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_size(const struct net_device *dev)
370 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
376 size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
377 nla_total_size(strlen(ops->kind) + 1); /* IFLA_INFO_KIND */
380 /* IFLA_INFO_DATA + nested data */
381 size += nla_total_size(sizeof(struct nlattr)) +
384 if (ops->get_xstats_size)
385 /* IFLA_INFO_XSTATS */
386 size += nla_total_size(ops->get_xstats_size(dev));
391 static LIST_HEAD(rtnl_af_ops);
393 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
395 const struct rtnl_af_ops *ops;
397 list_for_each_entry(ops, &rtnl_af_ops, list) {
398 if (ops->family == family)
406 * __rtnl_af_register - Register rtnl_af_ops with rtnetlink.
407 * @ops: struct rtnl_af_ops * to register
409 * The caller must hold the rtnl_mutex.
411 * Returns 0 on success or a negative error code.
413 int __rtnl_af_register(struct rtnl_af_ops *ops)
415 list_add_tail(&ops->list, &rtnl_af_ops);
418 EXPORT_SYMBOL_GPL(__rtnl_af_register);
421 * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
422 * @ops: struct rtnl_af_ops * to register
424 * Returns 0 on success or a negative error code.
426 int rtnl_af_register(struct rtnl_af_ops *ops)
431 err = __rtnl_af_register(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 int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
482 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
483 struct nlattr *linkinfo, *data;
486 linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
487 if (linkinfo == NULL)
490 if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
491 goto err_cancel_link;
492 if (ops->fill_xstats) {
493 err = ops->fill_xstats(skb, dev);
495 goto err_cancel_link;
497 if (ops->fill_info) {
498 data = nla_nest_start(skb, IFLA_INFO_DATA);
500 goto err_cancel_link;
501 err = ops->fill_info(skb, dev);
503 goto err_cancel_data;
504 nla_nest_end(skb, data);
507 nla_nest_end(skb, linkinfo);
511 nla_nest_cancel(skb, data);
513 nla_nest_cancel(skb, linkinfo);
518 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
520 struct sock *rtnl = net->rtnl;
523 NETLINK_CB(skb).dst_group = group;
525 atomic_inc(&skb->users);
526 netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
528 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
532 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
534 struct sock *rtnl = net->rtnl;
536 return nlmsg_unicast(rtnl, skb, pid);
538 EXPORT_SYMBOL(rtnl_unicast);
540 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
541 struct nlmsghdr *nlh, gfp_t flags)
543 struct sock *rtnl = net->rtnl;
547 report = nlmsg_report(nlh);
549 nlmsg_notify(rtnl, skb, pid, group, report, flags);
551 EXPORT_SYMBOL(rtnl_notify);
553 void rtnl_set_sk_err(struct net *net, u32 group, int error)
555 struct sock *rtnl = net->rtnl;
557 netlink_set_err(rtnl, 0, group, error);
559 EXPORT_SYMBOL(rtnl_set_sk_err);
561 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
566 mx = nla_nest_start(skb, RTA_METRICS);
570 for (i = 0; i < RTAX_MAX; i++) {
573 if (nla_put_u32(skb, i+1, metrics[i]))
574 goto nla_put_failure;
579 nla_nest_cancel(skb, mx);
583 return nla_nest_end(skb, mx);
586 nla_nest_cancel(skb, mx);
589 EXPORT_SYMBOL(rtnetlink_put_metrics);
591 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
592 long expires, u32 error)
594 struct rta_cacheinfo ci = {
595 .rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse),
596 .rta_used = dst->__use,
597 .rta_clntref = atomic_read(&(dst->__refcnt)),
605 clock = jiffies_to_clock_t(abs(expires));
606 clock = min_t(unsigned long, clock, INT_MAX);
607 ci.rta_expires = (expires > 0) ? clock : -clock;
609 return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
611 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
613 static void set_operstate(struct net_device *dev, unsigned char transition)
615 unsigned char operstate = dev->operstate;
617 switch (transition) {
619 if ((operstate == IF_OPER_DORMANT ||
620 operstate == IF_OPER_UNKNOWN) &&
622 operstate = IF_OPER_UP;
625 case IF_OPER_DORMANT:
626 if (operstate == IF_OPER_UP ||
627 operstate == IF_OPER_UNKNOWN)
628 operstate = IF_OPER_DORMANT;
632 if (dev->operstate != operstate) {
633 write_lock_bh(&dev_base_lock);
634 dev->operstate = operstate;
635 write_unlock_bh(&dev_base_lock);
636 netdev_state_change(dev);
640 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
642 return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
643 (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
646 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
647 const struct ifinfomsg *ifm)
649 unsigned int flags = ifm->ifi_flags;
651 /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
653 flags = (flags & ifm->ifi_change) |
654 (rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
659 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
660 const struct rtnl_link_stats64 *b)
662 a->rx_packets = b->rx_packets;
663 a->tx_packets = b->tx_packets;
664 a->rx_bytes = b->rx_bytes;
665 a->tx_bytes = b->tx_bytes;
666 a->rx_errors = b->rx_errors;
667 a->tx_errors = b->tx_errors;
668 a->rx_dropped = b->rx_dropped;
669 a->tx_dropped = b->tx_dropped;
671 a->multicast = b->multicast;
672 a->collisions = b->collisions;
674 a->rx_length_errors = b->rx_length_errors;
675 a->rx_over_errors = b->rx_over_errors;
676 a->rx_crc_errors = b->rx_crc_errors;
677 a->rx_frame_errors = b->rx_frame_errors;
678 a->rx_fifo_errors = b->rx_fifo_errors;
679 a->rx_missed_errors = b->rx_missed_errors;
681 a->tx_aborted_errors = b->tx_aborted_errors;
682 a->tx_carrier_errors = b->tx_carrier_errors;
683 a->tx_fifo_errors = b->tx_fifo_errors;
684 a->tx_heartbeat_errors = b->tx_heartbeat_errors;
685 a->tx_window_errors = b->tx_window_errors;
687 a->rx_compressed = b->rx_compressed;
688 a->tx_compressed = b->tx_compressed;
691 static void copy_rtnl_link_stats64(void *v, const struct rtnl_link_stats64 *b)
693 memcpy(v, b, sizeof(*b));
697 static inline int rtnl_vfinfo_size(const struct net_device *dev,
700 if (dev->dev.parent && dev_is_pci(dev->dev.parent) &&
701 (ext_filter_mask & RTEXT_FILTER_VF)) {
702 int num_vfs = dev_num_vf(dev->dev.parent);
703 size_t size = nla_total_size(sizeof(struct nlattr));
704 size += nla_total_size(num_vfs * sizeof(struct nlattr));
706 (nla_total_size(sizeof(struct ifla_vf_mac)) +
707 nla_total_size(sizeof(struct ifla_vf_vlan)) +
708 nla_total_size(sizeof(struct ifla_vf_tx_rate)) +
709 nla_total_size(sizeof(struct ifla_vf_spoofchk)));
715 static size_t rtnl_port_size(const struct net_device *dev)
717 size_t port_size = nla_total_size(4) /* PORT_VF */
718 + nla_total_size(PORT_PROFILE_MAX) /* PORT_PROFILE */
719 + nla_total_size(sizeof(struct ifla_port_vsi))
721 + nla_total_size(PORT_UUID_MAX) /* PORT_INSTANCE_UUID */
722 + nla_total_size(PORT_UUID_MAX) /* PORT_HOST_UUID */
723 + nla_total_size(1) /* PROT_VDP_REQUEST */
724 + nla_total_size(2); /* PORT_VDP_RESPONSE */
725 size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
726 size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
728 size_t port_self_size = nla_total_size(sizeof(struct nlattr))
731 if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent)
733 if (dev_num_vf(dev->dev.parent))
734 return port_self_size + vf_ports_size +
735 vf_port_size * dev_num_vf(dev->dev.parent);
737 return port_self_size;
740 static noinline size_t if_nlmsg_size(const struct net_device *dev,
743 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
744 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
745 + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
746 + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
747 + nla_total_size(sizeof(struct rtnl_link_ifmap))
748 + nla_total_size(sizeof(struct rtnl_link_stats))
749 + nla_total_size(sizeof(struct rtnl_link_stats64))
750 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
751 + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
752 + nla_total_size(4) /* IFLA_TXQLEN */
753 + nla_total_size(4) /* IFLA_WEIGHT */
754 + nla_total_size(4) /* IFLA_MTU */
755 + nla_total_size(4) /* IFLA_LINK */
756 + nla_total_size(4) /* IFLA_MASTER */
757 + nla_total_size(1) /* IFLA_CARRIER */
758 + nla_total_size(4) /* IFLA_PROMISCUITY */
759 + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
760 + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
761 + nla_total_size(1) /* IFLA_OPERSTATE */
762 + nla_total_size(1) /* IFLA_LINKMODE */
763 + nla_total_size(ext_filter_mask
764 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
765 + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
766 + rtnl_port_size(dev) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
767 + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
768 + rtnl_link_get_af_size(dev); /* IFLA_AF_SPEC */
771 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
773 struct nlattr *vf_ports;
774 struct nlattr *vf_port;
778 vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
782 for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
783 vf_port = nla_nest_start(skb, IFLA_VF_PORT);
785 goto nla_put_failure;
786 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
787 goto nla_put_failure;
788 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
789 if (err == -EMSGSIZE)
790 goto nla_put_failure;
792 nla_nest_cancel(skb, vf_port);
795 nla_nest_end(skb, vf_port);
798 nla_nest_end(skb, vf_ports);
803 nla_nest_cancel(skb, vf_ports);
807 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
809 struct nlattr *port_self;
812 port_self = nla_nest_start(skb, IFLA_PORT_SELF);
816 err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
818 nla_nest_cancel(skb, port_self);
819 return (err == -EMSGSIZE) ? err : 0;
822 nla_nest_end(skb, port_self);
827 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev)
831 if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent)
834 err = rtnl_port_self_fill(skb, dev);
838 if (dev_num_vf(dev->dev.parent)) {
839 err = rtnl_vf_ports_fill(skb, dev);
847 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
848 int type, u32 pid, u32 seq, u32 change,
849 unsigned int flags, u32 ext_filter_mask)
851 struct ifinfomsg *ifm;
852 struct nlmsghdr *nlh;
853 struct rtnl_link_stats64 temp;
854 const struct rtnl_link_stats64 *stats;
855 struct nlattr *attr, *af_spec;
856 struct rtnl_af_ops *af_ops;
857 struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
860 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
864 ifm = nlmsg_data(nlh);
865 ifm->ifi_family = AF_UNSPEC;
867 ifm->ifi_type = dev->type;
868 ifm->ifi_index = dev->ifindex;
869 ifm->ifi_flags = dev_get_flags(dev);
870 ifm->ifi_change = change;
872 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
873 nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
874 nla_put_u8(skb, IFLA_OPERSTATE,
875 netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
876 nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
877 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
878 nla_put_u32(skb, IFLA_GROUP, dev->group) ||
879 nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
880 nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
882 nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
884 (dev->ifindex != dev->iflink &&
885 nla_put_u32(skb, IFLA_LINK, dev->iflink)) ||
887 nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex)) ||
888 nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
890 nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
892 nla_put_string(skb, IFLA_IFALIAS, dev->ifalias)))
893 goto nla_put_failure;
896 struct rtnl_link_ifmap map = {
897 .mem_start = dev->mem_start,
898 .mem_end = dev->mem_end,
899 .base_addr = dev->base_addr,
902 .port = dev->if_port,
904 if (nla_put(skb, IFLA_MAP, sizeof(map), &map))
905 goto nla_put_failure;
909 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
910 nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
911 goto nla_put_failure;
914 attr = nla_reserve(skb, IFLA_STATS,
915 sizeof(struct rtnl_link_stats));
917 goto nla_put_failure;
919 stats = dev_get_stats(dev, &temp);
920 copy_rtnl_link_stats(nla_data(attr), stats);
922 attr = nla_reserve(skb, IFLA_STATS64,
923 sizeof(struct rtnl_link_stats64));
925 goto nla_put_failure;
926 copy_rtnl_link_stats64(nla_data(attr), stats);
928 if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF) &&
929 nla_put_u32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent)))
930 goto nla_put_failure;
932 if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent
933 && (ext_filter_mask & RTEXT_FILTER_VF)) {
936 struct nlattr *vfinfo, *vf;
937 int num_vfs = dev_num_vf(dev->dev.parent);
939 vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
941 goto nla_put_failure;
942 for (i = 0; i < num_vfs; i++) {
943 struct ifla_vf_info ivi;
944 struct ifla_vf_mac vf_mac;
945 struct ifla_vf_vlan vf_vlan;
946 struct ifla_vf_tx_rate vf_tx_rate;
947 struct ifla_vf_spoofchk vf_spoofchk;
950 * Not all SR-IOV capable drivers support the
951 * spoofcheck query. Preset to -1 so the user
952 * space tool can detect that the driver didn't
956 memset(ivi.mac, 0, sizeof(ivi.mac));
957 if (dev->netdev_ops->ndo_get_vf_config(dev, i, &ivi))
962 vf_spoofchk.vf = ivi.vf;
964 memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
965 vf_vlan.vlan = ivi.vlan;
966 vf_vlan.qos = ivi.qos;
967 vf_tx_rate.rate = ivi.tx_rate;
968 vf_spoofchk.setting = ivi.spoofchk;
969 vf = nla_nest_start(skb, IFLA_VF_INFO);
971 nla_nest_cancel(skb, vfinfo);
972 goto nla_put_failure;
974 if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
975 nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
976 nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
978 nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
980 goto nla_put_failure;
981 nla_nest_end(skb, vf);
983 nla_nest_end(skb, vfinfo);
986 if (rtnl_port_fill(skb, dev))
987 goto nla_put_failure;
989 if (dev->rtnl_link_ops) {
990 if (rtnl_link_fill(skb, dev) < 0)
991 goto nla_put_failure;
994 if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
995 goto nla_put_failure;
997 list_for_each_entry(af_ops, &rtnl_af_ops, list) {
998 if (af_ops->fill_link_af) {
1002 if (!(af = nla_nest_start(skb, af_ops->family)))
1003 goto nla_put_failure;
1005 err = af_ops->fill_link_af(skb, dev);
1008 * Caller may return ENODATA to indicate that there
1009 * was no data to be dumped. This is not an error, it
1010 * means we should trim the attribute header and
1013 if (err == -ENODATA)
1014 nla_nest_cancel(skb, af);
1016 goto nla_put_failure;
1018 nla_nest_end(skb, af);
1022 nla_nest_end(skb, af_spec);
1024 return nlmsg_end(skb, nlh);
1027 nlmsg_cancel(skb, nlh);
1031 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1033 struct net *net = sock_net(skb->sk);
1036 struct net_device *dev;
1037 struct hlist_head *head;
1038 struct nlattr *tb[IFLA_MAX+1];
1039 u32 ext_filter_mask = 0;
1042 s_idx = cb->args[1];
1045 cb->seq = net->dev_base_seq;
1047 if (nlmsg_parse(cb->nlh, sizeof(struct rtgenmsg), tb, IFLA_MAX,
1048 ifla_policy) >= 0) {
1050 if (tb[IFLA_EXT_MASK])
1051 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1054 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1056 head = &net->dev_index_head[h];
1057 hlist_for_each_entry_rcu(dev, head, index_hlist) {
1060 if (rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1061 NETLINK_CB(cb->skb).portid,
1062 cb->nlh->nlmsg_seq, 0,
1064 ext_filter_mask) <= 0)
1067 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1080 const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1081 [IFLA_IFNAME] = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1082 [IFLA_ADDRESS] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1083 [IFLA_BROADCAST] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1084 [IFLA_MAP] = { .len = sizeof(struct rtnl_link_ifmap) },
1085 [IFLA_MTU] = { .type = NLA_U32 },
1086 [IFLA_LINK] = { .type = NLA_U32 },
1087 [IFLA_MASTER] = { .type = NLA_U32 },
1088 [IFLA_CARRIER] = { .type = NLA_U8 },
1089 [IFLA_TXQLEN] = { .type = NLA_U32 },
1090 [IFLA_WEIGHT] = { .type = NLA_U32 },
1091 [IFLA_OPERSTATE] = { .type = NLA_U8 },
1092 [IFLA_LINKMODE] = { .type = NLA_U8 },
1093 [IFLA_LINKINFO] = { .type = NLA_NESTED },
1094 [IFLA_NET_NS_PID] = { .type = NLA_U32 },
1095 [IFLA_NET_NS_FD] = { .type = NLA_U32 },
1096 [IFLA_IFALIAS] = { .type = NLA_STRING, .len = IFALIASZ-1 },
1097 [IFLA_VFINFO_LIST] = {. type = NLA_NESTED },
1098 [IFLA_VF_PORTS] = { .type = NLA_NESTED },
1099 [IFLA_PORT_SELF] = { .type = NLA_NESTED },
1100 [IFLA_AF_SPEC] = { .type = NLA_NESTED },
1101 [IFLA_EXT_MASK] = { .type = NLA_U32 },
1102 [IFLA_PROMISCUITY] = { .type = NLA_U32 },
1103 [IFLA_NUM_TX_QUEUES] = { .type = NLA_U32 },
1104 [IFLA_NUM_RX_QUEUES] = { .type = NLA_U32 },
1106 EXPORT_SYMBOL(ifla_policy);
1108 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1109 [IFLA_INFO_KIND] = { .type = NLA_STRING },
1110 [IFLA_INFO_DATA] = { .type = NLA_NESTED },
1113 static const struct nla_policy ifla_vfinfo_policy[IFLA_VF_INFO_MAX+1] = {
1114 [IFLA_VF_INFO] = { .type = NLA_NESTED },
1117 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1118 [IFLA_VF_MAC] = { .type = NLA_BINARY,
1119 .len = sizeof(struct ifla_vf_mac) },
1120 [IFLA_VF_VLAN] = { .type = NLA_BINARY,
1121 .len = sizeof(struct ifla_vf_vlan) },
1122 [IFLA_VF_TX_RATE] = { .type = NLA_BINARY,
1123 .len = sizeof(struct ifla_vf_tx_rate) },
1124 [IFLA_VF_SPOOFCHK] = { .type = NLA_BINARY,
1125 .len = sizeof(struct ifla_vf_spoofchk) },
1128 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1129 [IFLA_PORT_VF] = { .type = NLA_U32 },
1130 [IFLA_PORT_PROFILE] = { .type = NLA_STRING,
1131 .len = PORT_PROFILE_MAX },
1132 [IFLA_PORT_VSI_TYPE] = { .type = NLA_BINARY,
1133 .len = sizeof(struct ifla_port_vsi)},
1134 [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1135 .len = PORT_UUID_MAX },
1136 [IFLA_PORT_HOST_UUID] = { .type = NLA_STRING,
1137 .len = PORT_UUID_MAX },
1138 [IFLA_PORT_REQUEST] = { .type = NLA_U8, },
1139 [IFLA_PORT_RESPONSE] = { .type = NLA_U16, },
1142 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1145 /* Examine the link attributes and figure out which
1146 * network namespace we are talking about.
1148 if (tb[IFLA_NET_NS_PID])
1149 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1150 else if (tb[IFLA_NET_NS_FD])
1151 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1153 net = get_net(src_net);
1156 EXPORT_SYMBOL(rtnl_link_get_net);
1158 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1161 if (tb[IFLA_ADDRESS] &&
1162 nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1165 if (tb[IFLA_BROADCAST] &&
1166 nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1170 if (tb[IFLA_AF_SPEC]) {
1174 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1175 const struct rtnl_af_ops *af_ops;
1177 if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1178 return -EAFNOSUPPORT;
1180 if (!af_ops->set_link_af)
1183 if (af_ops->validate_link_af) {
1184 err = af_ops->validate_link_af(dev, af);
1194 static int do_setvfinfo(struct net_device *dev, struct nlattr *attr)
1196 int rem, err = -EINVAL;
1198 const struct net_device_ops *ops = dev->netdev_ops;
1200 nla_for_each_nested(vf, attr, rem) {
1201 switch (nla_type(vf)) {
1203 struct ifla_vf_mac *ivm;
1206 if (ops->ndo_set_vf_mac)
1207 err = ops->ndo_set_vf_mac(dev, ivm->vf,
1211 case IFLA_VF_VLAN: {
1212 struct ifla_vf_vlan *ivv;
1215 if (ops->ndo_set_vf_vlan)
1216 err = ops->ndo_set_vf_vlan(dev, ivv->vf,
1221 case IFLA_VF_TX_RATE: {
1222 struct ifla_vf_tx_rate *ivt;
1225 if (ops->ndo_set_vf_tx_rate)
1226 err = ops->ndo_set_vf_tx_rate(dev, ivt->vf,
1230 case IFLA_VF_SPOOFCHK: {
1231 struct ifla_vf_spoofchk *ivs;
1234 if (ops->ndo_set_vf_spoofchk)
1235 err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1249 static int do_set_master(struct net_device *dev, int ifindex)
1251 struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1252 const struct net_device_ops *ops;
1256 if (upper_dev->ifindex == ifindex)
1258 ops = upper_dev->netdev_ops;
1259 if (ops->ndo_del_slave) {
1260 err = ops->ndo_del_slave(upper_dev, dev);
1269 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
1272 ops = upper_dev->netdev_ops;
1273 if (ops->ndo_add_slave) {
1274 err = ops->ndo_add_slave(upper_dev, dev);
1284 static int do_setlink(struct net_device *dev, struct ifinfomsg *ifm,
1285 struct nlattr **tb, char *ifname, int modified)
1287 const struct net_device_ops *ops = dev->netdev_ops;
1290 if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1291 struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1296 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) {
1300 err = dev_change_net_namespace(dev, net, ifname);
1308 struct rtnl_link_ifmap *u_map;
1311 if (!ops->ndo_set_config) {
1316 if (!netif_device_present(dev)) {
1321 u_map = nla_data(tb[IFLA_MAP]);
1322 k_map.mem_start = (unsigned long) u_map->mem_start;
1323 k_map.mem_end = (unsigned long) u_map->mem_end;
1324 k_map.base_addr = (unsigned short) u_map->base_addr;
1325 k_map.irq = (unsigned char) u_map->irq;
1326 k_map.dma = (unsigned char) u_map->dma;
1327 k_map.port = (unsigned char) u_map->port;
1329 err = ops->ndo_set_config(dev, &k_map);
1336 if (tb[IFLA_ADDRESS]) {
1337 struct sockaddr *sa;
1340 len = sizeof(sa_family_t) + dev->addr_len;
1341 sa = kmalloc(len, GFP_KERNEL);
1346 sa->sa_family = dev->type;
1347 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1349 err = dev_set_mac_address(dev, sa);
1357 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1363 if (tb[IFLA_GROUP]) {
1364 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1369 * Interface selected by interface index but interface
1370 * name provided implies that a name change has been
1373 if (ifm->ifi_index > 0 && ifname[0]) {
1374 err = dev_change_name(dev, ifname);
1380 if (tb[IFLA_IFALIAS]) {
1381 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
1382 nla_len(tb[IFLA_IFALIAS]));
1388 if (tb[IFLA_BROADCAST]) {
1389 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
1390 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
1393 if (ifm->ifi_flags || ifm->ifi_change) {
1394 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1399 if (tb[IFLA_MASTER]) {
1400 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
1406 if (tb[IFLA_CARRIER]) {
1407 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
1413 if (tb[IFLA_TXQLEN])
1414 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1416 if (tb[IFLA_OPERSTATE])
1417 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1419 if (tb[IFLA_LINKMODE]) {
1420 write_lock_bh(&dev_base_lock);
1421 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1422 write_unlock_bh(&dev_base_lock);
1425 if (tb[IFLA_VFINFO_LIST]) {
1426 struct nlattr *attr;
1428 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
1429 if (nla_type(attr) != IFLA_VF_INFO) {
1433 err = do_setvfinfo(dev, attr);
1441 if (tb[IFLA_VF_PORTS]) {
1442 struct nlattr *port[IFLA_PORT_MAX+1];
1443 struct nlattr *attr;
1448 if (!ops->ndo_set_vf_port)
1451 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
1452 if (nla_type(attr) != IFLA_VF_PORT)
1454 err = nla_parse_nested(port, IFLA_PORT_MAX,
1455 attr, ifla_port_policy);
1458 if (!port[IFLA_PORT_VF]) {
1462 vf = nla_get_u32(port[IFLA_PORT_VF]);
1463 err = ops->ndo_set_vf_port(dev, vf, port);
1471 if (tb[IFLA_PORT_SELF]) {
1472 struct nlattr *port[IFLA_PORT_MAX+1];
1474 err = nla_parse_nested(port, IFLA_PORT_MAX,
1475 tb[IFLA_PORT_SELF], ifla_port_policy);
1480 if (ops->ndo_set_vf_port)
1481 err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
1487 if (tb[IFLA_AF_SPEC]) {
1491 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1492 const struct rtnl_af_ops *af_ops;
1494 if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1497 err = af_ops->set_link_af(dev, af);
1507 if (err < 0 && modified)
1508 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",
1514 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
1516 struct net *net = sock_net(skb->sk);
1517 struct ifinfomsg *ifm;
1518 struct net_device *dev;
1520 struct nlattr *tb[IFLA_MAX+1];
1521 char ifname[IFNAMSIZ];
1523 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1527 if (tb[IFLA_IFNAME])
1528 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1533 ifm = nlmsg_data(nlh);
1534 if (ifm->ifi_index > 0)
1535 dev = __dev_get_by_index(net, ifm->ifi_index);
1536 else if (tb[IFLA_IFNAME])
1537 dev = __dev_get_by_name(net, ifname);
1546 err = validate_linkmsg(dev, tb);
1550 err = do_setlink(dev, ifm, tb, ifname, 0);
1555 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
1557 struct net *net = sock_net(skb->sk);
1558 const struct rtnl_link_ops *ops;
1559 struct net_device *dev;
1560 struct ifinfomsg *ifm;
1561 char ifname[IFNAMSIZ];
1562 struct nlattr *tb[IFLA_MAX+1];
1564 LIST_HEAD(list_kill);
1566 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1570 if (tb[IFLA_IFNAME])
1571 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1573 ifm = nlmsg_data(nlh);
1574 if (ifm->ifi_index > 0)
1575 dev = __dev_get_by_index(net, ifm->ifi_index);
1576 else if (tb[IFLA_IFNAME])
1577 dev = __dev_get_by_name(net, ifname);
1584 ops = dev->rtnl_link_ops;
1588 ops->dellink(dev, &list_kill);
1589 unregister_netdevice_many(&list_kill);
1590 list_del(&list_kill);
1594 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
1596 unsigned int old_flags;
1599 old_flags = dev->flags;
1600 if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
1601 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1606 dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
1607 rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U);
1609 __dev_notify_flags(dev, old_flags);
1612 EXPORT_SYMBOL(rtnl_configure_link);
1614 struct net_device *rtnl_create_link(struct net *net,
1615 char *ifname, const struct rtnl_link_ops *ops, struct nlattr *tb[])
1618 struct net_device *dev;
1619 unsigned int num_tx_queues = 1;
1620 unsigned int num_rx_queues = 1;
1622 if (tb[IFLA_NUM_TX_QUEUES])
1623 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
1624 else if (ops->get_num_tx_queues)
1625 num_tx_queues = ops->get_num_tx_queues();
1627 if (tb[IFLA_NUM_RX_QUEUES])
1628 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
1629 else if (ops->get_num_rx_queues)
1630 num_rx_queues = ops->get_num_rx_queues();
1633 dev = alloc_netdev_mqs(ops->priv_size, ifname, ops->setup,
1634 num_tx_queues, num_rx_queues);
1638 dev_net_set(dev, net);
1639 dev->rtnl_link_ops = ops;
1640 dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
1643 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
1644 if (tb[IFLA_ADDRESS]) {
1645 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
1646 nla_len(tb[IFLA_ADDRESS]));
1647 dev->addr_assign_type = NET_ADDR_SET;
1649 if (tb[IFLA_BROADCAST])
1650 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
1651 nla_len(tb[IFLA_BROADCAST]));
1652 if (tb[IFLA_TXQLEN])
1653 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1654 if (tb[IFLA_OPERSTATE])
1655 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1656 if (tb[IFLA_LINKMODE])
1657 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1659 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1664 return ERR_PTR(err);
1666 EXPORT_SYMBOL(rtnl_create_link);
1668 static int rtnl_group_changelink(struct net *net, int group,
1669 struct ifinfomsg *ifm,
1672 struct net_device *dev;
1675 for_each_netdev(net, dev) {
1676 if (dev->group == group) {
1677 err = do_setlink(dev, ifm, tb, NULL, 0);
1686 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh)
1688 struct net *net = sock_net(skb->sk);
1689 const struct rtnl_link_ops *ops;
1690 struct net_device *dev;
1691 struct ifinfomsg *ifm;
1692 char kind[MODULE_NAME_LEN];
1693 char ifname[IFNAMSIZ];
1694 struct nlattr *tb[IFLA_MAX+1];
1695 struct nlattr *linkinfo[IFLA_INFO_MAX+1];
1698 #ifdef CONFIG_MODULES
1701 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1705 if (tb[IFLA_IFNAME])
1706 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1710 ifm = nlmsg_data(nlh);
1711 if (ifm->ifi_index > 0)
1712 dev = __dev_get_by_index(net, ifm->ifi_index);
1715 dev = __dev_get_by_name(net, ifname);
1720 err = validate_linkmsg(dev, tb);
1724 if (tb[IFLA_LINKINFO]) {
1725 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
1726 tb[IFLA_LINKINFO], ifla_info_policy);
1730 memset(linkinfo, 0, sizeof(linkinfo));
1732 if (linkinfo[IFLA_INFO_KIND]) {
1733 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
1734 ops = rtnl_link_ops_get(kind);
1741 struct nlattr *attr[ops ? ops->maxtype + 1 : 0], **data = NULL;
1742 struct net *dest_net;
1745 if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
1746 err = nla_parse_nested(attr, ops->maxtype,
1747 linkinfo[IFLA_INFO_DATA],
1753 if (ops->validate) {
1754 err = ops->validate(tb, data);
1763 if (nlh->nlmsg_flags & NLM_F_EXCL)
1765 if (nlh->nlmsg_flags & NLM_F_REPLACE)
1768 if (linkinfo[IFLA_INFO_DATA]) {
1769 if (!ops || ops != dev->rtnl_link_ops ||
1773 err = ops->changelink(dev, tb, data);
1779 return do_setlink(dev, ifm, tb, ifname, modified);
1782 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
1783 if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
1784 return rtnl_group_changelink(net,
1785 nla_get_u32(tb[IFLA_GROUP]),
1790 if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
1794 #ifdef CONFIG_MODULES
1797 request_module("rtnl-link-%s", kind);
1799 ops = rtnl_link_ops_get(kind);
1808 snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
1810 dest_net = rtnl_link_get_net(net, tb);
1811 if (IS_ERR(dest_net))
1812 return PTR_ERR(dest_net);
1814 dev = rtnl_create_link(dest_net, ifname, ops, tb);
1820 dev->ifindex = ifm->ifi_index;
1823 err = ops->newlink(net, dev, tb, data);
1825 err = register_netdevice(dev);
1827 if (err < 0 && !IS_ERR(dev))
1832 err = rtnl_configure_link(dev, ifm);
1834 unregister_netdevice(dev);
1841 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh)
1843 struct net *net = sock_net(skb->sk);
1844 struct ifinfomsg *ifm;
1845 char ifname[IFNAMSIZ];
1846 struct nlattr *tb[IFLA_MAX+1];
1847 struct net_device *dev = NULL;
1848 struct sk_buff *nskb;
1850 u32 ext_filter_mask = 0;
1852 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1856 if (tb[IFLA_IFNAME])
1857 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1859 if (tb[IFLA_EXT_MASK])
1860 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1862 ifm = nlmsg_data(nlh);
1863 if (ifm->ifi_index > 0)
1864 dev = __dev_get_by_index(net, ifm->ifi_index);
1865 else if (tb[IFLA_IFNAME])
1866 dev = __dev_get_by_name(net, ifname);
1873 nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
1877 err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).portid,
1878 nlh->nlmsg_seq, 0, 0, ext_filter_mask);
1880 /* -EMSGSIZE implies BUG in if_nlmsg_size */
1881 WARN_ON(err == -EMSGSIZE);
1884 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
1889 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
1891 struct net *net = sock_net(skb->sk);
1892 struct net_device *dev;
1893 struct nlattr *tb[IFLA_MAX+1];
1894 u32 ext_filter_mask = 0;
1895 u16 min_ifinfo_dump_size = 0;
1897 if (nlmsg_parse(nlh, sizeof(struct rtgenmsg), tb, IFLA_MAX,
1898 ifla_policy) >= 0) {
1899 if (tb[IFLA_EXT_MASK])
1900 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1903 if (!ext_filter_mask)
1904 return NLMSG_GOODSIZE;
1906 * traverse the list of net devices and compute the minimum
1907 * buffer size based upon the filter mask.
1909 list_for_each_entry(dev, &net->dev_base_head, dev_list) {
1910 min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
1915 return min_ifinfo_dump_size;
1918 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
1921 int s_idx = cb->family;
1925 for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
1926 int type = cb->nlh->nlmsg_type-RTM_BASE;
1927 if (idx < s_idx || idx == PF_PACKET)
1929 if (rtnl_msg_handlers[idx] == NULL ||
1930 rtnl_msg_handlers[idx][type].dumpit == NULL)
1933 memset(&cb->args[0], 0, sizeof(cb->args));
1937 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
1945 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change)
1947 struct net *net = dev_net(dev);
1948 struct sk_buff *skb;
1950 size_t if_info_size;
1952 skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), GFP_KERNEL);
1956 err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0);
1958 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
1959 WARN_ON(err == -EMSGSIZE);
1963 rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_KERNEL);
1967 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
1969 EXPORT_SYMBOL(rtmsg_ifinfo);
1971 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
1972 struct net_device *dev,
1973 u8 *addr, u32 pid, u32 seq,
1974 int type, unsigned int flags)
1976 struct nlmsghdr *nlh;
1979 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), NLM_F_MULTI);
1983 ndm = nlmsg_data(nlh);
1984 ndm->ndm_family = AF_BRIDGE;
1987 ndm->ndm_flags = flags;
1989 ndm->ndm_ifindex = dev->ifindex;
1990 ndm->ndm_state = NUD_PERMANENT;
1992 if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
1993 goto nla_put_failure;
1995 return nlmsg_end(skb, nlh);
1998 nlmsg_cancel(skb, nlh);
2002 static inline size_t rtnl_fdb_nlmsg_size(void)
2004 return NLMSG_ALIGN(sizeof(struct ndmsg)) + nla_total_size(ETH_ALEN);
2007 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, int type)
2009 struct net *net = dev_net(dev);
2010 struct sk_buff *skb;
2013 skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
2017 err = nlmsg_populate_fdb_fill(skb, dev, addr, 0, 0, type, NTF_SELF);
2023 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2026 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
2030 * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
2032 int ndo_dflt_fdb_add(struct ndmsg *ndm,
2033 struct nlattr *tb[],
2034 struct net_device *dev,
2035 const unsigned char *addr,
2040 /* If aging addresses are supported device will need to
2041 * implement its own handler for this.
2043 if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
2044 pr_info("%s: FDB only supports static addresses\n", dev->name);
2048 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2049 err = dev_uc_add_excl(dev, addr);
2050 else if (is_multicast_ether_addr(addr))
2051 err = dev_mc_add_excl(dev, addr);
2053 /* Only return duplicate errors if NLM_F_EXCL is set */
2054 if (err == -EEXIST && !(flags & NLM_F_EXCL))
2059 EXPORT_SYMBOL(ndo_dflt_fdb_add);
2061 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh)
2063 struct net *net = sock_net(skb->sk);
2065 struct nlattr *tb[NDA_MAX+1];
2066 struct net_device *dev;
2070 err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2074 ndm = nlmsg_data(nlh);
2075 if (ndm->ndm_ifindex == 0) {
2076 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n");
2080 dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2082 pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n");
2086 if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2087 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n");
2091 addr = nla_data(tb[NDA_LLADDR]);
2092 if (is_zero_ether_addr(addr)) {
2093 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ether address\n");
2099 /* Support fdb on master device the net/bridge default case */
2100 if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2101 (dev->priv_flags & IFF_BRIDGE_PORT)) {
2102 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2103 const struct net_device_ops *ops = br_dev->netdev_ops;
2105 err = ops->ndo_fdb_add(ndm, tb, dev, addr, nlh->nlmsg_flags);
2109 ndm->ndm_flags &= ~NTF_MASTER;
2112 /* Embedded bridge, macvlan, and any other device support */
2113 if ((ndm->ndm_flags & NTF_SELF)) {
2114 if (dev->netdev_ops->ndo_fdb_add)
2115 err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
2118 err = ndo_dflt_fdb_add(ndm, tb, dev, addr,
2122 rtnl_fdb_notify(dev, addr, RTM_NEWNEIGH);
2123 ndm->ndm_flags &= ~NTF_SELF;
2131 * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
2133 int ndo_dflt_fdb_del(struct ndmsg *ndm,
2134 struct nlattr *tb[],
2135 struct net_device *dev,
2136 const unsigned char *addr)
2138 int err = -EOPNOTSUPP;
2140 /* If aging addresses are supported device will need to
2141 * implement its own handler for this.
2143 if (ndm->ndm_state & NUD_PERMANENT) {
2144 pr_info("%s: FDB only supports static addresses\n", dev->name);
2148 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2149 err = dev_uc_del(dev, addr);
2150 else if (is_multicast_ether_addr(addr))
2151 err = dev_mc_del(dev, addr);
2157 EXPORT_SYMBOL(ndo_dflt_fdb_del);
2159 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh)
2161 struct net *net = sock_net(skb->sk);
2163 struct nlattr *tb[NDA_MAX+1];
2164 struct net_device *dev;
2168 if (!capable(CAP_NET_ADMIN))
2171 err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2175 ndm = nlmsg_data(nlh);
2176 if (ndm->ndm_ifindex == 0) {
2177 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n");
2181 dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2183 pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n");
2187 if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2188 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid address\n");
2192 addr = nla_data(tb[NDA_LLADDR]);
2193 if (!is_valid_ether_addr(addr)) {
2194 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ether address\n");
2200 /* Support fdb on master device the net/bridge default case */
2201 if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2202 (dev->priv_flags & IFF_BRIDGE_PORT)) {
2203 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2204 const struct net_device_ops *ops = br_dev->netdev_ops;
2206 if (ops->ndo_fdb_del)
2207 err = ops->ndo_fdb_del(ndm, tb, dev, addr);
2212 ndm->ndm_flags &= ~NTF_MASTER;
2215 /* Embedded bridge, macvlan, and any other device support */
2216 if (ndm->ndm_flags & NTF_SELF) {
2217 if (dev->netdev_ops->ndo_fdb_del)
2218 err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr);
2220 err = ndo_dflt_fdb_del(ndm, tb, dev, addr);
2223 rtnl_fdb_notify(dev, addr, RTM_DELNEIGH);
2224 ndm->ndm_flags &= ~NTF_SELF;
2231 static int nlmsg_populate_fdb(struct sk_buff *skb,
2232 struct netlink_callback *cb,
2233 struct net_device *dev,
2235 struct netdev_hw_addr_list *list)
2237 struct netdev_hw_addr *ha;
2241 portid = NETLINK_CB(cb->skb).portid;
2242 seq = cb->nlh->nlmsg_seq;
2244 list_for_each_entry(ha, &list->list, list) {
2245 if (*idx < cb->args[0])
2248 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr,
2250 RTM_NEWNEIGH, NTF_SELF);
2260 * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
2261 * @nlh: netlink message header
2264 * Default netdevice operation to dump the existing unicast address list.
2265 * Returns zero on success.
2267 int ndo_dflt_fdb_dump(struct sk_buff *skb,
2268 struct netlink_callback *cb,
2269 struct net_device *dev,
2274 netif_addr_lock_bh(dev);
2275 err = nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->uc);
2278 nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->mc);
2280 netif_addr_unlock_bh(dev);
2283 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
2285 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
2288 struct net *net = sock_net(skb->sk);
2289 struct net_device *dev;
2292 for_each_netdev_rcu(net, dev) {
2293 if (dev->priv_flags & IFF_BRIDGE_PORT) {
2294 struct net_device *br_dev;
2295 const struct net_device_ops *ops;
2297 br_dev = netdev_master_upper_dev_get(dev);
2298 ops = br_dev->netdev_ops;
2299 if (ops->ndo_fdb_dump)
2300 idx = ops->ndo_fdb_dump(skb, cb, dev, idx);
2303 if (dev->netdev_ops->ndo_fdb_dump)
2304 idx = dev->netdev_ops->ndo_fdb_dump(skb, cb, dev, idx);
2306 ndo_dflt_fdb_dump(skb, cb, dev, idx);
2314 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
2315 struct net_device *dev, u16 mode)
2317 struct nlmsghdr *nlh;
2318 struct ifinfomsg *ifm;
2319 struct nlattr *br_afspec;
2320 u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
2321 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2323 nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), NLM_F_MULTI);
2327 ifm = nlmsg_data(nlh);
2328 ifm->ifi_family = AF_BRIDGE;
2330 ifm->ifi_type = dev->type;
2331 ifm->ifi_index = dev->ifindex;
2332 ifm->ifi_flags = dev_get_flags(dev);
2333 ifm->ifi_change = 0;
2336 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
2337 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
2338 nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
2340 nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
2342 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
2343 (dev->ifindex != dev->iflink &&
2344 nla_put_u32(skb, IFLA_LINK, dev->iflink)))
2345 goto nla_put_failure;
2347 br_afspec = nla_nest_start(skb, IFLA_AF_SPEC);
2349 goto nla_put_failure;
2351 if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF) ||
2352 nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
2353 nla_nest_cancel(skb, br_afspec);
2354 goto nla_put_failure;
2356 nla_nest_end(skb, br_afspec);
2358 return nlmsg_end(skb, nlh);
2360 nlmsg_cancel(skb, nlh);
2363 EXPORT_SYMBOL(ndo_dflt_bridge_getlink);
2365 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
2367 struct net *net = sock_net(skb->sk);
2368 struct net_device *dev;
2370 u32 portid = NETLINK_CB(cb->skb).portid;
2371 u32 seq = cb->nlh->nlmsg_seq;
2372 struct nlattr *extfilt;
2373 u32 filter_mask = 0;
2375 extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct rtgenmsg),
2378 filter_mask = nla_get_u32(extfilt);
2381 for_each_netdev_rcu(net, dev) {
2382 const struct net_device_ops *ops = dev->netdev_ops;
2383 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2385 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
2386 if (idx >= cb->args[0] &&
2387 br_dev->netdev_ops->ndo_bridge_getlink(
2388 skb, portid, seq, dev, filter_mask) < 0)
2393 if (ops->ndo_bridge_getlink) {
2394 if (idx >= cb->args[0] &&
2395 ops->ndo_bridge_getlink(skb, portid, seq, dev,
2407 static inline size_t bridge_nlmsg_size(void)
2409 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
2410 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
2411 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
2412 + nla_total_size(sizeof(u32)) /* IFLA_MASTER */
2413 + nla_total_size(sizeof(u32)) /* IFLA_MTU */
2414 + nla_total_size(sizeof(u32)) /* IFLA_LINK */
2415 + nla_total_size(sizeof(u32)) /* IFLA_OPERSTATE */
2416 + nla_total_size(sizeof(u8)) /* IFLA_PROTINFO */
2417 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
2418 + nla_total_size(sizeof(u16)) /* IFLA_BRIDGE_FLAGS */
2419 + nla_total_size(sizeof(u16)); /* IFLA_BRIDGE_MODE */
2422 static int rtnl_bridge_notify(struct net_device *dev, u16 flags)
2424 struct net *net = dev_net(dev);
2425 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2426 struct sk_buff *skb;
2427 int err = -EOPNOTSUPP;
2429 skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
2435 if ((!flags || (flags & BRIDGE_FLAGS_MASTER)) &&
2436 br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
2437 err = br_dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0);
2442 if ((flags & BRIDGE_FLAGS_SELF) &&
2443 dev->netdev_ops->ndo_bridge_getlink) {
2444 err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0);
2449 rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
2452 WARN_ON(err == -EMSGSIZE);
2454 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2458 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
2460 struct net *net = sock_net(skb->sk);
2461 struct ifinfomsg *ifm;
2462 struct net_device *dev;
2463 struct nlattr *br_spec, *attr = NULL;
2464 int rem, err = -EOPNOTSUPP;
2465 u16 oflags, flags = 0;
2466 bool have_flags = false;
2468 if (nlmsg_len(nlh) < sizeof(*ifm))
2471 ifm = nlmsg_data(nlh);
2472 if (ifm->ifi_family != AF_BRIDGE)
2473 return -EPFNOSUPPORT;
2475 dev = __dev_get_by_index(net, ifm->ifi_index);
2477 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
2481 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
2483 nla_for_each_nested(attr, br_spec, rem) {
2484 if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
2486 flags = nla_get_u16(attr);
2494 if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
2495 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2497 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
2502 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh);
2506 flags &= ~BRIDGE_FLAGS_MASTER;
2509 if ((flags & BRIDGE_FLAGS_SELF)) {
2510 if (!dev->netdev_ops->ndo_bridge_setlink)
2513 err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh);
2516 flags &= ~BRIDGE_FLAGS_SELF;
2520 memcpy(nla_data(attr), &flags, sizeof(flags));
2521 /* Generate event to notify upper layer of bridge change */
2523 err = rtnl_bridge_notify(dev, oflags);
2528 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
2530 struct net *net = sock_net(skb->sk);
2531 struct ifinfomsg *ifm;
2532 struct net_device *dev;
2533 struct nlattr *br_spec, *attr = NULL;
2534 int rem, err = -EOPNOTSUPP;
2535 u16 oflags, flags = 0;
2536 bool have_flags = false;
2538 if (nlmsg_len(nlh) < sizeof(*ifm))
2541 ifm = nlmsg_data(nlh);
2542 if (ifm->ifi_family != AF_BRIDGE)
2543 return -EPFNOSUPPORT;
2545 dev = __dev_get_by_index(net, ifm->ifi_index);
2547 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
2551 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
2553 nla_for_each_nested(attr, br_spec, rem) {
2554 if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
2556 flags = nla_get_u16(attr);
2564 if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
2565 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2567 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
2572 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh);
2576 flags &= ~BRIDGE_FLAGS_MASTER;
2579 if ((flags & BRIDGE_FLAGS_SELF)) {
2580 if (!dev->netdev_ops->ndo_bridge_dellink)
2583 err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh);
2586 flags &= ~BRIDGE_FLAGS_SELF;
2590 memcpy(nla_data(attr), &flags, sizeof(flags));
2591 /* Generate event to notify upper layer of bridge change */
2593 err = rtnl_bridge_notify(dev, oflags);
2598 /* Process one rtnetlink message. */
2600 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
2602 struct net *net = sock_net(skb->sk);
2603 rtnl_doit_func doit;
2609 type = nlh->nlmsg_type;
2615 /* All the messages must have at least 1 byte length */
2616 if (nlh->nlmsg_len < NLMSG_LENGTH(sizeof(struct rtgenmsg)))
2619 family = ((struct rtgenmsg *)NLMSG_DATA(nlh))->rtgen_family;
2623 if (kind != 2 && !ns_capable(net->user_ns, CAP_NET_ADMIN))
2626 if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
2628 rtnl_dumpit_func dumpit;
2629 rtnl_calcit_func calcit;
2630 u16 min_dump_alloc = 0;
2632 dumpit = rtnl_get_dumpit(family, type);
2635 calcit = rtnl_get_calcit(family, type);
2637 min_dump_alloc = calcit(skb, nlh);
2642 struct netlink_dump_control c = {
2644 .min_dump_alloc = min_dump_alloc,
2646 err = netlink_dump_start(rtnl, skb, nlh, &c);
2652 doit = rtnl_get_doit(family, type);
2656 return doit(skb, nlh);
2659 static void rtnetlink_rcv(struct sk_buff *skb)
2662 netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
2666 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
2668 struct net_device *dev = ptr;
2674 case NETDEV_POST_INIT:
2675 case NETDEV_REGISTER:
2677 case NETDEV_PRE_TYPE_CHANGE:
2678 case NETDEV_GOING_DOWN:
2679 case NETDEV_UNREGISTER:
2680 case NETDEV_UNREGISTER_FINAL:
2681 case NETDEV_RELEASE:
2685 rtmsg_ifinfo(RTM_NEWLINK, dev, 0);
2691 static struct notifier_block rtnetlink_dev_notifier = {
2692 .notifier_call = rtnetlink_event,
2696 static int __net_init rtnetlink_net_init(struct net *net)
2699 struct netlink_kernel_cfg cfg = {
2700 .groups = RTNLGRP_MAX,
2701 .input = rtnetlink_rcv,
2702 .cb_mutex = &rtnl_mutex,
2703 .flags = NL_CFG_F_NONROOT_RECV,
2706 sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
2713 static void __net_exit rtnetlink_net_exit(struct net *net)
2715 netlink_kernel_release(net->rtnl);
2719 static struct pernet_operations rtnetlink_net_ops = {
2720 .init = rtnetlink_net_init,
2721 .exit = rtnetlink_net_exit,
2724 void __init rtnetlink_init(void)
2726 if (register_pernet_subsys(&rtnetlink_net_ops))
2727 panic("rtnetlink_init: cannot initialize rtnetlink\n");
2729 register_netdevice_notifier(&rtnetlink_dev_notifier);
2731 rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
2732 rtnl_dump_ifinfo, rtnl_calcit);
2733 rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
2734 rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
2735 rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
2737 rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
2738 rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
2740 rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, NULL);
2741 rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, NULL);
2742 rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, NULL);
2744 rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, NULL);
2745 rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, NULL);
2746 rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, NULL);