2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
6 * Routing netlink socket interface: protocol independent part.
8 * Authors: Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
10 * This program is free software; you can redistribute it and/or
11 * modify it under the terms of the GNU General Public License
12 * as published by the Free Software Foundation; either version
13 * 2 of the License, or (at your option) any later version.
16 * Vitaly E. Lavrov RTA_OK arithmetics was wrong.
19 #include <linux/errno.h>
20 #include <linux/module.h>
21 #include <linux/types.h>
22 #include <linux/socket.h>
23 #include <linux/kernel.h>
24 #include <linux/timer.h>
25 #include <linux/string.h>
26 #include <linux/sockios.h>
27 #include <linux/net.h>
28 #include <linux/fcntl.h>
30 #include <linux/slab.h>
31 #include <linux/interrupt.h>
32 #include <linux/capability.h>
33 #include <linux/skbuff.h>
34 #include <linux/init.h>
35 #include <linux/security.h>
36 #include <linux/mutex.h>
37 #include <linux/if_addr.h>
38 #include <linux/if_bridge.h>
39 #include <linux/pci.h>
40 #include <linux/etherdevice.h>
42 #include <asm/uaccess.h>
44 #include <linux/inet.h>
45 #include <linux/netdevice.h>
47 #include <net/protocol.h>
49 #include <net/route.h>
52 #include <net/pkt_sched.h>
53 #include <net/fib_rules.h>
54 #include <net/rtnetlink.h>
55 #include <net/net_namespace.h>
59 rtnl_dumpit_func dumpit;
60 rtnl_calcit_func calcit;
63 static DEFINE_MUTEX(rtnl_mutex);
67 mutex_lock(&rtnl_mutex);
69 EXPORT_SYMBOL(rtnl_lock);
71 void __rtnl_unlock(void)
73 mutex_unlock(&rtnl_mutex);
76 void rtnl_unlock(void)
78 /* This fellow will unlock it for us. */
81 EXPORT_SYMBOL(rtnl_unlock);
83 int rtnl_trylock(void)
85 return mutex_trylock(&rtnl_mutex);
87 EXPORT_SYMBOL(rtnl_trylock);
89 int rtnl_is_locked(void)
91 return mutex_is_locked(&rtnl_mutex);
93 EXPORT_SYMBOL(rtnl_is_locked);
95 #ifdef CONFIG_PROVE_LOCKING
96 int lockdep_rtnl_is_held(void)
98 return lockdep_is_held(&rtnl_mutex);
100 EXPORT_SYMBOL(lockdep_rtnl_is_held);
101 #endif /* #ifdef CONFIG_PROVE_LOCKING */
103 static struct rtnl_link *rtnl_msg_handlers[RTNL_FAMILY_MAX + 1];
105 static inline int rtm_msgindex(int msgtype)
107 int msgindex = msgtype - RTM_BASE;
110 * msgindex < 0 implies someone tried to register a netlink
111 * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
112 * the message type has not been added to linux/rtnetlink.h
114 BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
119 static rtnl_doit_func rtnl_get_doit(int protocol, int msgindex)
121 struct rtnl_link *tab;
123 if (protocol <= RTNL_FAMILY_MAX)
124 tab = rtnl_msg_handlers[protocol];
128 if (tab == NULL || tab[msgindex].doit == NULL)
129 tab = rtnl_msg_handlers[PF_UNSPEC];
131 return tab[msgindex].doit;
134 static rtnl_dumpit_func rtnl_get_dumpit(int protocol, int msgindex)
136 struct rtnl_link *tab;
138 if (protocol <= RTNL_FAMILY_MAX)
139 tab = rtnl_msg_handlers[protocol];
143 if (tab == NULL || tab[msgindex].dumpit == NULL)
144 tab = rtnl_msg_handlers[PF_UNSPEC];
146 return tab[msgindex].dumpit;
149 static rtnl_calcit_func rtnl_get_calcit(int protocol, int msgindex)
151 struct rtnl_link *tab;
153 if (protocol <= RTNL_FAMILY_MAX)
154 tab = rtnl_msg_handlers[protocol];
158 if (tab == NULL || tab[msgindex].calcit == NULL)
159 tab = rtnl_msg_handlers[PF_UNSPEC];
161 return tab[msgindex].calcit;
165 * __rtnl_register - Register a rtnetlink message type
166 * @protocol: Protocol family or PF_UNSPEC
167 * @msgtype: rtnetlink message type
168 * @doit: Function pointer called for each request message
169 * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
170 * @calcit: Function pointer to calc size of dump message
172 * Registers the specified function pointers (at least one of them has
173 * to be non-NULL) to be called whenever a request message for the
174 * specified protocol family and message type is received.
176 * The special protocol family PF_UNSPEC may be used to define fallback
177 * function pointers for the case when no entry for the specific protocol
180 * Returns 0 on success or a negative error code.
182 int __rtnl_register(int protocol, int msgtype,
183 rtnl_doit_func doit, rtnl_dumpit_func dumpit,
184 rtnl_calcit_func calcit)
186 struct rtnl_link *tab;
189 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
190 msgindex = rtm_msgindex(msgtype);
192 tab = rtnl_msg_handlers[protocol];
194 tab = kcalloc(RTM_NR_MSGTYPES, sizeof(*tab), GFP_KERNEL);
198 rtnl_msg_handlers[protocol] = tab;
202 tab[msgindex].doit = doit;
205 tab[msgindex].dumpit = dumpit;
208 tab[msgindex].calcit = calcit;
212 EXPORT_SYMBOL_GPL(__rtnl_register);
215 * rtnl_register - Register a rtnetlink message type
217 * Identical to __rtnl_register() but panics on failure. This is useful
218 * as failure of this function is very unlikely, it can only happen due
219 * to lack of memory when allocating the chain to store all message
220 * handlers for a protocol. Meant for use in init functions where lack
221 * of memory implies no sense in continuing.
223 void rtnl_register(int protocol, int msgtype,
224 rtnl_doit_func doit, rtnl_dumpit_func dumpit,
225 rtnl_calcit_func calcit)
227 if (__rtnl_register(protocol, msgtype, doit, dumpit, calcit) < 0)
228 panic("Unable to register rtnetlink message handler, "
229 "protocol = %d, message type = %d\n",
232 EXPORT_SYMBOL_GPL(rtnl_register);
235 * rtnl_unregister - Unregister a rtnetlink message type
236 * @protocol: Protocol family or PF_UNSPEC
237 * @msgtype: rtnetlink message type
239 * Returns 0 on success or a negative error code.
241 int rtnl_unregister(int protocol, int msgtype)
245 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
246 msgindex = rtm_msgindex(msgtype);
248 if (rtnl_msg_handlers[protocol] == NULL)
251 rtnl_msg_handlers[protocol][msgindex].doit = NULL;
252 rtnl_msg_handlers[protocol][msgindex].dumpit = NULL;
256 EXPORT_SYMBOL_GPL(rtnl_unregister);
259 * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
260 * @protocol : Protocol family or PF_UNSPEC
262 * Identical to calling rtnl_unregster() for all registered message types
263 * of a certain protocol family.
265 void rtnl_unregister_all(int protocol)
267 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
269 kfree(rtnl_msg_handlers[protocol]);
270 rtnl_msg_handlers[protocol] = NULL;
272 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
274 static LIST_HEAD(link_ops);
276 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
278 const struct rtnl_link_ops *ops;
280 list_for_each_entry(ops, &link_ops, list) {
281 if (!strcmp(ops->kind, kind))
288 * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
289 * @ops: struct rtnl_link_ops * to register
291 * The caller must hold the rtnl_mutex. This function should be used
292 * by drivers that create devices during module initialization. It
293 * must be called before registering the devices.
295 * Returns 0 on success or a negative error code.
297 int __rtnl_link_register(struct rtnl_link_ops *ops)
299 if (rtnl_link_ops_get(ops->kind))
303 ops->dellink = unregister_netdevice_queue;
305 list_add_tail(&ops->list, &link_ops);
308 EXPORT_SYMBOL_GPL(__rtnl_link_register);
311 * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
312 * @ops: struct rtnl_link_ops * to register
314 * Returns 0 on success or a negative error code.
316 int rtnl_link_register(struct rtnl_link_ops *ops)
321 err = __rtnl_link_register(ops);
325 EXPORT_SYMBOL_GPL(rtnl_link_register);
327 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
329 struct net_device *dev;
330 LIST_HEAD(list_kill);
332 for_each_netdev(net, dev) {
333 if (dev->rtnl_link_ops == ops)
334 ops->dellink(dev, &list_kill);
336 unregister_netdevice_many(&list_kill);
340 * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
341 * @ops: struct rtnl_link_ops * to unregister
343 * The caller must hold the rtnl_mutex.
345 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
350 __rtnl_kill_links(net, ops);
352 list_del(&ops->list);
354 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
357 * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
358 * @ops: struct rtnl_link_ops * to unregister
360 void rtnl_link_unregister(struct rtnl_link_ops *ops)
363 __rtnl_link_unregister(ops);
366 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
368 static size_t rtnl_link_get_slave_info_data_size(const struct net_device *dev)
370 struct net_device *master_dev;
371 const struct rtnl_link_ops *ops;
373 master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
376 ops = master_dev->rtnl_link_ops;
377 if (!ops || !ops->get_slave_size)
379 /* IFLA_INFO_SLAVE_DATA + nested data */
380 return nla_total_size(sizeof(struct nlattr)) +
381 ops->get_slave_size(master_dev, dev);
384 static size_t rtnl_link_get_size(const struct net_device *dev)
386 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
392 size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
393 nla_total_size(strlen(ops->kind) + 1); /* IFLA_INFO_KIND */
396 /* IFLA_INFO_DATA + nested data */
397 size += nla_total_size(sizeof(struct nlattr)) +
400 if (ops->get_xstats_size)
401 /* IFLA_INFO_XSTATS */
402 size += nla_total_size(ops->get_xstats_size(dev));
404 size += rtnl_link_get_slave_info_data_size(dev);
409 static LIST_HEAD(rtnl_af_ops);
411 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
413 const struct rtnl_af_ops *ops;
415 list_for_each_entry(ops, &rtnl_af_ops, list) {
416 if (ops->family == family)
424 * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
425 * @ops: struct rtnl_af_ops * to register
427 * Returns 0 on success or a negative error code.
429 void rtnl_af_register(struct rtnl_af_ops *ops)
432 list_add_tail(&ops->list, &rtnl_af_ops);
435 EXPORT_SYMBOL_GPL(rtnl_af_register);
438 * __rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
439 * @ops: struct rtnl_af_ops * to unregister
441 * The caller must hold the rtnl_mutex.
443 void __rtnl_af_unregister(struct rtnl_af_ops *ops)
445 list_del(&ops->list);
447 EXPORT_SYMBOL_GPL(__rtnl_af_unregister);
450 * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
451 * @ops: struct rtnl_af_ops * to unregister
453 void rtnl_af_unregister(struct rtnl_af_ops *ops)
456 __rtnl_af_unregister(ops);
459 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
461 static size_t rtnl_link_get_af_size(const struct net_device *dev)
463 struct rtnl_af_ops *af_ops;
467 size = nla_total_size(sizeof(struct nlattr));
469 list_for_each_entry(af_ops, &rtnl_af_ops, list) {
470 if (af_ops->get_link_af_size) {
471 /* AF_* + nested data */
472 size += nla_total_size(sizeof(struct nlattr)) +
473 af_ops->get_link_af_size(dev);
480 static bool rtnl_have_link_slave_info(const struct net_device *dev)
482 struct net_device *master_dev;
484 master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
485 if (master_dev && master_dev->rtnl_link_ops)
490 static int rtnl_link_slave_info_fill(struct sk_buff *skb,
491 const struct net_device *dev)
493 struct net_device *master_dev;
494 const struct rtnl_link_ops *ops;
495 struct nlattr *slave_data;
498 master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
501 ops = master_dev->rtnl_link_ops;
504 if (nla_put_string(skb, IFLA_INFO_SLAVE_KIND, ops->kind) < 0)
506 if (ops->fill_slave_info) {
507 slave_data = nla_nest_start(skb, IFLA_INFO_SLAVE_DATA);
510 err = ops->fill_slave_info(skb, master_dev, dev);
512 goto err_cancel_slave_data;
513 nla_nest_end(skb, slave_data);
517 err_cancel_slave_data:
518 nla_nest_cancel(skb, slave_data);
522 static int rtnl_link_info_fill(struct sk_buff *skb,
523 const struct net_device *dev)
525 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
531 if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
533 if (ops->fill_xstats) {
534 err = ops->fill_xstats(skb, dev);
538 if (ops->fill_info) {
539 data = nla_nest_start(skb, IFLA_INFO_DATA);
542 err = ops->fill_info(skb, dev);
544 goto err_cancel_data;
545 nla_nest_end(skb, data);
550 nla_nest_cancel(skb, data);
554 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
556 struct nlattr *linkinfo;
559 linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
560 if (linkinfo == NULL)
563 err = rtnl_link_info_fill(skb, dev);
565 goto err_cancel_link;
567 err = rtnl_link_slave_info_fill(skb, dev);
569 goto err_cancel_link;
571 nla_nest_end(skb, linkinfo);
575 nla_nest_cancel(skb, linkinfo);
580 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
582 struct sock *rtnl = net->rtnl;
585 NETLINK_CB(skb).dst_group = group;
587 atomic_inc(&skb->users);
588 netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
590 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
594 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
596 struct sock *rtnl = net->rtnl;
598 return nlmsg_unicast(rtnl, skb, pid);
600 EXPORT_SYMBOL(rtnl_unicast);
602 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
603 struct nlmsghdr *nlh, gfp_t flags)
605 struct sock *rtnl = net->rtnl;
609 report = nlmsg_report(nlh);
611 nlmsg_notify(rtnl, skb, pid, group, report, flags);
613 EXPORT_SYMBOL(rtnl_notify);
615 void rtnl_set_sk_err(struct net *net, u32 group, int error)
617 struct sock *rtnl = net->rtnl;
619 netlink_set_err(rtnl, 0, group, error);
621 EXPORT_SYMBOL(rtnl_set_sk_err);
623 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
628 mx = nla_nest_start(skb, RTA_METRICS);
632 for (i = 0; i < RTAX_MAX; i++) {
635 if (nla_put_u32(skb, i+1, metrics[i]))
636 goto nla_put_failure;
641 nla_nest_cancel(skb, mx);
645 return nla_nest_end(skb, mx);
648 nla_nest_cancel(skb, mx);
651 EXPORT_SYMBOL(rtnetlink_put_metrics);
653 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
654 long expires, u32 error)
656 struct rta_cacheinfo ci = {
657 .rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse),
658 .rta_used = dst->__use,
659 .rta_clntref = atomic_read(&(dst->__refcnt)),
667 clock = jiffies_to_clock_t(abs(expires));
668 clock = min_t(unsigned long, clock, INT_MAX);
669 ci.rta_expires = (expires > 0) ? clock : -clock;
671 return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
673 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
675 static void set_operstate(struct net_device *dev, unsigned char transition)
677 unsigned char operstate = dev->operstate;
679 switch (transition) {
681 if ((operstate == IF_OPER_DORMANT ||
682 operstate == IF_OPER_UNKNOWN) &&
684 operstate = IF_OPER_UP;
687 case IF_OPER_DORMANT:
688 if (operstate == IF_OPER_UP ||
689 operstate == IF_OPER_UNKNOWN)
690 operstate = IF_OPER_DORMANT;
694 if (dev->operstate != operstate) {
695 write_lock_bh(&dev_base_lock);
696 dev->operstate = operstate;
697 write_unlock_bh(&dev_base_lock);
698 netdev_state_change(dev);
702 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
704 return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
705 (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
708 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
709 const struct ifinfomsg *ifm)
711 unsigned int flags = ifm->ifi_flags;
713 /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
715 flags = (flags & ifm->ifi_change) |
716 (rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
721 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
722 const struct rtnl_link_stats64 *b)
724 a->rx_packets = b->rx_packets;
725 a->tx_packets = b->tx_packets;
726 a->rx_bytes = b->rx_bytes;
727 a->tx_bytes = b->tx_bytes;
728 a->rx_errors = b->rx_errors;
729 a->tx_errors = b->tx_errors;
730 a->rx_dropped = b->rx_dropped;
731 a->tx_dropped = b->tx_dropped;
733 a->multicast = b->multicast;
734 a->collisions = b->collisions;
736 a->rx_length_errors = b->rx_length_errors;
737 a->rx_over_errors = b->rx_over_errors;
738 a->rx_crc_errors = b->rx_crc_errors;
739 a->rx_frame_errors = b->rx_frame_errors;
740 a->rx_fifo_errors = b->rx_fifo_errors;
741 a->rx_missed_errors = b->rx_missed_errors;
743 a->tx_aborted_errors = b->tx_aborted_errors;
744 a->tx_carrier_errors = b->tx_carrier_errors;
745 a->tx_fifo_errors = b->tx_fifo_errors;
746 a->tx_heartbeat_errors = b->tx_heartbeat_errors;
747 a->tx_window_errors = b->tx_window_errors;
749 a->rx_compressed = b->rx_compressed;
750 a->tx_compressed = b->tx_compressed;
753 static void copy_rtnl_link_stats64(void *v, const struct rtnl_link_stats64 *b)
755 memcpy(v, b, sizeof(*b));
759 static inline int rtnl_vfinfo_size(const struct net_device *dev,
762 if (dev->dev.parent && dev_is_pci(dev->dev.parent) &&
763 (ext_filter_mask & RTEXT_FILTER_VF)) {
764 int num_vfs = dev_num_vf(dev->dev.parent);
765 size_t size = nla_total_size(sizeof(struct nlattr));
766 size += nla_total_size(num_vfs * sizeof(struct nlattr));
768 (nla_total_size(sizeof(struct ifla_vf_mac)) +
769 nla_total_size(sizeof(struct ifla_vf_vlan)) +
770 nla_total_size(sizeof(struct ifla_vf_tx_rate)) +
771 nla_total_size(sizeof(struct ifla_vf_spoofchk)));
777 static size_t rtnl_port_size(const struct net_device *dev)
779 size_t port_size = nla_total_size(4) /* PORT_VF */
780 + nla_total_size(PORT_PROFILE_MAX) /* PORT_PROFILE */
781 + nla_total_size(sizeof(struct ifla_port_vsi))
783 + nla_total_size(PORT_UUID_MAX) /* PORT_INSTANCE_UUID */
784 + nla_total_size(PORT_UUID_MAX) /* PORT_HOST_UUID */
785 + nla_total_size(1) /* PROT_VDP_REQUEST */
786 + nla_total_size(2); /* PORT_VDP_RESPONSE */
787 size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
788 size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
790 size_t port_self_size = nla_total_size(sizeof(struct nlattr))
793 if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent)
795 if (dev_num_vf(dev->dev.parent))
796 return port_self_size + vf_ports_size +
797 vf_port_size * dev_num_vf(dev->dev.parent);
799 return port_self_size;
802 static noinline size_t if_nlmsg_size(const struct net_device *dev,
805 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
806 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
807 + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
808 + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
809 + nla_total_size(sizeof(struct rtnl_link_ifmap))
810 + nla_total_size(sizeof(struct rtnl_link_stats))
811 + nla_total_size(sizeof(struct rtnl_link_stats64))
812 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
813 + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
814 + nla_total_size(4) /* IFLA_TXQLEN */
815 + nla_total_size(4) /* IFLA_WEIGHT */
816 + nla_total_size(4) /* IFLA_MTU */
817 + nla_total_size(4) /* IFLA_LINK */
818 + nla_total_size(4) /* IFLA_MASTER */
819 + nla_total_size(1) /* IFLA_CARRIER */
820 + nla_total_size(4) /* IFLA_PROMISCUITY */
821 + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
822 + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
823 + nla_total_size(1) /* IFLA_OPERSTATE */
824 + nla_total_size(1) /* IFLA_LINKMODE */
825 + nla_total_size(ext_filter_mask
826 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
827 + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
828 + rtnl_port_size(dev) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
829 + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
830 + rtnl_link_get_af_size(dev) /* IFLA_AF_SPEC */
831 + nla_total_size(MAX_PHYS_PORT_ID_LEN); /* IFLA_PHYS_PORT_ID */
834 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
836 struct nlattr *vf_ports;
837 struct nlattr *vf_port;
841 vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
845 for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
846 vf_port = nla_nest_start(skb, IFLA_VF_PORT);
848 goto nla_put_failure;
849 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
850 goto nla_put_failure;
851 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
852 if (err == -EMSGSIZE)
853 goto nla_put_failure;
855 nla_nest_cancel(skb, vf_port);
858 nla_nest_end(skb, vf_port);
861 nla_nest_end(skb, vf_ports);
866 nla_nest_cancel(skb, vf_ports);
870 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
872 struct nlattr *port_self;
875 port_self = nla_nest_start(skb, IFLA_PORT_SELF);
879 err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
881 nla_nest_cancel(skb, port_self);
882 return (err == -EMSGSIZE) ? err : 0;
885 nla_nest_end(skb, port_self);
890 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev)
894 if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent)
897 err = rtnl_port_self_fill(skb, dev);
901 if (dev_num_vf(dev->dev.parent)) {
902 err = rtnl_vf_ports_fill(skb, dev);
910 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
913 struct netdev_phys_port_id ppid;
915 err = dev_get_phys_port_id(dev, &ppid);
917 if (err == -EOPNOTSUPP)
922 if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
928 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
929 int type, u32 pid, u32 seq, u32 change,
930 unsigned int flags, u32 ext_filter_mask)
932 struct ifinfomsg *ifm;
933 struct nlmsghdr *nlh;
934 struct rtnl_link_stats64 temp;
935 const struct rtnl_link_stats64 *stats;
936 struct nlattr *attr, *af_spec;
937 struct rtnl_af_ops *af_ops;
938 struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
941 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
945 ifm = nlmsg_data(nlh);
946 ifm->ifi_family = AF_UNSPEC;
948 ifm->ifi_type = dev->type;
949 ifm->ifi_index = dev->ifindex;
950 ifm->ifi_flags = dev_get_flags(dev);
951 ifm->ifi_change = change;
953 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
954 nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
955 nla_put_u8(skb, IFLA_OPERSTATE,
956 netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
957 nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
958 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
959 nla_put_u32(skb, IFLA_GROUP, dev->group) ||
960 nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
961 nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
963 nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
965 (dev->ifindex != dev->iflink &&
966 nla_put_u32(skb, IFLA_LINK, dev->iflink)) ||
968 nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex)) ||
969 nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
971 nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
973 nla_put_string(skb, IFLA_IFALIAS, dev->ifalias)))
974 goto nla_put_failure;
977 struct rtnl_link_ifmap map = {
978 .mem_start = dev->mem_start,
979 .mem_end = dev->mem_end,
980 .base_addr = dev->base_addr,
983 .port = dev->if_port,
985 if (nla_put(skb, IFLA_MAP, sizeof(map), &map))
986 goto nla_put_failure;
990 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
991 nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
992 goto nla_put_failure;
995 if (rtnl_phys_port_id_fill(skb, dev))
996 goto nla_put_failure;
998 attr = nla_reserve(skb, IFLA_STATS,
999 sizeof(struct rtnl_link_stats));
1001 goto nla_put_failure;
1003 stats = dev_get_stats(dev, &temp);
1004 copy_rtnl_link_stats(nla_data(attr), stats);
1006 attr = nla_reserve(skb, IFLA_STATS64,
1007 sizeof(struct rtnl_link_stats64));
1009 goto nla_put_failure;
1010 copy_rtnl_link_stats64(nla_data(attr), stats);
1012 if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF) &&
1013 nla_put_u32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent)))
1014 goto nla_put_failure;
1016 if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent
1017 && (ext_filter_mask & RTEXT_FILTER_VF)) {
1020 struct nlattr *vfinfo, *vf;
1021 int num_vfs = dev_num_vf(dev->dev.parent);
1023 vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
1025 goto nla_put_failure;
1026 for (i = 0; i < num_vfs; i++) {
1027 struct ifla_vf_info ivi;
1028 struct ifla_vf_mac vf_mac;
1029 struct ifla_vf_vlan vf_vlan;
1030 struct ifla_vf_tx_rate vf_tx_rate;
1031 struct ifla_vf_spoofchk vf_spoofchk;
1032 struct ifla_vf_link_state vf_linkstate;
1035 * Not all SR-IOV capable drivers support the
1036 * spoofcheck query. Preset to -1 so the user
1037 * space tool can detect that the driver didn't
1041 memset(ivi.mac, 0, sizeof(ivi.mac));
1042 /* The default value for VF link state is "auto"
1043 * IFLA_VF_LINK_STATE_AUTO which equals zero
1046 if (dev->netdev_ops->ndo_get_vf_config(dev, i, &ivi))
1052 vf_linkstate.vf = ivi.vf;
1054 memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1055 vf_vlan.vlan = ivi.vlan;
1056 vf_vlan.qos = ivi.qos;
1057 vf_tx_rate.rate = ivi.tx_rate;
1058 vf_spoofchk.setting = ivi.spoofchk;
1059 vf_linkstate.link_state = ivi.linkstate;
1060 vf = nla_nest_start(skb, IFLA_VF_INFO);
1062 nla_nest_cancel(skb, vfinfo);
1063 goto nla_put_failure;
1065 if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1066 nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1067 nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1069 nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1071 nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1073 goto nla_put_failure;
1074 nla_nest_end(skb, vf);
1076 nla_nest_end(skb, vfinfo);
1079 if (rtnl_port_fill(skb, dev))
1080 goto nla_put_failure;
1082 if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1083 if (rtnl_link_fill(skb, dev) < 0)
1084 goto nla_put_failure;
1087 if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1088 goto nla_put_failure;
1090 list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1091 if (af_ops->fill_link_af) {
1095 if (!(af = nla_nest_start(skb, af_ops->family)))
1096 goto nla_put_failure;
1098 err = af_ops->fill_link_af(skb, dev);
1101 * Caller may return ENODATA to indicate that there
1102 * was no data to be dumped. This is not an error, it
1103 * means we should trim the attribute header and
1106 if (err == -ENODATA)
1107 nla_nest_cancel(skb, af);
1109 goto nla_put_failure;
1111 nla_nest_end(skb, af);
1115 nla_nest_end(skb, af_spec);
1117 return nlmsg_end(skb, nlh);
1120 nlmsg_cancel(skb, nlh);
1124 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1126 struct net *net = sock_net(skb->sk);
1129 struct net_device *dev;
1130 struct hlist_head *head;
1131 struct nlattr *tb[IFLA_MAX+1];
1132 u32 ext_filter_mask = 0;
1135 s_idx = cb->args[1];
1138 cb->seq = net->dev_base_seq;
1140 if (nlmsg_parse(cb->nlh, sizeof(struct ifinfomsg), tb, IFLA_MAX,
1141 ifla_policy) >= 0) {
1143 if (tb[IFLA_EXT_MASK])
1144 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1147 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1149 head = &net->dev_index_head[h];
1150 hlist_for_each_entry_rcu(dev, head, index_hlist) {
1153 if (rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1154 NETLINK_CB(cb->skb).portid,
1155 cb->nlh->nlmsg_seq, 0,
1157 ext_filter_mask) <= 0)
1160 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1173 const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1174 [IFLA_IFNAME] = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1175 [IFLA_ADDRESS] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1176 [IFLA_BROADCAST] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1177 [IFLA_MAP] = { .len = sizeof(struct rtnl_link_ifmap) },
1178 [IFLA_MTU] = { .type = NLA_U32 },
1179 [IFLA_LINK] = { .type = NLA_U32 },
1180 [IFLA_MASTER] = { .type = NLA_U32 },
1181 [IFLA_CARRIER] = { .type = NLA_U8 },
1182 [IFLA_TXQLEN] = { .type = NLA_U32 },
1183 [IFLA_WEIGHT] = { .type = NLA_U32 },
1184 [IFLA_OPERSTATE] = { .type = NLA_U8 },
1185 [IFLA_LINKMODE] = { .type = NLA_U8 },
1186 [IFLA_LINKINFO] = { .type = NLA_NESTED },
1187 [IFLA_NET_NS_PID] = { .type = NLA_U32 },
1188 [IFLA_NET_NS_FD] = { .type = NLA_U32 },
1189 [IFLA_IFALIAS] = { .type = NLA_STRING, .len = IFALIASZ-1 },
1190 [IFLA_VFINFO_LIST] = {. type = NLA_NESTED },
1191 [IFLA_VF_PORTS] = { .type = NLA_NESTED },
1192 [IFLA_PORT_SELF] = { .type = NLA_NESTED },
1193 [IFLA_AF_SPEC] = { .type = NLA_NESTED },
1194 [IFLA_EXT_MASK] = { .type = NLA_U32 },
1195 [IFLA_PROMISCUITY] = { .type = NLA_U32 },
1196 [IFLA_NUM_TX_QUEUES] = { .type = NLA_U32 },
1197 [IFLA_NUM_RX_QUEUES] = { .type = NLA_U32 },
1198 [IFLA_PHYS_PORT_ID] = { .type = NLA_BINARY, .len = MAX_PHYS_PORT_ID_LEN },
1200 EXPORT_SYMBOL(ifla_policy);
1202 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1203 [IFLA_INFO_KIND] = { .type = NLA_STRING },
1204 [IFLA_INFO_DATA] = { .type = NLA_NESTED },
1205 [IFLA_INFO_SLAVE_KIND] = { .type = NLA_STRING },
1206 [IFLA_INFO_SLAVE_DATA] = { .type = NLA_NESTED },
1209 static const struct nla_policy ifla_vfinfo_policy[IFLA_VF_INFO_MAX+1] = {
1210 [IFLA_VF_INFO] = { .type = NLA_NESTED },
1213 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1214 [IFLA_VF_MAC] = { .type = NLA_BINARY,
1215 .len = sizeof(struct ifla_vf_mac) },
1216 [IFLA_VF_VLAN] = { .type = NLA_BINARY,
1217 .len = sizeof(struct ifla_vf_vlan) },
1218 [IFLA_VF_TX_RATE] = { .type = NLA_BINARY,
1219 .len = sizeof(struct ifla_vf_tx_rate) },
1220 [IFLA_VF_SPOOFCHK] = { .type = NLA_BINARY,
1221 .len = sizeof(struct ifla_vf_spoofchk) },
1224 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1225 [IFLA_PORT_VF] = { .type = NLA_U32 },
1226 [IFLA_PORT_PROFILE] = { .type = NLA_STRING,
1227 .len = PORT_PROFILE_MAX },
1228 [IFLA_PORT_VSI_TYPE] = { .type = NLA_BINARY,
1229 .len = sizeof(struct ifla_port_vsi)},
1230 [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1231 .len = PORT_UUID_MAX },
1232 [IFLA_PORT_HOST_UUID] = { .type = NLA_STRING,
1233 .len = PORT_UUID_MAX },
1234 [IFLA_PORT_REQUEST] = { .type = NLA_U8, },
1235 [IFLA_PORT_RESPONSE] = { .type = NLA_U16, },
1238 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1241 /* Examine the link attributes and figure out which
1242 * network namespace we are talking about.
1244 if (tb[IFLA_NET_NS_PID])
1245 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1246 else if (tb[IFLA_NET_NS_FD])
1247 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1249 net = get_net(src_net);
1252 EXPORT_SYMBOL(rtnl_link_get_net);
1254 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1257 if (tb[IFLA_ADDRESS] &&
1258 nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1261 if (tb[IFLA_BROADCAST] &&
1262 nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1266 if (tb[IFLA_AF_SPEC]) {
1270 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1271 const struct rtnl_af_ops *af_ops;
1273 if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1274 return -EAFNOSUPPORT;
1276 if (!af_ops->set_link_af)
1279 if (af_ops->validate_link_af) {
1280 err = af_ops->validate_link_af(dev, af);
1290 static int do_setvfinfo(struct net_device *dev, struct nlattr *attr)
1292 int rem, err = -EINVAL;
1294 const struct net_device_ops *ops = dev->netdev_ops;
1296 nla_for_each_nested(vf, attr, rem) {
1297 switch (nla_type(vf)) {
1299 struct ifla_vf_mac *ivm;
1302 if (ops->ndo_set_vf_mac)
1303 err = ops->ndo_set_vf_mac(dev, ivm->vf,
1307 case IFLA_VF_VLAN: {
1308 struct ifla_vf_vlan *ivv;
1311 if (ops->ndo_set_vf_vlan)
1312 err = ops->ndo_set_vf_vlan(dev, ivv->vf,
1317 case IFLA_VF_TX_RATE: {
1318 struct ifla_vf_tx_rate *ivt;
1321 if (ops->ndo_set_vf_tx_rate)
1322 err = ops->ndo_set_vf_tx_rate(dev, ivt->vf,
1326 case IFLA_VF_SPOOFCHK: {
1327 struct ifla_vf_spoofchk *ivs;
1330 if (ops->ndo_set_vf_spoofchk)
1331 err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1335 case IFLA_VF_LINK_STATE: {
1336 struct ifla_vf_link_state *ivl;
1339 if (ops->ndo_set_vf_link_state)
1340 err = ops->ndo_set_vf_link_state(dev, ivl->vf,
1354 static int do_set_master(struct net_device *dev, int ifindex)
1356 struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1357 const struct net_device_ops *ops;
1361 if (upper_dev->ifindex == ifindex)
1363 ops = upper_dev->netdev_ops;
1364 if (ops->ndo_del_slave) {
1365 err = ops->ndo_del_slave(upper_dev, dev);
1374 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
1377 ops = upper_dev->netdev_ops;
1378 if (ops->ndo_add_slave) {
1379 err = ops->ndo_add_slave(upper_dev, dev);
1389 static int do_setlink(struct net_device *dev, struct ifinfomsg *ifm,
1390 struct nlattr **tb, char *ifname, int modified)
1392 const struct net_device_ops *ops = dev->netdev_ops;
1395 if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1396 struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1401 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) {
1405 err = dev_change_net_namespace(dev, net, ifname);
1413 struct rtnl_link_ifmap *u_map;
1416 if (!ops->ndo_set_config) {
1421 if (!netif_device_present(dev)) {
1426 u_map = nla_data(tb[IFLA_MAP]);
1427 k_map.mem_start = (unsigned long) u_map->mem_start;
1428 k_map.mem_end = (unsigned long) u_map->mem_end;
1429 k_map.base_addr = (unsigned short) u_map->base_addr;
1430 k_map.irq = (unsigned char) u_map->irq;
1431 k_map.dma = (unsigned char) u_map->dma;
1432 k_map.port = (unsigned char) u_map->port;
1434 err = ops->ndo_set_config(dev, &k_map);
1441 if (tb[IFLA_ADDRESS]) {
1442 struct sockaddr *sa;
1445 len = sizeof(sa_family_t) + dev->addr_len;
1446 sa = kmalloc(len, GFP_KERNEL);
1451 sa->sa_family = dev->type;
1452 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1454 err = dev_set_mac_address(dev, sa);
1462 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1468 if (tb[IFLA_GROUP]) {
1469 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1474 * Interface selected by interface index but interface
1475 * name provided implies that a name change has been
1478 if (ifm->ifi_index > 0 && ifname[0]) {
1479 err = dev_change_name(dev, ifname);
1485 if (tb[IFLA_IFALIAS]) {
1486 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
1487 nla_len(tb[IFLA_IFALIAS]));
1493 if (tb[IFLA_BROADCAST]) {
1494 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
1495 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
1498 if (ifm->ifi_flags || ifm->ifi_change) {
1499 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1504 if (tb[IFLA_MASTER]) {
1505 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
1511 if (tb[IFLA_CARRIER]) {
1512 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
1518 if (tb[IFLA_TXQLEN])
1519 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1521 if (tb[IFLA_OPERSTATE])
1522 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1524 if (tb[IFLA_LINKMODE]) {
1525 write_lock_bh(&dev_base_lock);
1526 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1527 write_unlock_bh(&dev_base_lock);
1530 if (tb[IFLA_VFINFO_LIST]) {
1531 struct nlattr *attr;
1533 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
1534 if (nla_type(attr) != IFLA_VF_INFO) {
1538 err = do_setvfinfo(dev, attr);
1546 if (tb[IFLA_VF_PORTS]) {
1547 struct nlattr *port[IFLA_PORT_MAX+1];
1548 struct nlattr *attr;
1553 if (!ops->ndo_set_vf_port)
1556 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
1557 if (nla_type(attr) != IFLA_VF_PORT)
1559 err = nla_parse_nested(port, IFLA_PORT_MAX,
1560 attr, ifla_port_policy);
1563 if (!port[IFLA_PORT_VF]) {
1567 vf = nla_get_u32(port[IFLA_PORT_VF]);
1568 err = ops->ndo_set_vf_port(dev, vf, port);
1576 if (tb[IFLA_PORT_SELF]) {
1577 struct nlattr *port[IFLA_PORT_MAX+1];
1579 err = nla_parse_nested(port, IFLA_PORT_MAX,
1580 tb[IFLA_PORT_SELF], ifla_port_policy);
1585 if (ops->ndo_set_vf_port)
1586 err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
1592 if (tb[IFLA_AF_SPEC]) {
1596 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1597 const struct rtnl_af_ops *af_ops;
1599 if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1602 err = af_ops->set_link_af(dev, af);
1612 if (err < 0 && modified)
1613 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",
1619 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
1621 struct net *net = sock_net(skb->sk);
1622 struct ifinfomsg *ifm;
1623 struct net_device *dev;
1625 struct nlattr *tb[IFLA_MAX+1];
1626 char ifname[IFNAMSIZ];
1628 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1632 if (tb[IFLA_IFNAME])
1633 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1638 ifm = nlmsg_data(nlh);
1639 if (ifm->ifi_index > 0)
1640 dev = __dev_get_by_index(net, ifm->ifi_index);
1641 else if (tb[IFLA_IFNAME])
1642 dev = __dev_get_by_name(net, ifname);
1651 err = validate_linkmsg(dev, tb);
1655 err = do_setlink(dev, ifm, tb, ifname, 0);
1660 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
1662 struct net *net = sock_net(skb->sk);
1663 const struct rtnl_link_ops *ops;
1664 struct net_device *dev;
1665 struct ifinfomsg *ifm;
1666 char ifname[IFNAMSIZ];
1667 struct nlattr *tb[IFLA_MAX+1];
1669 LIST_HEAD(list_kill);
1671 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1675 if (tb[IFLA_IFNAME])
1676 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1678 ifm = nlmsg_data(nlh);
1679 if (ifm->ifi_index > 0)
1680 dev = __dev_get_by_index(net, ifm->ifi_index);
1681 else if (tb[IFLA_IFNAME])
1682 dev = __dev_get_by_name(net, ifname);
1689 ops = dev->rtnl_link_ops;
1693 ops->dellink(dev, &list_kill);
1694 unregister_netdevice_many(&list_kill);
1695 list_del(&list_kill);
1699 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
1701 unsigned int old_flags;
1704 old_flags = dev->flags;
1705 if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
1706 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1711 dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
1713 __dev_notify_flags(dev, old_flags, ~0U);
1716 EXPORT_SYMBOL(rtnl_configure_link);
1718 struct net_device *rtnl_create_link(struct net *net,
1719 char *ifname, const struct rtnl_link_ops *ops, struct nlattr *tb[])
1722 struct net_device *dev;
1723 unsigned int num_tx_queues = 1;
1724 unsigned int num_rx_queues = 1;
1726 if (tb[IFLA_NUM_TX_QUEUES])
1727 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
1728 else if (ops->get_num_tx_queues)
1729 num_tx_queues = ops->get_num_tx_queues();
1731 if (tb[IFLA_NUM_RX_QUEUES])
1732 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
1733 else if (ops->get_num_rx_queues)
1734 num_rx_queues = ops->get_num_rx_queues();
1737 dev = alloc_netdev_mqs(ops->priv_size, ifname, ops->setup,
1738 num_tx_queues, num_rx_queues);
1742 dev_net_set(dev, net);
1743 dev->rtnl_link_ops = ops;
1744 dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
1747 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
1748 if (tb[IFLA_ADDRESS]) {
1749 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
1750 nla_len(tb[IFLA_ADDRESS]));
1751 dev->addr_assign_type = NET_ADDR_SET;
1753 if (tb[IFLA_BROADCAST])
1754 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
1755 nla_len(tb[IFLA_BROADCAST]));
1756 if (tb[IFLA_TXQLEN])
1757 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1758 if (tb[IFLA_OPERSTATE])
1759 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1760 if (tb[IFLA_LINKMODE])
1761 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1763 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1768 return ERR_PTR(err);
1770 EXPORT_SYMBOL(rtnl_create_link);
1772 static int rtnl_group_changelink(struct net *net, int group,
1773 struct ifinfomsg *ifm,
1776 struct net_device *dev;
1779 for_each_netdev(net, dev) {
1780 if (dev->group == group) {
1781 err = do_setlink(dev, ifm, tb, NULL, 0);
1790 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh)
1792 struct net *net = sock_net(skb->sk);
1793 const struct rtnl_link_ops *ops;
1794 const struct rtnl_link_ops *m_ops = NULL;
1795 struct net_device *dev;
1796 struct net_device *master_dev = NULL;
1797 struct ifinfomsg *ifm;
1798 char kind[MODULE_NAME_LEN];
1799 char ifname[IFNAMSIZ];
1800 struct nlattr *tb[IFLA_MAX+1];
1801 struct nlattr *linkinfo[IFLA_INFO_MAX+1];
1804 #ifdef CONFIG_MODULES
1807 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1811 if (tb[IFLA_IFNAME])
1812 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1816 ifm = nlmsg_data(nlh);
1817 if (ifm->ifi_index > 0)
1818 dev = __dev_get_by_index(net, ifm->ifi_index);
1821 dev = __dev_get_by_name(net, ifname);
1827 master_dev = netdev_master_upper_dev_get(dev);
1829 m_ops = master_dev->rtnl_link_ops;
1832 err = validate_linkmsg(dev, tb);
1836 if (tb[IFLA_LINKINFO]) {
1837 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
1838 tb[IFLA_LINKINFO], ifla_info_policy);
1842 memset(linkinfo, 0, sizeof(linkinfo));
1844 if (linkinfo[IFLA_INFO_KIND]) {
1845 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
1846 ops = rtnl_link_ops_get(kind);
1853 struct nlattr *attr[ops ? ops->maxtype + 1 : 0];
1854 struct nlattr *slave_attr[m_ops ? m_ops->slave_maxtype + 1 : 0];
1855 struct nlattr **data = NULL;
1856 struct nlattr **slave_data = NULL;
1857 struct net *dest_net;
1860 if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
1861 err = nla_parse_nested(attr, ops->maxtype,
1862 linkinfo[IFLA_INFO_DATA],
1868 if (ops->validate) {
1869 err = ops->validate(tb, data);
1876 if (m_ops->slave_maxtype &&
1877 linkinfo[IFLA_INFO_SLAVE_DATA]) {
1878 err = nla_parse_nested(slave_attr,
1879 m_ops->slave_maxtype,
1880 linkinfo[IFLA_INFO_SLAVE_DATA],
1881 m_ops->slave_policy);
1884 slave_data = slave_attr;
1886 if (m_ops->slave_validate) {
1887 err = m_ops->slave_validate(tb, slave_data);
1896 if (nlh->nlmsg_flags & NLM_F_EXCL)
1898 if (nlh->nlmsg_flags & NLM_F_REPLACE)
1901 if (linkinfo[IFLA_INFO_DATA]) {
1902 if (!ops || ops != dev->rtnl_link_ops ||
1906 err = ops->changelink(dev, tb, data);
1912 if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
1913 if (!m_ops || !m_ops->slave_changelink)
1916 err = m_ops->slave_changelink(master_dev, dev,
1923 return do_setlink(dev, ifm, tb, ifname, modified);
1926 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
1927 if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
1928 return rtnl_group_changelink(net,
1929 nla_get_u32(tb[IFLA_GROUP]),
1934 if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
1938 #ifdef CONFIG_MODULES
1941 request_module("rtnl-link-%s", kind);
1943 ops = rtnl_link_ops_get(kind);
1952 snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
1954 dest_net = rtnl_link_get_net(net, tb);
1955 if (IS_ERR(dest_net))
1956 return PTR_ERR(dest_net);
1958 dev = rtnl_create_link(dest_net, ifname, ops, tb);
1964 dev->ifindex = ifm->ifi_index;
1967 err = ops->newlink(net, dev, tb, data);
1968 /* Drivers should call free_netdev() in ->destructor
1969 * and unregister it on failure so that device could be
1970 * finally freed in rtnl_unlock.
1975 err = register_netdevice(dev);
1981 err = rtnl_configure_link(dev, ifm);
1983 unregister_netdevice(dev);
1990 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh)
1992 struct net *net = sock_net(skb->sk);
1993 struct ifinfomsg *ifm;
1994 char ifname[IFNAMSIZ];
1995 struct nlattr *tb[IFLA_MAX+1];
1996 struct net_device *dev = NULL;
1997 struct sk_buff *nskb;
1999 u32 ext_filter_mask = 0;
2001 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
2005 if (tb[IFLA_IFNAME])
2006 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2008 if (tb[IFLA_EXT_MASK])
2009 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2011 ifm = nlmsg_data(nlh);
2012 if (ifm->ifi_index > 0)
2013 dev = __dev_get_by_index(net, ifm->ifi_index);
2014 else if (tb[IFLA_IFNAME])
2015 dev = __dev_get_by_name(net, ifname);
2022 nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
2026 err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).portid,
2027 nlh->nlmsg_seq, 0, 0, ext_filter_mask);
2029 /* -EMSGSIZE implies BUG in if_nlmsg_size */
2030 WARN_ON(err == -EMSGSIZE);
2033 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
2038 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
2040 struct net *net = sock_net(skb->sk);
2041 struct net_device *dev;
2042 struct nlattr *tb[IFLA_MAX+1];
2043 u32 ext_filter_mask = 0;
2044 u16 min_ifinfo_dump_size = 0;
2046 if (nlmsg_parse(nlh, sizeof(struct ifinfomsg), tb, IFLA_MAX,
2047 ifla_policy) >= 0) {
2048 if (tb[IFLA_EXT_MASK])
2049 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2052 if (!ext_filter_mask)
2053 return NLMSG_GOODSIZE;
2055 * traverse the list of net devices and compute the minimum
2056 * buffer size based upon the filter mask.
2058 list_for_each_entry(dev, &net->dev_base_head, dev_list) {
2059 min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
2064 return min_ifinfo_dump_size;
2067 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
2070 int s_idx = cb->family;
2074 for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
2075 int type = cb->nlh->nlmsg_type-RTM_BASE;
2076 if (idx < s_idx || idx == PF_PACKET)
2078 if (rtnl_msg_handlers[idx] == NULL ||
2079 rtnl_msg_handlers[idx][type].dumpit == NULL)
2082 memset(&cb->args[0], 0, sizeof(cb->args));
2086 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
2094 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
2097 struct net *net = dev_net(dev);
2098 struct sk_buff *skb;
2100 size_t if_info_size;
2102 skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
2106 err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0);
2108 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
2109 WARN_ON(err == -EMSGSIZE);
2113 rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
2117 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2119 EXPORT_SYMBOL(rtmsg_ifinfo);
2121 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
2122 struct net_device *dev,
2123 u8 *addr, u32 pid, u32 seq,
2124 int type, unsigned int flags)
2126 struct nlmsghdr *nlh;
2129 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), NLM_F_MULTI);
2133 ndm = nlmsg_data(nlh);
2134 ndm->ndm_family = AF_BRIDGE;
2137 ndm->ndm_flags = flags;
2139 ndm->ndm_ifindex = dev->ifindex;
2140 ndm->ndm_state = NUD_PERMANENT;
2142 if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
2143 goto nla_put_failure;
2145 return nlmsg_end(skb, nlh);
2148 nlmsg_cancel(skb, nlh);
2152 static inline size_t rtnl_fdb_nlmsg_size(void)
2154 return NLMSG_ALIGN(sizeof(struct ndmsg)) + nla_total_size(ETH_ALEN);
2157 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, int type)
2159 struct net *net = dev_net(dev);
2160 struct sk_buff *skb;
2163 skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
2167 err = nlmsg_populate_fdb_fill(skb, dev, addr, 0, 0, type, NTF_SELF);
2173 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2176 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
2180 * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
2182 int ndo_dflt_fdb_add(struct ndmsg *ndm,
2183 struct nlattr *tb[],
2184 struct net_device *dev,
2185 const unsigned char *addr,
2190 /* If aging addresses are supported device will need to
2191 * implement its own handler for this.
2193 if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
2194 pr_info("%s: FDB only supports static addresses\n", dev->name);
2198 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2199 err = dev_uc_add_excl(dev, addr);
2200 else if (is_multicast_ether_addr(addr))
2201 err = dev_mc_add_excl(dev, addr);
2203 /* Only return duplicate errors if NLM_F_EXCL is set */
2204 if (err == -EEXIST && !(flags & NLM_F_EXCL))
2209 EXPORT_SYMBOL(ndo_dflt_fdb_add);
2211 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh)
2213 struct net *net = sock_net(skb->sk);
2215 struct nlattr *tb[NDA_MAX+1];
2216 struct net_device *dev;
2220 err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2224 ndm = nlmsg_data(nlh);
2225 if (ndm->ndm_ifindex == 0) {
2226 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n");
2230 dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2232 pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n");
2236 if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2237 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n");
2241 addr = nla_data(tb[NDA_LLADDR]);
2245 /* Support fdb on master device the net/bridge default case */
2246 if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2247 (dev->priv_flags & IFF_BRIDGE_PORT)) {
2248 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2249 const struct net_device_ops *ops = br_dev->netdev_ops;
2251 err = ops->ndo_fdb_add(ndm, tb, dev, addr, nlh->nlmsg_flags);
2255 ndm->ndm_flags &= ~NTF_MASTER;
2258 /* Embedded bridge, macvlan, and any other device support */
2259 if ((ndm->ndm_flags & NTF_SELF)) {
2260 if (dev->netdev_ops->ndo_fdb_add)
2261 err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
2264 err = ndo_dflt_fdb_add(ndm, tb, dev, addr,
2268 rtnl_fdb_notify(dev, addr, RTM_NEWNEIGH);
2269 ndm->ndm_flags &= ~NTF_SELF;
2277 * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
2279 int ndo_dflt_fdb_del(struct ndmsg *ndm,
2280 struct nlattr *tb[],
2281 struct net_device *dev,
2282 const unsigned char *addr)
2284 int err = -EOPNOTSUPP;
2286 /* If aging addresses are supported device will need to
2287 * implement its own handler for this.
2289 if (!(ndm->ndm_state & NUD_PERMANENT)) {
2290 pr_info("%s: FDB only supports static addresses\n", dev->name);
2294 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2295 err = dev_uc_del(dev, addr);
2296 else if (is_multicast_ether_addr(addr))
2297 err = dev_mc_del(dev, addr);
2303 EXPORT_SYMBOL(ndo_dflt_fdb_del);
2305 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh)
2307 struct net *net = sock_net(skb->sk);
2309 struct nlattr *tb[NDA_MAX+1];
2310 struct net_device *dev;
2314 if (!capable(CAP_NET_ADMIN))
2317 err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2321 ndm = nlmsg_data(nlh);
2322 if (ndm->ndm_ifindex == 0) {
2323 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n");
2327 dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2329 pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n");
2333 if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2334 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid address\n");
2338 addr = nla_data(tb[NDA_LLADDR]);
2342 /* Support fdb on master device the net/bridge default case */
2343 if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2344 (dev->priv_flags & IFF_BRIDGE_PORT)) {
2345 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2346 const struct net_device_ops *ops = br_dev->netdev_ops;
2348 if (ops->ndo_fdb_del)
2349 err = ops->ndo_fdb_del(ndm, tb, dev, addr);
2354 ndm->ndm_flags &= ~NTF_MASTER;
2357 /* Embedded bridge, macvlan, and any other device support */
2358 if (ndm->ndm_flags & NTF_SELF) {
2359 if (dev->netdev_ops->ndo_fdb_del)
2360 err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr);
2362 err = ndo_dflt_fdb_del(ndm, tb, dev, addr);
2365 rtnl_fdb_notify(dev, addr, RTM_DELNEIGH);
2366 ndm->ndm_flags &= ~NTF_SELF;
2373 static int nlmsg_populate_fdb(struct sk_buff *skb,
2374 struct netlink_callback *cb,
2375 struct net_device *dev,
2377 struct netdev_hw_addr_list *list)
2379 struct netdev_hw_addr *ha;
2383 portid = NETLINK_CB(cb->skb).portid;
2384 seq = cb->nlh->nlmsg_seq;
2386 list_for_each_entry(ha, &list->list, list) {
2387 if (*idx < cb->args[0])
2390 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr,
2392 RTM_NEWNEIGH, NTF_SELF);
2402 * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
2403 * @nlh: netlink message header
2406 * Default netdevice operation to dump the existing unicast address list.
2407 * Returns number of addresses from list put in skb.
2409 int ndo_dflt_fdb_dump(struct sk_buff *skb,
2410 struct netlink_callback *cb,
2411 struct net_device *dev,
2416 netif_addr_lock_bh(dev);
2417 err = nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->uc);
2420 nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->mc);
2422 netif_addr_unlock_bh(dev);
2425 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
2427 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
2430 struct net *net = sock_net(skb->sk);
2431 struct net_device *dev;
2434 for_each_netdev_rcu(net, dev) {
2435 if (dev->priv_flags & IFF_BRIDGE_PORT) {
2436 struct net_device *br_dev;
2437 const struct net_device_ops *ops;
2439 br_dev = netdev_master_upper_dev_get(dev);
2440 ops = br_dev->netdev_ops;
2441 if (ops->ndo_fdb_dump)
2442 idx = ops->ndo_fdb_dump(skb, cb, dev, idx);
2445 if (dev->netdev_ops->ndo_fdb_dump)
2446 idx = dev->netdev_ops->ndo_fdb_dump(skb, cb, dev, idx);
2448 idx = ndo_dflt_fdb_dump(skb, cb, dev, idx);
2456 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
2457 struct net_device *dev, u16 mode)
2459 struct nlmsghdr *nlh;
2460 struct ifinfomsg *ifm;
2461 struct nlattr *br_afspec;
2462 u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
2463 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2465 nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), NLM_F_MULTI);
2469 ifm = nlmsg_data(nlh);
2470 ifm->ifi_family = AF_BRIDGE;
2472 ifm->ifi_type = dev->type;
2473 ifm->ifi_index = dev->ifindex;
2474 ifm->ifi_flags = dev_get_flags(dev);
2475 ifm->ifi_change = 0;
2478 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
2479 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
2480 nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
2482 nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
2484 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
2485 (dev->ifindex != dev->iflink &&
2486 nla_put_u32(skb, IFLA_LINK, dev->iflink)))
2487 goto nla_put_failure;
2489 br_afspec = nla_nest_start(skb, IFLA_AF_SPEC);
2491 goto nla_put_failure;
2493 if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF) ||
2494 nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
2495 nla_nest_cancel(skb, br_afspec);
2496 goto nla_put_failure;
2498 nla_nest_end(skb, br_afspec);
2500 return nlmsg_end(skb, nlh);
2502 nlmsg_cancel(skb, nlh);
2505 EXPORT_SYMBOL(ndo_dflt_bridge_getlink);
2507 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
2509 struct net *net = sock_net(skb->sk);
2510 struct net_device *dev;
2512 u32 portid = NETLINK_CB(cb->skb).portid;
2513 u32 seq = cb->nlh->nlmsg_seq;
2514 struct nlattr *extfilt;
2515 u32 filter_mask = 0;
2517 extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct ifinfomsg),
2520 filter_mask = nla_get_u32(extfilt);
2523 for_each_netdev_rcu(net, dev) {
2524 const struct net_device_ops *ops = dev->netdev_ops;
2525 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2527 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
2528 if (idx >= cb->args[0] &&
2529 br_dev->netdev_ops->ndo_bridge_getlink(
2530 skb, portid, seq, dev, filter_mask) < 0)
2535 if (ops->ndo_bridge_getlink) {
2536 if (idx >= cb->args[0] &&
2537 ops->ndo_bridge_getlink(skb, portid, seq, dev,
2549 static inline size_t bridge_nlmsg_size(void)
2551 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
2552 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
2553 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
2554 + nla_total_size(sizeof(u32)) /* IFLA_MASTER */
2555 + nla_total_size(sizeof(u32)) /* IFLA_MTU */
2556 + nla_total_size(sizeof(u32)) /* IFLA_LINK */
2557 + nla_total_size(sizeof(u32)) /* IFLA_OPERSTATE */
2558 + nla_total_size(sizeof(u8)) /* IFLA_PROTINFO */
2559 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
2560 + nla_total_size(sizeof(u16)) /* IFLA_BRIDGE_FLAGS */
2561 + nla_total_size(sizeof(u16)); /* IFLA_BRIDGE_MODE */
2564 static int rtnl_bridge_notify(struct net_device *dev, u16 flags)
2566 struct net *net = dev_net(dev);
2567 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2568 struct sk_buff *skb;
2569 int err = -EOPNOTSUPP;
2571 skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
2577 if ((!flags || (flags & BRIDGE_FLAGS_MASTER)) &&
2578 br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
2579 err = br_dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0);
2584 if ((flags & BRIDGE_FLAGS_SELF) &&
2585 dev->netdev_ops->ndo_bridge_getlink) {
2586 err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0);
2591 rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
2594 WARN_ON(err == -EMSGSIZE);
2596 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2600 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
2602 struct net *net = sock_net(skb->sk);
2603 struct ifinfomsg *ifm;
2604 struct net_device *dev;
2605 struct nlattr *br_spec, *attr = NULL;
2606 int rem, err = -EOPNOTSUPP;
2607 u16 oflags, flags = 0;
2608 bool have_flags = false;
2610 if (nlmsg_len(nlh) < sizeof(*ifm))
2613 ifm = nlmsg_data(nlh);
2614 if (ifm->ifi_family != AF_BRIDGE)
2615 return -EPFNOSUPPORT;
2617 dev = __dev_get_by_index(net, ifm->ifi_index);
2619 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
2623 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
2625 nla_for_each_nested(attr, br_spec, rem) {
2626 if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
2628 flags = nla_get_u16(attr);
2636 if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
2637 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2639 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
2644 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh);
2648 flags &= ~BRIDGE_FLAGS_MASTER;
2651 if ((flags & BRIDGE_FLAGS_SELF)) {
2652 if (!dev->netdev_ops->ndo_bridge_setlink)
2655 err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh);
2658 flags &= ~BRIDGE_FLAGS_SELF;
2662 memcpy(nla_data(attr), &flags, sizeof(flags));
2663 /* Generate event to notify upper layer of bridge change */
2665 err = rtnl_bridge_notify(dev, oflags);
2670 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
2672 struct net *net = sock_net(skb->sk);
2673 struct ifinfomsg *ifm;
2674 struct net_device *dev;
2675 struct nlattr *br_spec, *attr = NULL;
2676 int rem, err = -EOPNOTSUPP;
2677 u16 oflags, flags = 0;
2678 bool have_flags = false;
2680 if (nlmsg_len(nlh) < sizeof(*ifm))
2683 ifm = nlmsg_data(nlh);
2684 if (ifm->ifi_family != AF_BRIDGE)
2685 return -EPFNOSUPPORT;
2687 dev = __dev_get_by_index(net, ifm->ifi_index);
2689 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
2693 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
2695 nla_for_each_nested(attr, br_spec, rem) {
2696 if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
2698 flags = nla_get_u16(attr);
2706 if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
2707 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2709 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
2714 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh);
2718 flags &= ~BRIDGE_FLAGS_MASTER;
2721 if ((flags & BRIDGE_FLAGS_SELF)) {
2722 if (!dev->netdev_ops->ndo_bridge_dellink)
2725 err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh);
2728 flags &= ~BRIDGE_FLAGS_SELF;
2732 memcpy(nla_data(attr), &flags, sizeof(flags));
2733 /* Generate event to notify upper layer of bridge change */
2735 err = rtnl_bridge_notify(dev, oflags);
2740 /* Process one rtnetlink message. */
2742 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
2744 struct net *net = sock_net(skb->sk);
2745 rtnl_doit_func doit;
2751 type = nlh->nlmsg_type;
2757 /* All the messages must have at least 1 byte length */
2758 if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
2761 family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
2765 if (kind != 2 && !ns_capable(net->user_ns, CAP_NET_ADMIN))
2768 if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
2770 rtnl_dumpit_func dumpit;
2771 rtnl_calcit_func calcit;
2772 u16 min_dump_alloc = 0;
2774 dumpit = rtnl_get_dumpit(family, type);
2777 calcit = rtnl_get_calcit(family, type);
2779 min_dump_alloc = calcit(skb, nlh);
2784 struct netlink_dump_control c = {
2786 .min_dump_alloc = min_dump_alloc,
2788 err = netlink_dump_start(rtnl, skb, nlh, &c);
2794 doit = rtnl_get_doit(family, type);
2798 return doit(skb, nlh);
2801 static void rtnetlink_rcv(struct sk_buff *skb)
2804 netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
2808 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
2810 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2816 case NETDEV_POST_INIT:
2817 case NETDEV_REGISTER:
2819 case NETDEV_PRE_TYPE_CHANGE:
2820 case NETDEV_GOING_DOWN:
2821 case NETDEV_UNREGISTER:
2822 case NETDEV_UNREGISTER_FINAL:
2823 case NETDEV_RELEASE:
2827 rtmsg_ifinfo(RTM_NEWLINK, dev, 0, GFP_KERNEL);
2833 static struct notifier_block rtnetlink_dev_notifier = {
2834 .notifier_call = rtnetlink_event,
2838 static int __net_init rtnetlink_net_init(struct net *net)
2841 struct netlink_kernel_cfg cfg = {
2842 .groups = RTNLGRP_MAX,
2843 .input = rtnetlink_rcv,
2844 .cb_mutex = &rtnl_mutex,
2845 .flags = NL_CFG_F_NONROOT_RECV,
2848 sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
2855 static void __net_exit rtnetlink_net_exit(struct net *net)
2857 netlink_kernel_release(net->rtnl);
2861 static struct pernet_operations rtnetlink_net_ops = {
2862 .init = rtnetlink_net_init,
2863 .exit = rtnetlink_net_exit,
2866 void __init rtnetlink_init(void)
2868 if (register_pernet_subsys(&rtnetlink_net_ops))
2869 panic("rtnetlink_init: cannot initialize rtnetlink\n");
2871 register_netdevice_notifier(&rtnetlink_dev_notifier);
2873 rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
2874 rtnl_dump_ifinfo, rtnl_calcit);
2875 rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
2876 rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
2877 rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
2879 rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
2880 rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
2882 rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, NULL);
2883 rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, NULL);
2884 rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, NULL);
2886 rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, NULL);
2887 rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, NULL);
2888 rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, NULL);