices");
2866 addr = nla_data(tb[NDA_LLADDR]);
2868 err = fdb_vid_parse(tb[NDA_VLAN], &vid);
2874 /* Support fdb on master device the net/bridge default case */
2875 if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2876 (dev->priv_flags & IFF_BRIDGE_PORT)) {
2877 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2878 const struct net_device_ops *ops = br_dev->netdev_ops;
2880 if (ops->ndo_fdb_del)
2881 err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid);
2886 ndm->ndm_flags &= ~NTF_MASTER;
2889 /* Embedded bridge, macvlan, and any other device support */
2890 if (ndm->ndm_flags & NTF_SELF) {
2891 if (dev->netdev_ops->ndo_fdb_del)
2892 err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr,
2895 err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
2898 rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH);
2899 ndm->ndm_flags &= ~NTF_SELF;
2906 static int nlmsg_populate_fdb(struct sk_buff *skb,
2907 struct netlink_callback *cb,
2908 struct net_device *dev,
2910 struct netdev_hw_addr_list *list)
2912 struct netdev_hw_addr *ha;
2916 portid = NETLINK_CB(cb->skb).portid;
2917 seq = cb->nlh->nlmsg_seq;
2919 list_for_each_entry(ha, &list->list, list) {
2920 if (*idx < cb->args[0])
2923 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
2925 RTM_NEWNEIGH, NTF_SELF,
2936 * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
2937 * @nlh: netlink message header
2940 * Default netdevice operation to dump the existing unicast address list.
2941 * Returns number of addresses from list put in skb.
2943 int ndo_dflt_fdb_dump(struct sk_buff *skb,
2944 struct netlink_callback *cb,
2945 struct net_device *dev,
2946 struct net_device *filter_dev,
2951 if (dev->type != ARPHRD_ETHER)
2954 netif_addr_lock_bh(dev);
2955 err = nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->uc);
2958 nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->mc);
2960 netif_addr_unlock_bh(dev);
2963 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
2965 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
2967 struct net_device *dev;
2968 struct nlattr *tb[IFLA_MAX+1];
2969 struct net_device *br_dev = NULL;
2970 const struct net_device_ops *ops = NULL;
2971 const struct net_device_ops *cops = NULL;
2972 struct ifinfomsg *ifm = nlmsg_data(cb->nlh);
2973 struct net *net = sock_net(skb->sk);
2978 if (nlmsg_parse(cb->nlh, sizeof(struct ifinfomsg), tb, IFLA_MAX,
2979 ifla_policy) == 0) {
2980 if (tb[IFLA_MASTER])
2981 br_idx = nla_get_u32(tb[IFLA_MASTER]);
2984 brport_idx = ifm->ifi_index;
2987 br_dev = __dev_get_by_index(net, br_idx);
2991 ops = br_dev->netdev_ops;
2994 for_each_netdev(net, dev) {
2995 if (brport_idx && (dev->ifindex != brport_idx))
2998 if (!br_idx) { /* user did not specify a specific bridge */
2999 if (dev->priv_flags & IFF_BRIDGE_PORT) {
3000 br_dev = netdev_master_upper_dev_get(dev);
3001 cops = br_dev->netdev_ops;
3005 if (dev != br_dev &&
3006 !(dev->priv_flags & IFF_BRIDGE_PORT))
3009 if (br_dev != netdev_master_upper_dev_get(dev) &&
3010 !(dev->priv_flags & IFF_EBRIDGE))
3016 if (dev->priv_flags & IFF_BRIDGE_PORT) {
3017 if (cops && cops->ndo_fdb_dump)
3018 idx = cops->ndo_fdb_dump(skb, cb, br_dev, dev,
3022 if (dev->netdev_ops->ndo_fdb_dump)
3023 idx = dev->netdev_ops->ndo_fdb_dump(skb, cb, dev, NULL,
3026 idx = ndo_dflt_fdb_dump(skb, cb, dev, NULL, idx);
3035 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
3036 unsigned int attrnum, unsigned int flag)
3039 return nla_put_u8(skb, attrnum, !!(flags & flag));
3043 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
3044 struct net_device *dev, u16 mode,
3045 u32 flags, u32 mask, int nlflags,
3047 int (*vlan_fill)(struct sk_buff *skb,
3048 struct net_device *dev,
3051 struct nlmsghdr *nlh;
3052 struct ifinfomsg *ifm;
3053 struct nlattr *br_afspec;
3054 struct nlattr *protinfo;
3055 u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
3056 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3059 nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
3063 ifm = nlmsg_data(nlh);
3064 ifm->ifi_family = AF_BRIDGE;
3066 ifm->ifi_type = dev->type;
3067 ifm->ifi_index = dev->ifindex;
3068 ifm->ifi_flags = dev_get_flags(dev);
3069 ifm->ifi_change = 0;
3072 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
3073 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
3074 nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
3076 nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
3078 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
3079 (dev->ifindex != dev_get_iflink(dev) &&
3080 nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
3081 goto nla_put_failure;
3083 br_afspec = nla_nest_start(skb, IFLA_AF_SPEC);
3085 goto nla_put_failure;
3087 if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
3088 nla_nest_cancel(skb, br_afspec);
3089 goto nla_put_failure;
3092 if (mode != BRIDGE_MODE_UNDEF) {
3093 if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
3094 nla_nest_cancel(skb, br_afspec);
3095 goto nla_put_failure;
3099 err = vlan_fill(skb, dev, filter_mask);
3101 nla_nest_cancel(skb, br_afspec);
3102 goto nla_put_failure;
3105 nla_nest_end(skb, br_afspec);
3107 protinfo = nla_nest_start(skb, IFLA_PROTINFO | NLA_F_NESTED);
3109 goto nla_put_failure;
3111 if (brport_nla_put_flag(skb, flags, mask,
3112 IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
3113 brport_nla_put_flag(skb, flags, mask,
3114 IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
3115 brport_nla_put_flag(skb, flags, mask,
3116 IFLA_BRPORT_FAST_LEAVE,
3117 BR_MULTICAST_FAST_LEAVE) ||
3118 brport_nla_put_flag(skb, flags, mask,
3119 IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
3120 brport_nla_put_flag(skb, flags, mask,
3121 IFLA_BRPORT_LEARNING, BR_LEARNING) ||
3122 brport_nla_put_flag(skb, flags, mask,
3123 IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
3124 brport_nla_put_flag(skb, flags, mask,
3125 IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
3126 brport_nla_put_flag(skb, flags, mask,
3127 IFLA_BRPORT_PROXYARP, BR_PROXYARP)) {
3128 nla_nest_cancel(skb, protinfo);
3129 goto nla_put_failure;
3132 nla_nest_end(skb, protinfo);
3134 nlmsg_end(skb, nlh);
3137 nlmsg_cancel(skb, nlh);
3138 return err ? err : -EMSGSIZE;
3140 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink);
3142 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
3144 struct net *net = sock_net(skb->sk);
3145 struct net_device *dev;
3147 u32 portid = NETLINK_CB(cb->skb).portid;
3148 u32 seq = cb->nlh->nlmsg_seq;
3149 u32 filter_mask = 0;
3152 if (nlmsg_len(cb->nlh) > sizeof(struct ifinfomsg)) {
3153 struct nlattr *extfilt;
3155 extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct ifinfomsg),
3158 if (nla_len(extfilt) < sizeof(filter_mask))
3161 filter_mask = nla_get_u32(extfilt);
3166 for_each_netdev_rcu(net, dev) {
3167 const struct net_device_ops *ops = dev->netdev_ops;
3168 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3170 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
3171 if (idx >= cb->args[0]) {
3172 err = br_dev->netdev_ops->ndo_bridge_getlink(
3173 skb, portid, seq, dev,
3174 filter_mask, NLM_F_MULTI);
3175 if (err < 0 && err != -EOPNOTSUPP) {
3176 if (likely(skb->len))
3185 if (ops->ndo_bridge_getlink) {
3186 if (idx >= cb->args[0]) {
3187 err = ops->ndo_bridge_getlink(skb, portid,
3191 if (err < 0 && err != -EOPNOTSUPP) {
3192 if (likely(skb->len))
3209 static inline size_t bridge_nlmsg_size(void)
3211 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3212 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3213 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3214 + nla_total_size(sizeof(u32)) /* IFLA_MASTER */
3215 + nla_total_size(sizeof(u32)) /* IFLA_MTU */
3216 + nla_total_size(sizeof(u32)) /* IFLA_LINK */
3217 + nla_total_size(sizeof(u32)) /* IFLA_OPERSTATE */
3218 + nla_total_size(sizeof(u8)) /* IFLA_PROTINFO */
3219 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
3220 + nla_total_size(sizeof(u16)) /* IFLA_BRIDGE_FLAGS */
3221 + nla_total_size(sizeof(u16)); /* IFLA_BRIDGE_MODE */
3224 static int rtnl_bridge_notify(struct net_device *dev)
3226 struct net *net = dev_net(dev);
3227 struct sk_buff *skb;
3228 int err = -EOPNOTSUPP;
3230 if (!dev->netdev_ops->ndo_bridge_getlink)
3233 skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
3239 err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
3243 /* Notification info is only filled for bridge ports, not the bridge
3244 * device itself. Therefore, a zero notification length is valid and
3245 * should not result in an error.
3250 rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
3253 WARN_ON(err == -EMSGSIZE);
3256 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
3260 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
3262 struct net *net = sock_net(skb->sk);
3263 struct ifinfomsg *ifm;
3264 struct net_device *dev;
3265 struct nlattr *br_spec, *attr = NULL;
3266 int rem, err = -EOPNOTSUPP;
3268 bool have_flags = false;
3270 if (nlmsg_len(nlh) < sizeof(*ifm))
3273 ifm = nlmsg_data(nlh);
3274 if (ifm->ifi_family != AF_BRIDGE)
3275 return -EPFNOSUPPORT;
3277 dev = __dev_get_by_index(net, ifm->ifi_index);
3279 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3283 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3285 nla_for_each_nested(attr, br_spec, rem) {
3286 if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3287 if (nla_len(attr) < sizeof(flags))
3291 flags = nla_get_u16(attr);
3297 if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3298 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3300 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
3305 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags);
3309 flags &= ~BRIDGE_FLAGS_MASTER;
3312 if ((flags & BRIDGE_FLAGS_SELF)) {
3313 if (!dev->netdev_ops->ndo_bridge_setlink)
3316 err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
3319 flags &= ~BRIDGE_FLAGS_SELF;
3321 /* Generate event to notify upper layer of bridge
3324 err = rtnl_bridge_notify(dev);
3329 memcpy(nla_data(attr), &flags, sizeof(flags));
3334 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
3336 struct net *net = sock_net(skb->sk);
3337 struct ifinfomsg *ifm;
3338 struct net_device *dev;
3339 struct nlattr *br_spec, *attr = NULL;
3340 int rem, err = -EOPNOTSUPP;
3342 bool have_flags = false;
3344 if (nlmsg_len(nlh) < sizeof(*ifm))
3347 ifm = nlmsg_data(nlh);
3348 if (ifm->ifi_family != AF_BRIDGE)
3349 return -EPFNOSUPPORT;
3351 dev = __dev_get_by_index(net, ifm->ifi_index);
3353 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3357 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3359 nla_for_each_nested(attr, br_spec, rem) {
3360 if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3361 if (nla_len(attr) < sizeof(flags))
3365 flags = nla_get_u16(attr);
3371 if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3372 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3374 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
3379 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
3383 flags &= ~BRIDGE_FLAGS_MASTER;
3386 if ((flags & BRIDGE_FLAGS_SELF)) {
3387 if (!dev->netdev_ops->ndo_bridge_dellink)
3390 err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
3394 flags &= ~BRIDGE_FLAGS_SELF;
3396 /* Generate event to notify upper layer of bridge
3399 err = rtnl_bridge_notify(dev);
3404 memcpy(nla_data(attr), &flags, sizeof(flags));
3409 /* Process one rtnetlink message. */
3411 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
3413 struct net *net = sock_net(skb->sk);
3414 rtnl_doit_func doit;
3420 type = nlh->nlmsg_type;
3426 /* All the messages must have at least 1 byte length */
3427 if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
3430 family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
3434 if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
3437 if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
3439 rtnl_dumpit_func dumpit;
3440 rtnl_calcit_func calcit;
3441 u16 min_dump_alloc = 0;
3443 dumpit = rtnl_get_dumpit(family, type);
3446 calcit = rtnl_get_calcit(family, type);
3448 min_dump_alloc = calcit(skb, nlh);
3453 struct netlink_dump_control c = {
3455 .min_dump_alloc = min_dump_alloc,
3457 err = netlink_dump_start(rtnl, skb, nlh, &c);
3463 doit = rtnl_get_doit(family, type);
3467 return doit(skb, nlh);
3470 static void rtnetlink_rcv(struct sk_buff *skb)
3473 netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
3477 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
3479 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3485 case NETDEV_POST_INIT:
3486 case NETDEV_REGISTER:
3488 case NETDEV_PRE_TYPE_CHANGE:
3489 case NETDEV_GOING_DOWN:
3490 case NETDEV_UNREGISTER:
3491 case NETDEV_UNREGISTER_FINAL:
3492 case NETDEV_RELEASE:
3494 case NETDEV_BONDING_INFO:
3497 rtmsg_ifinfo(RTM_NEWLINK, dev, 0, GFP_KERNEL);
3503 static struct notifier_block rtnetlink_dev_notifier = {
3504 .notifier_call = rtnetlink_event,
3508 static int __net_init rtnetlink_net_init(struct net *net)
3511 struct netlink_kernel_cfg cfg = {
3512 .groups = RTNLGRP_MAX,
3513 .input = rtnetlink_rcv,
3514 .cb_mutex = &rtnl_mutex,
3515 .flags = NL_CFG_F_NONROOT_RECV,
3518 sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
3525 static void __net_exit rtnetlink_net_exit(struct net *net)
3527 netlink_kernel_release(net->rtnl);
3531 static struct pernet_operations rtnetlink_net_ops = {
3532 .init = rtnetlink_net_init,
3533 .exit = rtnetlink_net_exit,
3536 void __init rtnetlink_init(void)
3538 if (register_pernet_subsys(&rtnetlink_net_ops))
3539 panic("rtnetlink_init: cannot initialize rtnetlink\n");
3541 register_netdevice_notifier(&rtnetlink_dev_notifier);
3543 rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
3544 rtnl_dump_ifinfo, rtnl_calcit);
3545 rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
3546 rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
3547 rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
3549 rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
3550 rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
3552 rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, NULL);
3553 rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, NULL);
3554 rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, NULL);
3556 rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, NULL);
3557 rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, NULL);
3558 rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, NULL);