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;
259 EXPORT_SYMBOL_GPL(rtnl_unregister);
262 * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
263 * @protocol : Protocol family or PF_UNSPEC
265 * Identical to calling rtnl_unregster() for all registered message types
266 * of a certain protocol family.
268 void rtnl_unregister_all(int protocol)
270 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
272 kfree(rtnl_msg_handlers[protocol]);
273 rtnl_msg_handlers[protocol] = NULL;
275 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
277 static LIST_HEAD(link_ops);
279 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
281 const struct rtnl_link_ops *ops;
283 list_for_each_entry(ops, &link_ops, list) {
284 if (!strcmp(ops->kind, kind))
291 * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
292 * @ops: struct rtnl_link_ops * to register
294 * The caller must hold the rtnl_mutex. This function should be used
295 * by drivers that create devices during module initialization. It
296 * must be called before registering the devices.
298 * Returns 0 on success or a negative error code.
300 int __rtnl_link_register(struct rtnl_link_ops *ops)
302 if (rtnl_link_ops_get(ops->kind))
305 /* The check for setup is here because if ops
306 * does not have that filled up, it is not possible
307 * to use the ops for creating device. So do not
308 * fill up dellink as well. That disables rtnl_dellink.
310 if (ops->setup && !ops->dellink)
311 ops->dellink = unregister_netdevice_queue;
313 list_add_tail(&ops->list, &link_ops);
316 EXPORT_SYMBOL_GPL(__rtnl_link_register);
319 * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
320 * @ops: struct rtnl_link_ops * to register
322 * Returns 0 on success or a negative error code.
324 int rtnl_link_register(struct rtnl_link_ops *ops)
329 err = __rtnl_link_register(ops);
333 EXPORT_SYMBOL_GPL(rtnl_link_register);
335 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
337 struct net_device *dev;
338 LIST_HEAD(list_kill);
340 for_each_netdev(net, dev) {
341 if (dev->rtnl_link_ops == ops)
342 ops->dellink(dev, &list_kill);
344 unregister_netdevice_many(&list_kill);
348 * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
349 * @ops: struct rtnl_link_ops * to unregister
351 * The caller must hold the rtnl_mutex.
353 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
358 __rtnl_kill_links(net, ops);
360 list_del(&ops->list);
362 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
364 /* Return with the rtnl_lock held when there are no network
365 * devices unregistering in any network namespace.
367 static void rtnl_lock_unregistering_all(void)
371 DEFINE_WAIT_FUNC(wait, woken_wake_function);
373 add_wait_queue(&netdev_unregistering_wq, &wait);
375 unregistering = false;
378 if (net->dev_unreg_count > 0) {
379 unregistering = true;
387 wait_woken(&wait, TASK_UNINTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
389 remove_wait_queue(&netdev_unregistering_wq, &wait);
393 * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
394 * @ops: struct rtnl_link_ops * to unregister
396 void rtnl_link_unregister(struct rtnl_link_ops *ops)
398 /* Close the race with cleanup_net() */
399 mutex_lock(&net_mutex);
400 rtnl_lock_unregistering_all();
401 __rtnl_link_unregister(ops);
403 mutex_unlock(&net_mutex);
405 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
407 static size_t rtnl_link_get_slave_info_data_size(const struct net_device *dev)
409 struct net_device *master_dev;
410 const struct rtnl_link_ops *ops;
412 master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
415 ops = master_dev->rtnl_link_ops;
416 if (!ops || !ops->get_slave_size)
418 /* IFLA_INFO_SLAVE_DATA + nested data */
419 return nla_total_size(sizeof(struct nlattr)) +
420 ops->get_slave_size(master_dev, dev);
423 static size_t rtnl_link_get_size(const struct net_device *dev)
425 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
431 size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
432 nla_total_size(strlen(ops->kind) + 1); /* IFLA_INFO_KIND */
435 /* IFLA_INFO_DATA + nested data */
436 size += nla_total_size(sizeof(struct nlattr)) +
439 if (ops->get_xstats_size)
440 /* IFLA_INFO_XSTATS */
441 size += nla_total_size(ops->get_xstats_size(dev));
443 size += rtnl_link_get_slave_info_data_size(dev);
448 static LIST_HEAD(rtnl_af_ops);
450 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
452 const struct rtnl_af_ops *ops;
454 list_for_each_entry(ops, &rtnl_af_ops, list) {
455 if (ops->family == family)
463 * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
464 * @ops: struct rtnl_af_ops * to register
466 * Returns 0 on success or a negative error code.
468 void rtnl_af_register(struct rtnl_af_ops *ops)
471 list_add_tail(&ops->list, &rtnl_af_ops);
474 EXPORT_SYMBOL_GPL(rtnl_af_register);
477 * __rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
478 * @ops: struct rtnl_af_ops * to unregister
480 * The caller must hold the rtnl_mutex.
482 void __rtnl_af_unregister(struct rtnl_af_ops *ops)
484 list_del(&ops->list);
486 EXPORT_SYMBOL_GPL(__rtnl_af_unregister);
489 * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
490 * @ops: struct rtnl_af_ops * to unregister
492 void rtnl_af_unregister(struct rtnl_af_ops *ops)
495 __rtnl_af_unregister(ops);
498 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
500 static size_t rtnl_link_get_af_size(const struct net_device *dev,
503 struct rtnl_af_ops *af_ops;
507 size = nla_total_size(sizeof(struct nlattr));
509 list_for_each_entry(af_ops, &rtnl_af_ops, list) {
510 if (af_ops->get_link_af_size) {
511 /* AF_* + nested data */
512 size += nla_total_size(sizeof(struct nlattr)) +
513 af_ops->get_link_af_size(dev, ext_filter_mask);
520 static bool rtnl_have_link_slave_info(const struct net_device *dev)
522 struct net_device *master_dev;
524 master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
525 if (master_dev && master_dev->rtnl_link_ops)
530 static int rtnl_link_slave_info_fill(struct sk_buff *skb,
531 const struct net_device *dev)
533 struct net_device *master_dev;
534 const struct rtnl_link_ops *ops;
535 struct nlattr *slave_data;
538 master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
541 ops = master_dev->rtnl_link_ops;
544 if (nla_put_string(skb, IFLA_INFO_SLAVE_KIND, ops->kind) < 0)
546 if (ops->fill_slave_info) {
547 slave_data = nla_nest_start(skb, IFLA_INFO_SLAVE_DATA);
550 err = ops->fill_slave_info(skb, master_dev, dev);
552 goto err_cancel_slave_data;
553 nla_nest_end(skb, slave_data);
557 err_cancel_slave_data:
558 nla_nest_cancel(skb, slave_data);
562 static int rtnl_link_info_fill(struct sk_buff *skb,
563 const struct net_device *dev)
565 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
571 if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
573 if (ops->fill_xstats) {
574 err = ops->fill_xstats(skb, dev);
578 if (ops->fill_info) {
579 data = nla_nest_start(skb, IFLA_INFO_DATA);
582 err = ops->fill_info(skb, dev);
584 goto err_cancel_data;
585 nla_nest_end(skb, data);
590 nla_nest_cancel(skb, data);
594 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
596 struct nlattr *linkinfo;
599 linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
600 if (linkinfo == NULL)
603 err = rtnl_link_info_fill(skb, dev);
605 goto err_cancel_link;
607 err = rtnl_link_slave_info_fill(skb, dev);
609 goto err_cancel_link;
611 nla_nest_end(skb, linkinfo);
615 nla_nest_cancel(skb, linkinfo);
620 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
622 struct sock *rtnl = net->rtnl;
625 NETLINK_CB(skb).dst_group = group;
627 atomic_inc(&skb->users);
628 netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
630 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
634 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
636 struct sock *rtnl = net->rtnl;
638 return nlmsg_unicast(rtnl, skb, pid);
640 EXPORT_SYMBOL(rtnl_unicast);
642 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
643 struct nlmsghdr *nlh, gfp_t flags)
645 struct sock *rtnl = net->rtnl;
649 report = nlmsg_report(nlh);
651 nlmsg_notify(rtnl, skb, pid, group, report, flags);
653 EXPORT_SYMBOL(rtnl_notify);
655 void rtnl_set_sk_err(struct net *net, u32 group, int error)
657 struct sock *rtnl = net->rtnl;
659 netlink_set_err(rtnl, 0, group, error);
661 EXPORT_SYMBOL(rtnl_set_sk_err);
663 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
668 mx = nla_nest_start(skb, RTA_METRICS);
672 for (i = 0; i < RTAX_MAX; i++) {
674 if (i == RTAX_CC_ALGO - 1) {
675 char tmp[TCP_CA_NAME_MAX], *name;
677 name = tcp_ca_get_name_by_key(metrics[i], tmp);
680 if (nla_put_string(skb, i + 1, name))
681 goto nla_put_failure;
682 } else if (i == RTAX_FEATURES - 1) {
683 u32 user_features = metrics[i] & RTAX_FEATURE_MASK;
685 BUILD_BUG_ON(RTAX_FEATURE_MASK & DST_FEATURE_MASK);
686 if (nla_put_u32(skb, i + 1, user_features))
687 goto nla_put_failure;
689 if (nla_put_u32(skb, i + 1, metrics[i]))
690 goto nla_put_failure;
697 nla_nest_cancel(skb, mx);
701 return nla_nest_end(skb, mx);
704 nla_nest_cancel(skb, mx);
707 EXPORT_SYMBOL(rtnetlink_put_metrics);
709 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
710 long expires, u32 error)
712 struct rta_cacheinfo ci = {
713 .rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse),
714 .rta_used = dst->__use,
715 .rta_clntref = atomic_read(&(dst->__refcnt)),
723 clock = jiffies_to_clock_t(abs(expires));
724 clock = min_t(unsigned long, clock, INT_MAX);
725 ci.rta_expires = (expires > 0) ? clock : -clock;
727 return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
729 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
731 static void set_operstate(struct net_device *dev, unsigned char transition)
733 unsigned char operstate = dev->operstate;
735 switch (transition) {
737 if ((operstate == IF_OPER_DORMANT ||
738 operstate == IF_OPER_UNKNOWN) &&
740 operstate = IF_OPER_UP;
743 case IF_OPER_DORMANT:
744 if (operstate == IF_OPER_UP ||
745 operstate == IF_OPER_UNKNOWN)
746 operstate = IF_OPER_DORMANT;
750 if (dev->operstate != operstate) {
751 write_lock_bh(&dev_base_lock);
752 dev->operstate = operstate;
753 write_unlock_bh(&dev_base_lock);
754 netdev_state_change(dev);
758 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
760 return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
761 (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
764 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
765 const struct ifinfomsg *ifm)
767 unsigned int flags = ifm->ifi_flags;
769 /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
771 flags = (flags & ifm->ifi_change) |
772 (rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
777 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
778 const struct rtnl_link_stats64 *b)
780 a->rx_packets = b->rx_packets;
781 a->tx_packets = b->tx_packets;
782 a->rx_bytes = b->rx_bytes;
783 a->tx_bytes = b->tx_bytes;
784 a->rx_errors = b->rx_errors;
785 a->tx_errors = b->tx_errors;
786 a->rx_dropped = b->rx_dropped;
787 a->tx_dropped = b->tx_dropped;
789 a->multicast = b->multicast;
790 a->collisions = b->collisions;
792 a->rx_length_errors = b->rx_length_errors;
793 a->rx_over_errors = b->rx_over_errors;
794 a->rx_crc_errors = b->rx_crc_errors;
795 a->rx_frame_errors = b->rx_frame_errors;
796 a->rx_fifo_errors = b->rx_fifo_errors;
797 a->rx_missed_errors = b->rx_missed_errors;
799 a->tx_aborted_errors = b->tx_aborted_errors;
800 a->tx_carrier_errors = b->tx_carrier_errors;
801 a->tx_fifo_errors = b->tx_fifo_errors;
802 a->tx_heartbeat_errors = b->tx_heartbeat_errors;
803 a->tx_window_errors = b->tx_window_errors;
805 a->rx_compressed = b->rx_compressed;
806 a->tx_compressed = b->tx_compressed;
809 static void copy_rtnl_link_stats64(void *v, const struct rtnl_link_stats64 *b)
811 memcpy(v, b, sizeof(*b));
815 static inline int rtnl_vfinfo_size(const struct net_device *dev,
818 if (dev->dev.parent && dev_is_pci(dev->dev.parent) &&
819 (ext_filter_mask & RTEXT_FILTER_VF)) {
820 int num_vfs = dev_num_vf(dev->dev.parent);
821 size_t size = nla_total_size(sizeof(struct nlattr));
822 size += nla_total_size(num_vfs * sizeof(struct nlattr));
824 (nla_total_size(sizeof(struct ifla_vf_mac)) +
825 nla_total_size(sizeof(struct ifla_vf_vlan)) +
826 nla_total_size(sizeof(struct ifla_vf_spoofchk)) +
827 nla_total_size(sizeof(struct ifla_vf_rate)) +
828 nla_total_size(sizeof(struct ifla_vf_link_state)) +
829 nla_total_size(sizeof(struct ifla_vf_rss_query_en)) +
830 /* IFLA_VF_STATS_RX_PACKETS */
831 nla_total_size(sizeof(__u64)) +
832 /* IFLA_VF_STATS_TX_PACKETS */
833 nla_total_size(sizeof(__u64)) +
834 /* IFLA_VF_STATS_RX_BYTES */
835 nla_total_size(sizeof(__u64)) +
836 /* IFLA_VF_STATS_TX_BYTES */
837 nla_total_size(sizeof(__u64)) +
838 /* IFLA_VF_STATS_BROADCAST */
839 nla_total_size(sizeof(__u64)) +
840 /* IFLA_VF_STATS_MULTICAST */
841 nla_total_size(sizeof(__u64)) +
842 nla_total_size(sizeof(struct ifla_vf_trust)));
848 static size_t rtnl_port_size(const struct net_device *dev,
851 size_t port_size = nla_total_size(4) /* PORT_VF */
852 + nla_total_size(PORT_PROFILE_MAX) /* PORT_PROFILE */
853 + nla_total_size(sizeof(struct ifla_port_vsi))
855 + nla_total_size(PORT_UUID_MAX) /* PORT_INSTANCE_UUID */
856 + nla_total_size(PORT_UUID_MAX) /* PORT_HOST_UUID */
857 + nla_total_size(1) /* PROT_VDP_REQUEST */
858 + nla_total_size(2); /* PORT_VDP_RESPONSE */
859 size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
860 size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
862 size_t port_self_size = nla_total_size(sizeof(struct nlattr))
865 if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
866 !(ext_filter_mask & RTEXT_FILTER_VF))
868 if (dev_num_vf(dev->dev.parent))
869 return port_self_size + vf_ports_size +
870 vf_port_size * dev_num_vf(dev->dev.parent);
872 return port_self_size;
875 static noinline size_t if_nlmsg_size(const struct net_device *dev,
878 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
879 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
880 + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
881 + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
882 + nla_total_size(sizeof(struct rtnl_link_ifmap))
883 + nla_total_size(sizeof(struct rtnl_link_stats))
884 + nla_total_size(sizeof(struct rtnl_link_stats64))
885 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
886 + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
887 + nla_total_size(4) /* IFLA_TXQLEN */
888 + nla_total_size(4) /* IFLA_WEIGHT */
889 + nla_total_size(4) /* IFLA_MTU */
890 + nla_total_size(4) /* IFLA_LINK */
891 + nla_total_size(4) /* IFLA_MASTER */
892 + nla_total_size(1) /* IFLA_CARRIER */
893 + nla_total_size(4) /* IFLA_PROMISCUITY */
894 + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
895 + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
896 + nla_total_size(1) /* IFLA_OPERSTATE */
897 + nla_total_size(1) /* IFLA_LINKMODE */
898 + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
899 + nla_total_size(4) /* IFLA_LINK_NETNSID */
900 + nla_total_size(ext_filter_mask
901 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
902 + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
903 + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
904 + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
905 + rtnl_link_get_af_size(dev, ext_filter_mask) /* IFLA_AF_SPEC */
906 + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */
907 + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_SWITCH_ID */
908 + nla_total_size(IFNAMSIZ) /* IFLA_PHYS_PORT_NAME */
909 + nla_total_size(1); /* IFLA_PROTO_DOWN */
913 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
915 struct nlattr *vf_ports;
916 struct nlattr *vf_port;
920 vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
924 for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
925 vf_port = nla_nest_start(skb, IFLA_VF_PORT);
927 goto nla_put_failure;
928 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
929 goto nla_put_failure;
930 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
931 if (err == -EMSGSIZE)
932 goto nla_put_failure;
934 nla_nest_cancel(skb, vf_port);
937 nla_nest_end(skb, vf_port);
940 nla_nest_end(skb, vf_ports);
945 nla_nest_cancel(skb, vf_ports);
949 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
951 struct nlattr *port_self;
954 port_self = nla_nest_start(skb, IFLA_PORT_SELF);
958 err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
960 nla_nest_cancel(skb, port_self);
961 return (err == -EMSGSIZE) ? err : 0;
964 nla_nest_end(skb, port_self);
969 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
974 if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
975 !(ext_filter_mask & RTEXT_FILTER_VF))
978 err = rtnl_port_self_fill(skb, dev);
982 if (dev_num_vf(dev->dev.parent)) {
983 err = rtnl_vf_ports_fill(skb, dev);
991 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
994 struct netdev_phys_item_id ppid;
996 err = dev_get_phys_port_id(dev, &ppid);
998 if (err == -EOPNOTSUPP)
1003 if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
1009 static int rtnl_phys_port_name_fill(struct sk_buff *skb, struct net_device *dev)
1011 char name[IFNAMSIZ];
1014 err = dev_get_phys_port_name(dev, name, sizeof(name));
1016 if (err == -EOPNOTSUPP)
1021 if (nla_put(skb, IFLA_PHYS_PORT_NAME, strlen(name), name))
1027 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev)
1030 struct switchdev_attr attr = {
1031 .id = SWITCHDEV_ATTR_ID_PORT_PARENT_ID,
1032 .flags = SWITCHDEV_F_NO_RECURSE,
1035 err = switchdev_port_attr_get(dev, &attr);
1037 if (err == -EOPNOTSUPP)
1042 if (nla_put(skb, IFLA_PHYS_SWITCH_ID, attr.u.ppid.id_len,
1049 static noinline_for_stack int rtnl_fill_stats(struct sk_buff *skb,
1050 struct net_device *dev)
1052 const struct rtnl_link_stats64 *stats;
1053 struct rtnl_link_stats64 temp;
1054 struct nlattr *attr;
1056 stats = dev_get_stats(dev, &temp);
1058 attr = nla_reserve(skb, IFLA_STATS,
1059 sizeof(struct rtnl_link_stats));
1063 copy_rtnl_link_stats(nla_data(attr), stats);
1065 attr = nla_reserve(skb, IFLA_STATS64,
1066 sizeof(struct rtnl_link_stats64));
1070 copy_rtnl_link_stats64(nla_data(attr), stats);
1075 static noinline_for_stack int rtnl_fill_vfinfo(struct sk_buff *skb,
1076 struct net_device *dev,
1078 struct nlattr *vfinfo)
1080 struct ifla_vf_rss_query_en vf_rss_query_en;
1081 struct ifla_vf_link_state vf_linkstate;
1082 struct ifla_vf_spoofchk vf_spoofchk;
1083 struct ifla_vf_tx_rate vf_tx_rate;
1084 struct ifla_vf_stats vf_stats;
1085 struct ifla_vf_trust vf_trust;
1086 struct ifla_vf_vlan vf_vlan;
1087 struct ifla_vf_rate vf_rate;
1088 struct nlattr *vf, *vfstats;
1089 struct ifla_vf_mac vf_mac;
1090 struct ifla_vf_info ivi;
1092 /* Not all SR-IOV capable drivers support the
1093 * spoofcheck and "RSS query enable" query. Preset to
1094 * -1 so the user space tool can detect that the driver
1095 * didn't report anything.
1098 ivi.rss_query_en = -1;
1100 memset(ivi.mac, 0, sizeof(ivi.mac));
1101 /* The default value for VF link state is "auto"
1102 * IFLA_VF_LINK_STATE_AUTO which equals zero
1105 if (dev->netdev_ops->ndo_get_vf_config(dev, vfs_num, &ivi))
1114 vf_rss_query_en.vf =
1115 vf_trust.vf = ivi.vf;
1117 memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1118 vf_vlan.vlan = ivi.vlan;
1119 vf_vlan.qos = ivi.qos;
1120 vf_tx_rate.rate = ivi.max_tx_rate;
1121 vf_rate.min_tx_rate = ivi.min_tx_rate;
1122 vf_rate.max_tx_rate = ivi.max_tx_rate;
1123 vf_spoofchk.setting = ivi.spoofchk;
1124 vf_linkstate.link_state = ivi.linkstate;
1125 vf_rss_query_en.setting = ivi.rss_query_en;
1126 vf_trust.setting = ivi.trusted;
1127 vf = nla_nest_start(skb, IFLA_VF_INFO);
1129 nla_nest_cancel(skb, vfinfo);
1132 if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1133 nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1134 nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1136 nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1138 nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1140 nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1142 nla_put(skb, IFLA_VF_RSS_QUERY_EN,
1143 sizeof(vf_rss_query_en),
1144 &vf_rss_query_en) ||
1145 nla_put(skb, IFLA_VF_TRUST,
1146 sizeof(vf_trust), &vf_trust))
1148 memset(&vf_stats, 0, sizeof(vf_stats));
1149 if (dev->netdev_ops->ndo_get_vf_stats)
1150 dev->netdev_ops->ndo_get_vf_stats(dev, vfs_num,
1152 vfstats = nla_nest_start(skb, IFLA_VF_STATS);
1154 nla_nest_cancel(skb, vf);
1155 nla_nest_cancel(skb, vfinfo);
1158 if (nla_put_u64(skb, IFLA_VF_STATS_RX_PACKETS,
1159 vf_stats.rx_packets) ||
1160 nla_put_u64(skb, IFLA_VF_STATS_TX_PACKETS,
1161 vf_stats.tx_packets) ||
1162 nla_put_u64(skb, IFLA_VF_STATS_RX_BYTES,
1163 vf_stats.rx_bytes) ||
1164 nla_put_u64(skb, IFLA_VF_STATS_TX_BYTES,
1165 vf_stats.tx_bytes) ||
1166 nla_put_u64(skb, IFLA_VF_STATS_BROADCAST,
1167 vf_stats.broadcast) ||
1168 nla_put_u64(skb, IFLA_VF_STATS_MULTICAST,
1169 vf_stats.multicast))
1171 nla_nest_end(skb, vfstats);
1172 nla_nest_end(skb, vf);
1176 static int rtnl_fill_link_ifmap(struct sk_buff *skb, struct net_device *dev)
1178 struct rtnl_link_ifmap map;
1180 memset(&map, 0, sizeof(map));
1181 map.mem_start = dev->mem_start;
1182 map.mem_end = dev->mem_end;
1183 map.base_addr = dev->base_addr;
1186 map.port = dev->if_port;
1188 if (nla_put(skb, IFLA_MAP, sizeof(map), &map))
1194 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
1195 int type, u32 pid, u32 seq, u32 change,
1196 unsigned int flags, u32 ext_filter_mask)
1198 struct ifinfomsg *ifm;
1199 struct nlmsghdr *nlh;
1200 struct nlattr *af_spec;
1201 struct rtnl_af_ops *af_ops;
1202 struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1205 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
1209 ifm = nlmsg_data(nlh);
1210 ifm->ifi_family = AF_UNSPEC;
1212 ifm->ifi_type = dev->type;
1213 ifm->ifi_index = dev->ifindex;
1214 ifm->ifi_flags = dev_get_flags(dev);
1215 ifm->ifi_change = change;
1217 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
1218 nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
1219 nla_put_u8(skb, IFLA_OPERSTATE,
1220 netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
1221 nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
1222 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
1223 nla_put_u32(skb, IFLA_GROUP, dev->group) ||
1224 nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
1225 nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
1227 nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1229 (dev->ifindex != dev_get_iflink(dev) &&
1230 nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
1232 nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex)) ||
1233 nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1235 nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
1237 nla_put_string(skb, IFLA_IFALIAS, dev->ifalias)) ||
1238 nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1239 atomic_read(&dev->carrier_changes)) ||
1240 nla_put_u8(skb, IFLA_PROTO_DOWN, dev->proto_down))
1241 goto nla_put_failure;
1243 if (rtnl_fill_link_ifmap(skb, dev))
1244 goto nla_put_failure;
1246 if (dev->addr_len) {
1247 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1248 nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1249 goto nla_put_failure;
1252 if (rtnl_phys_port_id_fill(skb, dev))
1253 goto nla_put_failure;
1255 if (rtnl_phys_port_name_fill(skb, dev))
1256 goto nla_put_failure;
1258 if (rtnl_phys_switch_id_fill(skb, dev))
1259 goto nla_put_failure;
1261 if (rtnl_fill_stats(skb, dev))
1262 goto nla_put_failure;
1264 if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF) &&
1265 nla_put_u32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent)))
1266 goto nla_put_failure;
1268 if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent &&
1269 ext_filter_mask & RTEXT_FILTER_VF) {
1271 struct nlattr *vfinfo;
1272 int num_vfs = dev_num_vf(dev->dev.parent);
1274 vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
1276 goto nla_put_failure;
1277 for (i = 0; i < num_vfs; i++) {
1278 if (rtnl_fill_vfinfo(skb, dev, i, vfinfo))
1279 goto nla_put_failure;
1282 nla_nest_end(skb, vfinfo);
1285 if (rtnl_port_fill(skb, dev, ext_filter_mask))
1286 goto nla_put_failure;
1288 if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1289 if (rtnl_link_fill(skb, dev) < 0)
1290 goto nla_put_failure;
1293 if (dev->rtnl_link_ops &&
1294 dev->rtnl_link_ops->get_link_net) {
1295 struct net *link_net = dev->rtnl_link_ops->get_link_net(dev);
1297 if (!net_eq(dev_net(dev), link_net)) {
1298 int id = peernet2id_alloc(dev_net(dev), link_net);
1300 if (nla_put_s32(skb, IFLA_LINK_NETNSID, id))
1301 goto nla_put_failure;
1305 if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1306 goto nla_put_failure;
1308 list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1309 if (af_ops->fill_link_af) {
1313 if (!(af = nla_nest_start(skb, af_ops->family)))
1314 goto nla_put_failure;
1316 err = af_ops->fill_link_af(skb, dev, ext_filter_mask);
1319 * Caller may return ENODATA to indicate that there
1320 * was no data to be dumped. This is not an error, it
1321 * means we should trim the attribute header and
1324 if (err == -ENODATA)
1325 nla_nest_cancel(skb, af);
1327 goto nla_put_failure;
1329 nla_nest_end(skb, af);
1333 nla_nest_end(skb, af_spec);
1335 nlmsg_end(skb, nlh);
1339 nlmsg_cancel(skb, nlh);
1343 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1344 [IFLA_IFNAME] = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1345 [IFLA_ADDRESS] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1346 [IFLA_BROADCAST] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1347 [IFLA_MAP] = { .len = sizeof(struct rtnl_link_ifmap) },
1348 [IFLA_MTU] = { .type = NLA_U32 },
1349 [IFLA_LINK] = { .type = NLA_U32 },
1350 [IFLA_MASTER] = { .type = NLA_U32 },
1351 [IFLA_CARRIER] = { .type = NLA_U8 },
1352 [IFLA_TXQLEN] = { .type = NLA_U32 },
1353 [IFLA_WEIGHT] = { .type = NLA_U32 },
1354 [IFLA_OPERSTATE] = { .type = NLA_U8 },
1355 [IFLA_LINKMODE] = { .type = NLA_U8 },
1356 [IFLA_LINKINFO] = { .type = NLA_NESTED },
1357 [IFLA_NET_NS_PID] = { .type = NLA_U32 },
1358 [IFLA_NET_NS_FD] = { .type = NLA_U32 },
1359 [IFLA_IFALIAS] = { .type = NLA_STRING, .len = IFALIASZ-1 },
1360 [IFLA_VFINFO_LIST] = {. type = NLA_NESTED },
1361 [IFLA_VF_PORTS] = { .type = NLA_NESTED },
1362 [IFLA_PORT_SELF] = { .type = NLA_NESTED },
1363 [IFLA_AF_SPEC] = { .type = NLA_NESTED },
1364 [IFLA_EXT_MASK] = { .type = NLA_U32 },
1365 [IFLA_PROMISCUITY] = { .type = NLA_U32 },
1366 [IFLA_NUM_TX_QUEUES] = { .type = NLA_U32 },
1367 [IFLA_NUM_RX_QUEUES] = { .type = NLA_U32 },
1368 [IFLA_PHYS_PORT_ID] = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1369 [IFLA_CARRIER_CHANGES] = { .type = NLA_U32 }, /* ignored */
1370 [IFLA_PHYS_SWITCH_ID] = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1371 [IFLA_LINK_NETNSID] = { .type = NLA_S32 },
1372 [IFLA_PROTO_DOWN] = { .type = NLA_U8 },
1375 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1376 [IFLA_INFO_KIND] = { .type = NLA_STRING },
1377 [IFLA_INFO_DATA] = { .type = NLA_NESTED },
1378 [IFLA_INFO_SLAVE_KIND] = { .type = NLA_STRING },
1379 [IFLA_INFO_SLAVE_DATA] = { .type = NLA_NESTED },
1382 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1383 [IFLA_VF_MAC] = { .len = sizeof(struct ifla_vf_mac) },
1384 [IFLA_VF_VLAN] = { .len = sizeof(struct ifla_vf_vlan) },
1385 [IFLA_VF_TX_RATE] = { .len = sizeof(struct ifla_vf_tx_rate) },
1386 [IFLA_VF_SPOOFCHK] = { .len = sizeof(struct ifla_vf_spoofchk) },
1387 [IFLA_VF_RATE] = { .len = sizeof(struct ifla_vf_rate) },
1388 [IFLA_VF_LINK_STATE] = { .len = sizeof(struct ifla_vf_link_state) },
1389 [IFLA_VF_RSS_QUERY_EN] = { .len = sizeof(struct ifla_vf_rss_query_en) },
1390 [IFLA_VF_STATS] = { .type = NLA_NESTED },
1391 [IFLA_VF_TRUST] = { .len = sizeof(struct ifla_vf_trust) },
1394 static const struct nla_policy ifla_vf_stats_policy[IFLA_VF_STATS_MAX + 1] = {
1395 [IFLA_VF_STATS_RX_PACKETS] = { .type = NLA_U64 },
1396 [IFLA_VF_STATS_TX_PACKETS] = { .type = NLA_U64 },
1397 [IFLA_VF_STATS_RX_BYTES] = { .type = NLA_U64 },
1398 [IFLA_VF_STATS_TX_BYTES] = { .type = NLA_U64 },
1399 [IFLA_VF_STATS_BROADCAST] = { .type = NLA_U64 },
1400 [IFLA_VF_STATS_MULTICAST] = { .type = NLA_U64 },
1403 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1404 [IFLA_PORT_VF] = { .type = NLA_U32 },
1405 [IFLA_PORT_PROFILE] = { .type = NLA_STRING,
1406 .len = PORT_PROFILE_MAX },
1407 [IFLA_PORT_VSI_TYPE] = { .type = NLA_BINARY,
1408 .len = sizeof(struct ifla_port_vsi)},
1409 [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1410 .len = PORT_UUID_MAX },
1411 [IFLA_PORT_HOST_UUID] = { .type = NLA_STRING,
1412 .len = PORT_UUID_MAX },
1413 [IFLA_PORT_REQUEST] = { .type = NLA_U8, },
1414 [IFLA_PORT_RESPONSE] = { .type = NLA_U16, },
1417 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1419 struct net *net = sock_net(skb->sk);
1422 struct net_device *dev;
1423 struct hlist_head *head;
1424 struct nlattr *tb[IFLA_MAX+1];
1425 u32 ext_filter_mask = 0;
1430 s_idx = cb->args[1];
1432 cb->seq = net->dev_base_seq;
1434 /* A hack to preserve kernel<->userspace interface.
1435 * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
1436 * However, before Linux v3.9 the code here assumed rtgenmsg and that's
1437 * what iproute2 < v3.9.0 used.
1438 * We can detect the old iproute2. Even including the IFLA_EXT_MASK
1439 * attribute, its netlink message is shorter than struct ifinfomsg.
1441 hdrlen = nlmsg_len(cb->nlh) < sizeof(struct ifinfomsg) ?
1442 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
1444 if (nlmsg_parse(cb->nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
1446 if (tb[IFLA_EXT_MASK])
1447 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1450 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1452 head = &net->dev_index_head[h];
1453 hlist_for_each_entry(dev, head, index_hlist) {
1456 err = rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1457 NETLINK_CB(cb->skb).portid,
1458 cb->nlh->nlmsg_seq, 0,
1461 /* If we ran out of room on the first message,
1464 WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
1469 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1481 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len)
1483 return nla_parse(tb, IFLA_MAX, head, len, ifla_policy);
1485 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
1487 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1490 /* Examine the link attributes and figure out which
1491 * network namespace we are talking about.
1493 if (tb[IFLA_NET_NS_PID])
1494 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1495 else if (tb[IFLA_NET_NS_FD])
1496 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1498 net = get_net(src_net);
1501 EXPORT_SYMBOL(rtnl_link_get_net);
1503 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1506 if (tb[IFLA_ADDRESS] &&
1507 nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1510 if (tb[IFLA_BROADCAST] &&
1511 nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1515 if (tb[IFLA_AF_SPEC]) {
1519 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1520 const struct rtnl_af_ops *af_ops;
1522 if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1523 return -EAFNOSUPPORT;
1525 if (!af_ops->set_link_af)
1528 if (af_ops->validate_link_af) {
1529 err = af_ops->validate_link_af(dev, af);
1539 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb)
1541 const struct net_device_ops *ops = dev->netdev_ops;
1544 if (tb[IFLA_VF_MAC]) {
1545 struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]);
1548 if (ops->ndo_set_vf_mac)
1549 err = ops->ndo_set_vf_mac(dev, ivm->vf,
1555 if (tb[IFLA_VF_VLAN]) {
1556 struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]);
1559 if (ops->ndo_set_vf_vlan)
1560 err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan,
1566 if (tb[IFLA_VF_TX_RATE]) {
1567 struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]);
1568 struct ifla_vf_info ivf;
1571 if (ops->ndo_get_vf_config)
1572 err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf);
1577 if (ops->ndo_set_vf_rate)
1578 err = ops->ndo_set_vf_rate(dev, ivt->vf,
1585 if (tb[IFLA_VF_RATE]) {
1586 struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]);
1589 if (ops->ndo_set_vf_rate)
1590 err = ops->ndo_set_vf_rate(dev, ivt->vf,
1597 if (tb[IFLA_VF_SPOOFCHK]) {
1598 struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]);
1601 if (ops->ndo_set_vf_spoofchk)
1602 err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1608 if (tb[IFLA_VF_LINK_STATE]) {
1609 struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]);
1612 if (ops->ndo_set_vf_link_state)
1613 err = ops->ndo_set_vf_link_state(dev, ivl->vf,
1619 if (tb[IFLA_VF_RSS_QUERY_EN]) {
1620 struct ifla_vf_rss_query_en *ivrssq_en;
1623 ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]);
1624 if (ops->ndo_set_vf_rss_query_en)
1625 err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf,
1626 ivrssq_en->setting);
1631 if (tb[IFLA_VF_TRUST]) {
1632 struct ifla_vf_trust *ivt = nla_data(tb[IFLA_VF_TRUST]);
1635 if (ops->ndo_set_vf_trust)
1636 err = ops->ndo_set_vf_trust(dev, ivt->vf, ivt->setting);
1644 static int do_set_master(struct net_device *dev, int ifindex)
1646 struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1647 const struct net_device_ops *ops;
1651 if (upper_dev->ifindex == ifindex)
1653 ops = upper_dev->netdev_ops;
1654 if (ops->ndo_del_slave) {
1655 err = ops->ndo_del_slave(upper_dev, dev);
1664 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
1667 ops = upper_dev->netdev_ops;
1668 if (ops->ndo_add_slave) {
1669 err = ops->ndo_add_slave(upper_dev, dev);
1679 #define DO_SETLINK_MODIFIED 0x01
1680 /* notify flag means notify + modified. */
1681 #define DO_SETLINK_NOTIFY 0x03
1682 static int do_setlink(const struct sk_buff *skb,
1683 struct net_device *dev, struct ifinfomsg *ifm,
1684 struct nlattr **tb, char *ifname, int status)
1686 const struct net_device_ops *ops = dev->netdev_ops;
1689 if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1690 struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1695 if (!netlink_ns_capable(skb, net->user_ns, CAP_NET_ADMIN)) {
1700 err = dev_change_net_namespace(dev, net, ifname);
1704 status |= DO_SETLINK_MODIFIED;
1708 struct rtnl_link_ifmap *u_map;
1711 if (!ops->ndo_set_config) {
1716 if (!netif_device_present(dev)) {
1721 u_map = nla_data(tb[IFLA_MAP]);
1722 k_map.mem_start = (unsigned long) u_map->mem_start;
1723 k_map.mem_end = (unsigned long) u_map->mem_end;
1724 k_map.base_addr = (unsigned short) u_map->base_addr;
1725 k_map.irq = (unsigned char) u_map->irq;
1726 k_map.dma = (unsigned char) u_map->dma;
1727 k_map.port = (unsigned char) u_map->port;
1729 err = ops->ndo_set_config(dev, &k_map);
1733 status |= DO_SETLINK_NOTIFY;
1736 if (tb[IFLA_ADDRESS]) {
1737 struct sockaddr *sa;
1740 len = sizeof(sa_family_t) + dev->addr_len;
1741 sa = kmalloc(len, GFP_KERNEL);
1746 sa->sa_family = dev->type;
1747 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1749 err = dev_set_mac_address(dev, sa);
1753 status |= DO_SETLINK_MODIFIED;
1757 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1760 status |= DO_SETLINK_MODIFIED;
1763 if (tb[IFLA_GROUP]) {
1764 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1765 status |= DO_SETLINK_NOTIFY;
1769 * Interface selected by interface index but interface
1770 * name provided implies that a name change has been
1773 if (ifm->ifi_index > 0 && ifname[0]) {
1774 err = dev_change_name(dev, ifname);
1777 status |= DO_SETLINK_MODIFIED;
1780 if (tb[IFLA_IFALIAS]) {
1781 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
1782 nla_len(tb[IFLA_IFALIAS]));
1785 status |= DO_SETLINK_NOTIFY;
1788 if (tb[IFLA_BROADCAST]) {
1789 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
1790 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
1793 if (ifm->ifi_flags || ifm->ifi_change) {
1794 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1799 if (tb[IFLA_MASTER]) {
1800 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
1803 status |= DO_SETLINK_MODIFIED;
1806 if (tb[IFLA_CARRIER]) {
1807 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
1810 status |= DO_SETLINK_MODIFIED;
1813 if (tb[IFLA_TXQLEN]) {
1814 unsigned long value = nla_get_u32(tb[IFLA_TXQLEN]);
1816 if (dev->tx_queue_len ^ value)
1817 status |= DO_SETLINK_NOTIFY;
1819 dev->tx_queue_len = value;
1822 if (tb[IFLA_OPERSTATE])
1823 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1825 if (tb[IFLA_LINKMODE]) {
1826 unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
1828 write_lock_bh(&dev_base_lock);
1829 if (dev->link_mode ^ value)
1830 status |= DO_SETLINK_NOTIFY;
1831 dev->link_mode = value;
1832 write_unlock_bh(&dev_base_lock);
1835 if (tb[IFLA_VFINFO_LIST]) {
1836 struct nlattr *vfinfo[IFLA_VF_MAX + 1];
1837 struct nlattr *attr;
1840 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
1841 if (nla_type(attr) != IFLA_VF_INFO ||
1842 nla_len(attr) < NLA_HDRLEN) {
1846 err = nla_parse_nested(vfinfo, IFLA_VF_MAX, attr,
1850 err = do_setvfinfo(dev, vfinfo);
1853 status |= DO_SETLINK_NOTIFY;
1858 if (tb[IFLA_VF_PORTS]) {
1859 struct nlattr *port[IFLA_PORT_MAX+1];
1860 struct nlattr *attr;
1865 if (!ops->ndo_set_vf_port)
1868 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
1869 if (nla_type(attr) != IFLA_VF_PORT ||
1870 nla_len(attr) < NLA_HDRLEN) {
1874 err = nla_parse_nested(port, IFLA_PORT_MAX, attr,
1878 if (!port[IFLA_PORT_VF]) {
1882 vf = nla_get_u32(port[IFLA_PORT_VF]);
1883 err = ops->ndo_set_vf_port(dev, vf, port);
1886 status |= DO_SETLINK_NOTIFY;
1891 if (tb[IFLA_PORT_SELF]) {
1892 struct nlattr *port[IFLA_PORT_MAX+1];
1894 err = nla_parse_nested(port, IFLA_PORT_MAX,
1895 tb[IFLA_PORT_SELF], ifla_port_policy);
1900 if (ops->ndo_set_vf_port)
1901 err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
1904 status |= DO_SETLINK_NOTIFY;
1907 if (tb[IFLA_AF_SPEC]) {
1911 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1912 const struct rtnl_af_ops *af_ops;
1914 if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1917 err = af_ops->set_link_af(dev, af);
1921 status |= DO_SETLINK_NOTIFY;
1926 if (tb[IFLA_PROTO_DOWN]) {
1927 err = dev_change_proto_down(dev,
1928 nla_get_u8(tb[IFLA_PROTO_DOWN]));
1931 status |= DO_SETLINK_NOTIFY;
1935 if (status & DO_SETLINK_MODIFIED) {
1936 if (status & DO_SETLINK_NOTIFY)
1937 netdev_state_change(dev);
1940 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",
1947 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
1949 struct net *net = sock_net(skb->sk);
1950 struct ifinfomsg *ifm;
1951 struct net_device *dev;
1953 struct nlattr *tb[IFLA_MAX+1];
1954 char ifname[IFNAMSIZ];
1956 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1960 if (tb[IFLA_IFNAME])
1961 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1966 ifm = nlmsg_data(nlh);
1967 if (ifm->ifi_index > 0)
1968 dev = __dev_get_by_index(net, ifm->ifi_index);
1969 else if (tb[IFLA_IFNAME])
1970 dev = __dev_get_by_name(net, ifname);
1979 err = validate_linkmsg(dev, tb);
1983 err = do_setlink(skb, dev, ifm, tb, ifname, 0);
1988 static int rtnl_group_dellink(const struct net *net, int group)
1990 struct net_device *dev, *aux;
1991 LIST_HEAD(list_kill);
1997 for_each_netdev(net, dev) {
1998 if (dev->group == group) {
1999 const struct rtnl_link_ops *ops;
2002 ops = dev->rtnl_link_ops;
2003 if (!ops || !ops->dellink)
2011 for_each_netdev_safe(net, dev, aux) {
2012 if (dev->group == group) {
2013 const struct rtnl_link_ops *ops;
2015 ops = dev->rtnl_link_ops;
2016 ops->dellink(dev, &list_kill);
2019 unregister_netdevice_many(&list_kill);
2024 int rtnl_delete_link(struct net_device *dev)
2026 const struct rtnl_link_ops *ops;
2027 LIST_HEAD(list_kill);
2029 ops = dev->rtnl_link_ops;
2030 if (!ops || !ops->dellink)
2033 ops->dellink(dev, &list_kill);
2034 unregister_netdevice_many(&list_kill);
2038 EXPORT_SYMBOL_GPL(rtnl_delete_link);
2040 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
2042 struct net *net = sock_net(skb->sk);
2043 struct net_device *dev;
2044 struct ifinfomsg *ifm;
2045 char ifname[IFNAMSIZ];
2046 struct nlattr *tb[IFLA_MAX+1];
2049 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2053 if (tb[IFLA_IFNAME])
2054 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2056 ifm = nlmsg_data(nlh);
2057 if (ifm->ifi_index > 0)
2058 dev = __dev_get_by_index(net, ifm->ifi_index);
2059 else if (tb[IFLA_IFNAME])
2060 dev = __dev_get_by_name(net, ifname);
2061 else if (tb[IFLA_GROUP])
2062 return rtnl_group_dellink(net, nla_get_u32(tb[IFLA_GROUP]));
2069 return rtnl_delete_link(dev);
2072 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
2074 unsigned int old_flags;
2077 old_flags = dev->flags;
2078 if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
2079 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
2084 dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
2086 __dev_notify_flags(dev, old_flags, ~0U);
2089 EXPORT_SYMBOL(rtnl_configure_link);
2091 struct net_device *rtnl_create_link(struct net *net,
2092 const char *ifname, unsigned char name_assign_type,
2093 const struct rtnl_link_ops *ops, struct nlattr *tb[])
2096 struct net_device *dev;
2097 unsigned int num_tx_queues = 1;
2098 unsigned int num_rx_queues = 1;
2100 if (tb[IFLA_NUM_TX_QUEUES])
2101 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
2102 else if (ops->get_num_tx_queues)
2103 num_tx_queues = ops->get_num_tx_queues();
2105 if (tb[IFLA_NUM_RX_QUEUES])
2106 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
2107 else if (ops->get_num_rx_queues)
2108 num_rx_queues = ops->get_num_rx_queues();
2111 dev = alloc_netdev_mqs(ops->priv_size, ifname, name_assign_type,
2112 ops->setup, num_tx_queues, num_rx_queues);
2116 dev_net_set(dev, net);
2117 dev->rtnl_link_ops = ops;
2118 dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
2121 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
2122 if (tb[IFLA_ADDRESS]) {
2123 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
2124 nla_len(tb[IFLA_ADDRESS]));
2125 dev->addr_assign_type = NET_ADDR_SET;
2127 if (tb[IFLA_BROADCAST])
2128 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
2129 nla_len(tb[IFLA_BROADCAST]));
2130 if (tb[IFLA_TXQLEN])
2131 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
2132 if (tb[IFLA_OPERSTATE])
2133 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2134 if (tb[IFLA_LINKMODE])
2135 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
2137 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2142 return ERR_PTR(err);
2144 EXPORT_SYMBOL(rtnl_create_link);
2146 static int rtnl_group_changelink(const struct sk_buff *skb,
2147 struct net *net, int group,
2148 struct ifinfomsg *ifm,
2151 struct net_device *dev, *aux;
2154 for_each_netdev_safe(net, dev, aux) {
2155 if (dev->group == group) {
2156 err = do_setlink(skb, dev, ifm, tb, NULL, 0);
2165 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh)
2167 struct net *net = sock_net(skb->sk);
2168 const struct rtnl_link_ops *ops;
2169 const struct rtnl_link_ops *m_ops = NULL;
2170 struct net_device *dev;
2171 struct net_device *master_dev = NULL;
2172 struct ifinfomsg *ifm;
2173 char kind[MODULE_NAME_LEN];
2174 char ifname[IFNAMSIZ];
2175 struct nlattr *tb[IFLA_MAX+1];
2176 struct nlattr *linkinfo[IFLA_INFO_MAX+1];
2177 unsigned char name_assign_type = NET_NAME_USER;
2180 #ifdef CONFIG_MODULES
2183 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2187 if (tb[IFLA_IFNAME])
2188 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2192 ifm = nlmsg_data(nlh);
2193 if (ifm->ifi_index > 0)
2194 dev = __dev_get_by_index(net, ifm->ifi_index);
2197 dev = __dev_get_by_name(net, ifname);
2203 master_dev = netdev_master_upper_dev_get(dev);
2205 m_ops = master_dev->rtnl_link_ops;
2208 err = validate_linkmsg(dev, tb);
2212 if (tb[IFLA_LINKINFO]) {
2213 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
2214 tb[IFLA_LINKINFO], ifla_info_policy);
2218 memset(linkinfo, 0, sizeof(linkinfo));
2220 if (linkinfo[IFLA_INFO_KIND]) {
2221 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
2222 ops = rtnl_link_ops_get(kind);
2229 struct nlattr *attr[ops ? ops->maxtype + 1 : 1];
2230 struct nlattr *slave_attr[m_ops ? m_ops->slave_maxtype + 1 : 1];
2231 struct nlattr **data = NULL;
2232 struct nlattr **slave_data = NULL;
2233 struct net *dest_net, *link_net = NULL;
2236 if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
2237 err = nla_parse_nested(attr, ops->maxtype,
2238 linkinfo[IFLA_INFO_DATA],
2244 if (ops->validate) {
2245 err = ops->validate(tb, data);
2252 if (m_ops->slave_maxtype &&
2253 linkinfo[IFLA_INFO_SLAVE_DATA]) {
2254 err = nla_parse_nested(slave_attr,
2255 m_ops->slave_maxtype,
2256 linkinfo[IFLA_INFO_SLAVE_DATA],
2257 m_ops->slave_policy);
2260 slave_data = slave_attr;
2262 if (m_ops->slave_validate) {
2263 err = m_ops->slave_validate(tb, slave_data);
2272 if (nlh->nlmsg_flags & NLM_F_EXCL)
2274 if (nlh->nlmsg_flags & NLM_F_REPLACE)
2277 if (linkinfo[IFLA_INFO_DATA]) {
2278 if (!ops || ops != dev->rtnl_link_ops ||
2282 err = ops->changelink(dev, tb, data);
2285 status |= DO_SETLINK_NOTIFY;
2288 if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
2289 if (!m_ops || !m_ops->slave_changelink)
2292 err = m_ops->slave_changelink(master_dev, dev,
2296 status |= DO_SETLINK_NOTIFY;
2299 return do_setlink(skb, dev, ifm, tb, ifname, status);
2302 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
2303 if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
2304 return rtnl_group_changelink(skb, net,
2305 nla_get_u32(tb[IFLA_GROUP]),
2310 if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
2314 #ifdef CONFIG_MODULES
2317 request_module("rtnl-link-%s", kind);
2319 ops = rtnl_link_ops_get(kind);
2331 snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
2332 name_assign_type = NET_NAME_ENUM;
2335 dest_net = rtnl_link_get_net(net, tb);
2336 if (IS_ERR(dest_net))
2337 return PTR_ERR(dest_net);
2340 if (!netlink_ns_capable(skb, dest_net->user_ns, CAP_NET_ADMIN))
2343 if (tb[IFLA_LINK_NETNSID]) {
2344 int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
2346 link_net = get_net_ns_by_id(dest_net, id);
2352 if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
2356 dev = rtnl_create_link(link_net ? : dest_net, ifname,
2357 name_assign_type, ops, tb);
2363 dev->ifindex = ifm->ifi_index;
2366 err = ops->newlink(link_net ? : net, dev, tb, data);
2367 /* Drivers should call free_netdev() in ->destructor
2368 * and unregister it on failure after registration
2369 * so that device could be finally freed in rtnl_unlock.
2372 /* If device is not registered at all, free it now */
2373 if (dev->reg_state == NETREG_UNINITIALIZED)
2378 err = register_netdevice(dev);
2384 err = rtnl_configure_link(dev, ifm);
2386 goto out_unregister;
2388 err = dev_change_net_namespace(dev, dest_net, ifname);
2390 goto out_unregister;
2399 LIST_HEAD(list_kill);
2401 ops->dellink(dev, &list_kill);
2402 unregister_netdevice_many(&list_kill);
2404 unregister_netdevice(dev);
2410 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh)
2412 struct net *net = sock_net(skb->sk);
2413 struct ifinfomsg *ifm;
2414 char ifname[IFNAMSIZ];
2415 struct nlattr *tb[IFLA_MAX+1];
2416 struct net_device *dev = NULL;
2417 struct sk_buff *nskb;
2419 u32 ext_filter_mask = 0;
2421 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2425 if (tb[IFLA_IFNAME])
2426 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2428 if (tb[IFLA_EXT_MASK])
2429 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2431 ifm = nlmsg_data(nlh);
2432 if (ifm->ifi_index > 0)
2433 dev = __dev_get_by_index(net, ifm->ifi_index);
2434 else if (tb[IFLA_IFNAME])
2435 dev = __dev_get_by_name(net, ifname);
2442 nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
2446 err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).portid,
2447 nlh->nlmsg_seq, 0, 0, ext_filter_mask);
2449 /* -EMSGSIZE implies BUG in if_nlmsg_size */
2450 WARN_ON(err == -EMSGSIZE);
2453 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
2458 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
2460 struct net *net = sock_net(skb->sk);
2461 struct net_device *dev;
2462 struct nlattr *tb[IFLA_MAX+1];
2463 u32 ext_filter_mask = 0;
2464 u16 min_ifinfo_dump_size = 0;
2467 /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
2468 hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
2469 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
2471 if (nlmsg_parse(nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
2472 if (tb[IFLA_EXT_MASK])
2473 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2476 if (!ext_filter_mask)
2477 return NLMSG_GOODSIZE;
2479 * traverse the list of net devices and compute the minimum
2480 * buffer size based upon the filter mask.
2482 list_for_each_entry(dev, &net->dev_base_head, dev_list) {
2483 min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
2488 return min_ifinfo_dump_size;
2491 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
2494 int s_idx = cb->family;
2498 for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
2499 int type = cb->nlh->nlmsg_type-RTM_BASE;
2500 if (idx < s_idx || idx == PF_PACKET)
2502 if (rtnl_msg_handlers[idx] == NULL ||
2503 rtnl_msg_handlers[idx][type].dumpit == NULL)
2506 memset(&cb->args[0], 0, sizeof(cb->args));
2510 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
2518 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
2519 unsigned int change, gfp_t flags)
2521 struct net *net = dev_net(dev);
2522 struct sk_buff *skb;
2524 size_t if_info_size;
2526 skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
2530 err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0);
2532 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
2533 WARN_ON(err == -EMSGSIZE);
2540 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2544 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags)
2546 struct net *net = dev_net(dev);
2548 rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
2551 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
2554 struct sk_buff *skb;
2556 if (dev->reg_state != NETREG_REGISTERED)
2559 skb = rtmsg_ifinfo_build_skb(type, dev, change, flags);
2561 rtmsg_ifinfo_send(skb, dev, flags);
2563 EXPORT_SYMBOL(rtmsg_ifinfo);
2565 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
2566 struct net_device *dev,
2567 u8 *addr, u16 vid, u32 pid, u32 seq,
2568 int type, unsigned int flags,
2571 struct nlmsghdr *nlh;
2574 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
2578 ndm = nlmsg_data(nlh);
2579 ndm->ndm_family = AF_BRIDGE;
2582 ndm->ndm_flags = flags;
2584 ndm->ndm_ifindex = dev->ifindex;
2585 ndm->ndm_state = NUD_PERMANENT;
2587 if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
2588 goto nla_put_failure;
2590 if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
2591 goto nla_put_failure;
2593 nlmsg_end(skb, nlh);
2597 nlmsg_cancel(skb, nlh);
2601 static inline size_t rtnl_fdb_nlmsg_size(void)
2603 return NLMSG_ALIGN(sizeof(struct ndmsg)) +
2604 nla_total_size(ETH_ALEN) + /* NDA_LLADDR */
2605 nla_total_size(sizeof(u16)) + /* NDA_VLAN */
2609 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type)
2611 struct net *net = dev_net(dev);
2612 struct sk_buff *skb;
2615 skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
2619 err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
2620 0, 0, type, NTF_SELF, 0);
2626 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2629 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
2633 * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
2635 int ndo_dflt_fdb_add(struct ndmsg *ndm,
2636 struct nlattr *tb[],
2637 struct net_device *dev,
2638 const unsigned char *addr, u16 vid,
2643 /* If aging addresses are supported device will need to
2644 * implement its own handler for this.
2646 if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
2647 pr_info("%s: FDB only supports static addresses\n", dev->name);
2652 pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
2656 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2657 err = dev_uc_add_excl(dev, addr);
2658 else if (is_multicast_ether_addr(addr))
2659 err = dev_mc_add_excl(dev, addr);
2661 /* Only return duplicate errors if NLM_F_EXCL is set */
2662 if (err == -EEXIST && !(flags & NLM_F_EXCL))
2667 EXPORT_SYMBOL(ndo_dflt_fdb_add);
2669 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid)
2674 if (nla_len(vlan_attr) != sizeof(u16)) {
2675 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan\n");
2679 vid = nla_get_u16(vlan_attr);
2681 if (!vid || vid >= VLAN_VID_MASK) {
2682 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan id %d\n",
2691 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh)
2693 struct net *net = sock_net(skb->sk);
2695 struct nlattr *tb[NDA_MAX+1];
2696 struct net_device *dev;
2701 err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2705 ndm = nlmsg_data(nlh);
2706 if (ndm->ndm_ifindex == 0) {
2707 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n");
2711 dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2713 pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n");
2717 if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2718 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n");
2722 addr = nla_data(tb[NDA_LLADDR]);
2724 err = fdb_vid_parse(tb[NDA_VLAN], &vid);
2730 /* Support fdb on master device the net/bridge default case */
2731 if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2732 (dev->priv_flags & IFF_BRIDGE_PORT)) {
2733 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2734 const struct net_device_ops *ops = br_dev->netdev_ops;
2736 err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
2741 ndm->ndm_flags &= ~NTF_MASTER;
2744 /* Embedded bridge, macvlan, and any other device support */
2745 if ((ndm->ndm_flags & NTF_SELF)) {
2746 if (dev->netdev_ops->ndo_fdb_add)
2747 err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
2751 err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
2755 rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH);
2756 ndm->ndm_flags &= ~NTF_SELF;
2764 * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
2766 int ndo_dflt_fdb_del(struct ndmsg *ndm,
2767 struct nlattr *tb[],
2768 struct net_device *dev,
2769 const unsigned char *addr, u16 vid)
2773 /* If aging addresses are supported device will need to
2774 * implement its own handler for this.
2776 if (!(ndm->ndm_state & NUD_PERMANENT)) {
2777 pr_info("%s: FDB only supports static addresses\n", dev->name);
2781 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2782 err = dev_uc_del(dev, addr);
2783 else if (is_multicast_ether_addr(addr))
2784 err = dev_mc_del(dev, addr);
2788 EXPORT_SYMBOL(ndo_dflt_fdb_del);
2790 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh)
2792 struct net *net = sock_net(skb->sk);
2794 struct nlattr *tb[NDA_MAX+1];
2795 struct net_device *dev;
2800 if (!netlink_capable(skb, CAP_NET_ADMIN))
2803 err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2807 ndm = nlmsg_data(nlh);
2808 if (ndm->ndm_ifindex == 0) {
2809 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n");
2813 dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2815 pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n");
2819 if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2820 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid address\n");
2824 addr = nla_data(tb[NDA_LLADDR]);
2826 err = fdb_vid_parse(tb[NDA_VLAN], &vid);
2832 /* Support fdb on master device the net/bridge default case */
2833 if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2834 (dev->priv_flags & IFF_BRIDGE_PORT)) {
2835 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2836 const struct net_device_ops *ops = br_dev->netdev_ops;
2838 if (ops->ndo_fdb_del)
2839 err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid);
2844 ndm->ndm_flags &= ~NTF_MASTER;
2847 /* Embedded bridge, macvlan, and any other device support */
2848 if (ndm->ndm_flags & NTF_SELF) {
2849 if (dev->netdev_ops->ndo_fdb_del)
2850 err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr,
2853 err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
2856 rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH);
2857 ndm->ndm_flags &= ~NTF_SELF;
2864 static int nlmsg_populate_fdb(struct sk_buff *skb,
2865 struct netlink_callback *cb,
2866 struct net_device *dev,
2868 struct netdev_hw_addr_list *list)
2870 struct netdev_hw_addr *ha;
2874 portid = NETLINK_CB(cb->skb).portid;
2875 seq = cb->nlh->nlmsg_seq;
2877 list_for_each_entry(ha, &list->list, list) {
2878 if (*idx < cb->args[0])
2881 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
2883 RTM_NEWNEIGH, NTF_SELF,
2894 * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
2895 * @nlh: netlink message header
2898 * Default netdevice operation to dump the existing unicast address list.
2899 * Returns number of addresses from list put in skb.
2901 int ndo_dflt_fdb_dump(struct sk_buff *skb,
2902 struct netlink_callback *cb,
2903 struct net_device *dev,
2904 struct net_device *filter_dev,
2909 netif_addr_lock_bh(dev);
2910 err = nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->uc);
2913 nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->mc);
2915 netif_addr_unlock_bh(dev);
2918 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
2920 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
2922 struct net_device *dev;
2923 struct nlattr *tb[IFLA_MAX+1];
2924 struct net_device *br_dev = NULL;
2925 const struct net_device_ops *ops = NULL;
2926 const struct net_device_ops *cops = NULL;
2927 struct ifinfomsg *ifm = nlmsg_data(cb->nlh);
2928 struct net *net = sock_net(skb->sk);
2933 if (nlmsg_parse(cb->nlh, sizeof(struct ifinfomsg), tb, IFLA_MAX,
2934 ifla_policy) == 0) {
2935 if (tb[IFLA_MASTER])
2936 br_idx = nla_get_u32(tb[IFLA_MASTER]);
2939 brport_idx = ifm->ifi_index;
2942 br_dev = __dev_get_by_index(net, br_idx);
2946 ops = br_dev->netdev_ops;
2949 for_each_netdev(net, dev) {
2950 if (brport_idx && (dev->ifindex != brport_idx))
2953 if (!br_idx) { /* user did not specify a specific bridge */
2954 if (dev->priv_flags & IFF_BRIDGE_PORT) {
2955 br_dev = netdev_master_upper_dev_get(dev);
2956 cops = br_dev->netdev_ops;
2960 if (dev != br_dev &&
2961 !(dev->priv_flags & IFF_BRIDGE_PORT))
2964 if (br_dev != netdev_master_upper_dev_get(dev) &&
2965 !(dev->priv_flags & IFF_EBRIDGE))
2971 if (dev->priv_flags & IFF_BRIDGE_PORT) {
2972 if (cops && cops->ndo_fdb_dump)
2973 idx = cops->ndo_fdb_dump(skb, cb, br_dev, dev,
2977 if (dev->netdev_ops->ndo_fdb_dump)
2978 idx = dev->netdev_ops->ndo_fdb_dump(skb, cb, dev, NULL,
2981 idx = ndo_dflt_fdb_dump(skb, cb, dev, NULL, idx);
2990 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
2991 unsigned int attrnum, unsigned int flag)
2994 return nla_put_u8(skb, attrnum, !!(flags & flag));
2998 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
2999 struct net_device *dev, u16 mode,
3000 u32 flags, u32 mask, int nlflags,
3002 int (*vlan_fill)(struct sk_buff *skb,
3003 struct net_device *dev,
3006 struct nlmsghdr *nlh;
3007 struct ifinfomsg *ifm;
3008 struct nlattr *br_afspec;
3009 struct nlattr *protinfo;
3010 u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
3011 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3014 nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
3018 ifm = nlmsg_data(nlh);
3019 ifm->ifi_family = AF_BRIDGE;
3021 ifm->ifi_type = dev->type;
3022 ifm->ifi_index = dev->ifindex;
3023 ifm->ifi_flags = dev_get_flags(dev);
3024 ifm->ifi_change = 0;
3027 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
3028 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
3029 nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
3031 nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
3033 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
3034 (dev->ifindex != dev_get_iflink(dev) &&
3035 nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
3036 goto nla_put_failure;
3038 br_afspec = nla_nest_start(skb, IFLA_AF_SPEC);
3040 goto nla_put_failure;
3042 if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
3043 nla_nest_cancel(skb, br_afspec);
3044 goto nla_put_failure;
3047 if (mode != BRIDGE_MODE_UNDEF) {
3048 if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
3049 nla_nest_cancel(skb, br_afspec);
3050 goto nla_put_failure;
3054 err = vlan_fill(skb, dev, filter_mask);
3056 nla_nest_cancel(skb, br_afspec);
3057 goto nla_put_failure;
3060 nla_nest_end(skb, br_afspec);
3062 protinfo = nla_nest_start(skb, IFLA_PROTINFO | NLA_F_NESTED);
3064 goto nla_put_failure;
3066 if (brport_nla_put_flag(skb, flags, mask,
3067 IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
3068 brport_nla_put_flag(skb, flags, mask,
3069 IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
3070 brport_nla_put_flag(skb, flags, mask,
3071 IFLA_BRPORT_FAST_LEAVE,
3072 BR_MULTICAST_FAST_LEAVE) ||
3073 brport_nla_put_flag(skb, flags, mask,
3074 IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
3075 brport_nla_put_flag(skb, flags, mask,
3076 IFLA_BRPORT_LEARNING, BR_LEARNING) ||
3077 brport_nla_put_flag(skb, flags, mask,
3078 IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
3079 brport_nla_put_flag(skb, flags, mask,
3080 IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
3081 brport_nla_put_flag(skb, flags, mask,
3082 IFLA_BRPORT_PROXYARP, BR_PROXYARP)) {
3083 nla_nest_cancel(skb, protinfo);
3084 goto nla_put_failure;
3087 nla_nest_end(skb, protinfo);
3089 nlmsg_end(skb, nlh);
3092 nlmsg_cancel(skb, nlh);
3093 return err ? err : -EMSGSIZE;
3095 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink);
3097 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
3099 struct net *net = sock_net(skb->sk);
3100 struct net_device *dev;
3102 u32 portid = NETLINK_CB(cb->skb).portid;
3103 u32 seq = cb->nlh->nlmsg_seq;
3104 u32 filter_mask = 0;
3107 if (nlmsg_len(cb->nlh) > sizeof(struct ifinfomsg)) {
3108 struct nlattr *extfilt;
3110 extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct ifinfomsg),
3113 if (nla_len(extfilt) < sizeof(filter_mask))
3116 filter_mask = nla_get_u32(extfilt);
3121 for_each_netdev_rcu(net, dev) {
3122 const struct net_device_ops *ops = dev->netdev_ops;
3123 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3125 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
3126 if (idx >= cb->args[0]) {
3127 err = br_dev->netdev_ops->ndo_bridge_getlink(
3128 skb, portid, seq, dev,
3129 filter_mask, NLM_F_MULTI);
3130 if (err < 0 && err != -EOPNOTSUPP)
3136 if (ops->ndo_bridge_getlink) {
3137 if (idx >= cb->args[0]) {
3138 err = ops->ndo_bridge_getlink(skb, portid,
3142 if (err < 0 && err != -EOPNOTSUPP)
3154 static inline size_t bridge_nlmsg_size(void)
3156 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3157 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3158 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3159 + nla_total_size(sizeof(u32)) /* IFLA_MASTER */
3160 + nla_total_size(sizeof(u32)) /* IFLA_MTU */
3161 + nla_total_size(sizeof(u32)) /* IFLA_LINK */
3162 + nla_total_size(sizeof(u32)) /* IFLA_OPERSTATE */
3163 + nla_total_size(sizeof(u8)) /* IFLA_PROTINFO */
3164 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
3165 + nla_total_size(sizeof(u16)) /* IFLA_BRIDGE_FLAGS */
3166 + nla_total_size(sizeof(u16)); /* IFLA_BRIDGE_MODE */
3169 static int rtnl_bridge_notify(struct net_device *dev)
3171 struct net *net = dev_net(dev);
3172 struct sk_buff *skb;
3173 int err = -EOPNOTSUPP;
3175 if (!dev->netdev_ops->ndo_bridge_getlink)
3178 skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
3184 err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
3191 rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
3194 WARN_ON(err == -EMSGSIZE);
3197 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
3201 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
3203 struct net *net = sock_net(skb->sk);
3204 struct ifinfomsg *ifm;
3205 struct net_device *dev;
3206 struct nlattr *br_spec, *attr = NULL;
3207 int rem, err = -EOPNOTSUPP;
3209 bool have_flags = false;
3211 if (nlmsg_len(nlh) < sizeof(*ifm))
3214 ifm = nlmsg_data(nlh);
3215 if (ifm->ifi_family != AF_BRIDGE)
3216 return -EPFNOSUPPORT;
3218 dev = __dev_get_by_index(net, ifm->ifi_index);
3220 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3224 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3226 nla_for_each_nested(attr, br_spec, rem) {
3227 if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3228 if (nla_len(attr) < sizeof(flags))
3232 flags = nla_get_u16(attr);
3238 if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3239 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3241 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
3246 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags);
3250 flags &= ~BRIDGE_FLAGS_MASTER;
3253 if ((flags & BRIDGE_FLAGS_SELF)) {
3254 if (!dev->netdev_ops->ndo_bridge_setlink)
3257 err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
3260 flags &= ~BRIDGE_FLAGS_SELF;
3262 /* Generate event to notify upper layer of bridge
3265 err = rtnl_bridge_notify(dev);
3270 memcpy(nla_data(attr), &flags, sizeof(flags));
3275 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
3277 struct net *net = sock_net(skb->sk);
3278 struct ifinfomsg *ifm;
3279 struct net_device *dev;
3280 struct nlattr *br_spec, *attr = NULL;
3281 int rem, err = -EOPNOTSUPP;
3283 bool have_flags = false;
3285 if (nlmsg_len(nlh) < sizeof(*ifm))
3288 ifm = nlmsg_data(nlh);
3289 if (ifm->ifi_family != AF_BRIDGE)
3290 return -EPFNOSUPPORT;
3292 dev = __dev_get_by_index(net, ifm->ifi_index);
3294 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3298 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3300 nla_for_each_nested(attr, br_spec, rem) {
3301 if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3302 if (nla_len(attr) < sizeof(flags))
3306 flags = nla_get_u16(attr);
3312 if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3313 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3315 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
3320 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
3324 flags &= ~BRIDGE_FLAGS_MASTER;
3327 if ((flags & BRIDGE_FLAGS_SELF)) {
3328 if (!dev->netdev_ops->ndo_bridge_dellink)
3331 err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
3335 flags &= ~BRIDGE_FLAGS_SELF;
3337 /* Generate event to notify upper layer of bridge
3340 err = rtnl_bridge_notify(dev);
3345 memcpy(nla_data(attr), &flags, sizeof(flags));
3350 /* Process one rtnetlink message. */
3352 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
3354 struct net *net = sock_net(skb->sk);
3355 rtnl_doit_func doit;
3361 type = nlh->nlmsg_type;
3367 /* All the messages must have at least 1 byte length */
3368 if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
3371 family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
3375 if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
3378 if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
3380 rtnl_dumpit_func dumpit;
3381 rtnl_calcit_func calcit;
3382 u16 min_dump_alloc = 0;
3384 dumpit = rtnl_get_dumpit(family, type);
3387 calcit = rtnl_get_calcit(family, type);
3389 min_dump_alloc = calcit(skb, nlh);
3394 struct netlink_dump_control c = {
3396 .min_dump_alloc = min_dump_alloc,
3398 err = netlink_dump_start(rtnl, skb, nlh, &c);
3404 doit = rtnl_get_doit(family, type);
3408 return doit(skb, nlh);
3411 static void rtnetlink_rcv(struct sk_buff *skb)
3414 netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
3418 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
3420 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3426 case NETDEV_POST_INIT:
3427 case NETDEV_REGISTER:
3429 case NETDEV_PRE_TYPE_CHANGE:
3430 case NETDEV_GOING_DOWN:
3431 case NETDEV_UNREGISTER:
3432 case NETDEV_UNREGISTER_FINAL:
3433 case NETDEV_RELEASE:
3435 case NETDEV_BONDING_INFO:
3438 rtmsg_ifinfo(RTM_NEWLINK, dev, 0, GFP_KERNEL);
3444 static struct notifier_block rtnetlink_dev_notifier = {
3445 .notifier_call = rtnetlink_event,
3449 static int __net_init rtnetlink_net_init(struct net *net)
3452 struct netlink_kernel_cfg cfg = {
3453 .groups = RTNLGRP_MAX,
3454 .input = rtnetlink_rcv,
3455 .cb_mutex = &rtnl_mutex,
3456 .flags = NL_CFG_F_NONROOT_RECV,
3459 sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
3466 static void __net_exit rtnetlink_net_exit(struct net *net)
3468 netlink_kernel_release(net->rtnl);
3472 static struct pernet_operations rtnetlink_net_ops = {
3473 .init = rtnetlink_net_init,
3474 .exit = rtnetlink_net_exit,
3477 void __init rtnetlink_init(void)
3479 if (register_pernet_subsys(&rtnetlink_net_ops))
3480 panic("rtnetlink_init: cannot initialize rtnetlink\n");
3482 register_netdevice_notifier(&rtnetlink_dev_notifier);
3484 rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
3485 rtnl_dump_ifinfo, rtnl_calcit);
3486 rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
3487 rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
3488 rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
3490 rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
3491 rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
3493 rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, NULL);
3494 rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, NULL);
3495 rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, NULL);
3497 rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, NULL);
3498 rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, NULL);
3499 rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, NULL);