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);
501 goto err_cancel_link;
503 err = ops->fill_info(skb, dev);
505 goto err_cancel_data;
506 nla_nest_end(skb, data);
509 nla_nest_end(skb, linkinfo);
513 nla_nest_cancel(skb, data);
515 nla_nest_cancel(skb, linkinfo);
520 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
522 struct sock *rtnl = net->rtnl;
525 NETLINK_CB(skb).dst_group = group;
527 atomic_inc(&skb->users);
528 netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
530 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
534 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
536 struct sock *rtnl = net->rtnl;
538 return nlmsg_unicast(rtnl, skb, pid);
540 EXPORT_SYMBOL(rtnl_unicast);
542 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
543 struct nlmsghdr *nlh, gfp_t flags)
545 struct sock *rtnl = net->rtnl;
549 report = nlmsg_report(nlh);
551 nlmsg_notify(rtnl, skb, pid, group, report, flags);
553 EXPORT_SYMBOL(rtnl_notify);
555 void rtnl_set_sk_err(struct net *net, u32 group, int error)
557 struct sock *rtnl = net->rtnl;
559 netlink_set_err(rtnl, 0, group, error);
561 EXPORT_SYMBOL(rtnl_set_sk_err);
563 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
568 mx = nla_nest_start(skb, RTA_METRICS);
572 for (i = 0; i < RTAX_MAX; i++) {
575 if (nla_put_u32(skb, i+1, metrics[i]))
576 goto nla_put_failure;
581 nla_nest_cancel(skb, mx);
585 return nla_nest_end(skb, mx);
588 nla_nest_cancel(skb, mx);
591 EXPORT_SYMBOL(rtnetlink_put_metrics);
593 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
594 long expires, u32 error)
596 struct rta_cacheinfo ci = {
597 .rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse),
598 .rta_used = dst->__use,
599 .rta_clntref = atomic_read(&(dst->__refcnt)),
607 clock = jiffies_to_clock_t(abs(expires));
608 clock = min_t(unsigned long, clock, INT_MAX);
609 ci.rta_expires = (expires > 0) ? clock : -clock;
611 return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
613 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
615 static void set_operstate(struct net_device *dev, unsigned char transition)
617 unsigned char operstate = dev->operstate;
619 switch (transition) {
621 if ((operstate == IF_OPER_DORMANT ||
622 operstate == IF_OPER_UNKNOWN) &&
624 operstate = IF_OPER_UP;
627 case IF_OPER_DORMANT:
628 if (operstate == IF_OPER_UP ||
629 operstate == IF_OPER_UNKNOWN)
630 operstate = IF_OPER_DORMANT;
634 if (dev->operstate != operstate) {
635 write_lock_bh(&dev_base_lock);
636 dev->operstate = operstate;
637 write_unlock_bh(&dev_base_lock);
638 netdev_state_change(dev);
642 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
644 return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
645 (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
648 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
649 const struct ifinfomsg *ifm)
651 unsigned int flags = ifm->ifi_flags;
653 /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
655 flags = (flags & ifm->ifi_change) |
656 (rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
661 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
662 const struct rtnl_link_stats64 *b)
664 a->rx_packets = b->rx_packets;
665 a->tx_packets = b->tx_packets;
666 a->rx_bytes = b->rx_bytes;
667 a->tx_bytes = b->tx_bytes;
668 a->rx_errors = b->rx_errors;
669 a->tx_errors = b->tx_errors;
670 a->rx_dropped = b->rx_dropped;
671 a->tx_dropped = b->tx_dropped;
673 a->multicast = b->multicast;
674 a->collisions = b->collisions;
676 a->rx_length_errors = b->rx_length_errors;
677 a->rx_over_errors = b->rx_over_errors;
678 a->rx_crc_errors = b->rx_crc_errors;
679 a->rx_frame_errors = b->rx_frame_errors;
680 a->rx_fifo_errors = b->rx_fifo_errors;
681 a->rx_missed_errors = b->rx_missed_errors;
683 a->tx_aborted_errors = b->tx_aborted_errors;
684 a->tx_carrier_errors = b->tx_carrier_errors;
685 a->tx_fifo_errors = b->tx_fifo_errors;
686 a->tx_heartbeat_errors = b->tx_heartbeat_errors;
687 a->tx_window_errors = b->tx_window_errors;
689 a->rx_compressed = b->rx_compressed;
690 a->tx_compressed = b->tx_compressed;
693 static void copy_rtnl_link_stats64(void *v, const struct rtnl_link_stats64 *b)
695 memcpy(v, b, sizeof(*b));
699 static inline int rtnl_vfinfo_size(const struct net_device *dev,
702 if (dev->dev.parent && dev_is_pci(dev->dev.parent) &&
703 (ext_filter_mask & RTEXT_FILTER_VF)) {
704 int num_vfs = dev_num_vf(dev->dev.parent);
705 size_t size = nla_total_size(sizeof(struct nlattr));
706 size += nla_total_size(num_vfs * sizeof(struct nlattr));
708 (nla_total_size(sizeof(struct ifla_vf_mac)) +
709 nla_total_size(sizeof(struct ifla_vf_vlan)) +
710 nla_total_size(sizeof(struct ifla_vf_tx_rate)) +
711 nla_total_size(sizeof(struct ifla_vf_spoofchk)));
717 static size_t rtnl_port_size(const struct net_device *dev,
720 size_t port_size = nla_total_size(4) /* PORT_VF */
721 + nla_total_size(PORT_PROFILE_MAX) /* PORT_PROFILE */
722 + nla_total_size(sizeof(struct ifla_port_vsi))
724 + nla_total_size(PORT_UUID_MAX) /* PORT_INSTANCE_UUID */
725 + nla_total_size(PORT_UUID_MAX) /* PORT_HOST_UUID */
726 + nla_total_size(1) /* PROT_VDP_REQUEST */
727 + nla_total_size(2); /* PORT_VDP_RESPONSE */
728 size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
729 size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
731 size_t port_self_size = nla_total_size(sizeof(struct nlattr))
734 if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
735 !(ext_filter_mask & RTEXT_FILTER_VF))
737 if (dev_num_vf(dev->dev.parent))
738 return port_self_size + vf_ports_size +
739 vf_port_size * dev_num_vf(dev->dev.parent);
741 return port_self_size;
744 static noinline size_t if_nlmsg_size(const struct net_device *dev,
747 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
748 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
749 + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
750 + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
751 + nla_total_size(sizeof(struct rtnl_link_ifmap))
752 + nla_total_size(sizeof(struct rtnl_link_stats))
753 + nla_total_size(sizeof(struct rtnl_link_stats64))
754 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
755 + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
756 + nla_total_size(4) /* IFLA_TXQLEN */
757 + nla_total_size(4) /* IFLA_WEIGHT */
758 + nla_total_size(4) /* IFLA_MTU */
759 + nla_total_size(4) /* IFLA_LINK */
760 + nla_total_size(4) /* IFLA_MASTER */
761 + nla_total_size(1) /* IFLA_CARRIER */
762 + nla_total_size(4) /* IFLA_PROMISCUITY */
763 + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
764 + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
765 + nla_total_size(1) /* IFLA_OPERSTATE */
766 + nla_total_size(1) /* IFLA_LINKMODE */
767 + nla_total_size(ext_filter_mask
768 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
769 + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
770 + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
771 + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
772 + rtnl_link_get_af_size(dev); /* IFLA_AF_SPEC */
775 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
777 struct nlattr *vf_ports;
778 struct nlattr *vf_port;
782 vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
786 for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
787 vf_port = nla_nest_start(skb, IFLA_VF_PORT);
789 goto nla_put_failure;
790 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
791 goto nla_put_failure;
792 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
793 if (err == -EMSGSIZE)
794 goto nla_put_failure;
796 nla_nest_cancel(skb, vf_port);
799 nla_nest_end(skb, vf_port);
802 nla_nest_end(skb, vf_ports);
807 nla_nest_cancel(skb, vf_ports);
811 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
813 struct nlattr *port_self;
816 port_self = nla_nest_start(skb, IFLA_PORT_SELF);
820 err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
822 nla_nest_cancel(skb, port_self);
823 return (err == -EMSGSIZE) ? err : 0;
826 nla_nest_end(skb, port_self);
831 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
836 if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
837 !(ext_filter_mask & RTEXT_FILTER_VF))
840 err = rtnl_port_self_fill(skb, dev);
844 if (dev_num_vf(dev->dev.parent)) {
845 err = rtnl_vf_ports_fill(skb, dev);
853 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
854 int type, u32 pid, u32 seq, u32 change,
855 unsigned int flags, u32 ext_filter_mask)
857 struct ifinfomsg *ifm;
858 struct nlmsghdr *nlh;
859 struct rtnl_link_stats64 temp;
860 const struct rtnl_link_stats64 *stats;
861 struct nlattr *attr, *af_spec;
862 struct rtnl_af_ops *af_ops;
863 struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
866 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
870 ifm = nlmsg_data(nlh);
871 ifm->ifi_family = AF_UNSPEC;
873 ifm->ifi_type = dev->type;
874 ifm->ifi_index = dev->ifindex;
875 ifm->ifi_flags = dev_get_flags(dev);
876 ifm->ifi_change = change;
878 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
879 nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
880 nla_put_u8(skb, IFLA_OPERSTATE,
881 netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
882 nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
883 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
884 nla_put_u32(skb, IFLA_GROUP, dev->group) ||
885 nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
886 nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
888 nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
890 (dev->ifindex != dev->iflink &&
891 nla_put_u32(skb, IFLA_LINK, dev->iflink)) ||
893 nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex)) ||
894 nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
896 nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
898 nla_put_string(skb, IFLA_IFALIAS, dev->ifalias)))
899 goto nla_put_failure;
902 struct rtnl_link_ifmap map = {
903 .mem_start = dev->mem_start,
904 .mem_end = dev->mem_end,
905 .base_addr = dev->base_addr,
908 .port = dev->if_port,
910 if (nla_put(skb, IFLA_MAP, sizeof(map), &map))
911 goto nla_put_failure;
915 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
916 nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
917 goto nla_put_failure;
920 attr = nla_reserve(skb, IFLA_STATS,
921 sizeof(struct rtnl_link_stats));
923 goto nla_put_failure;
925 stats = dev_get_stats(dev, &temp);
926 copy_rtnl_link_stats(nla_data(attr), stats);
928 attr = nla_reserve(skb, IFLA_STATS64,
929 sizeof(struct rtnl_link_stats64));
931 goto nla_put_failure;
932 copy_rtnl_link_stats64(nla_data(attr), stats);
934 if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF) &&
935 nla_put_u32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent)))
936 goto nla_put_failure;
938 if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent
939 && (ext_filter_mask & RTEXT_FILTER_VF)) {
942 struct nlattr *vfinfo, *vf;
943 int num_vfs = dev_num_vf(dev->dev.parent);
945 vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
947 goto nla_put_failure;
948 for (i = 0; i < num_vfs; i++) {
949 struct ifla_vf_info ivi;
950 struct ifla_vf_mac vf_mac;
951 struct ifla_vf_vlan vf_vlan;
952 struct ifla_vf_tx_rate vf_tx_rate;
953 struct ifla_vf_spoofchk vf_spoofchk;
956 * Not all SR-IOV capable drivers support the
957 * spoofcheck query. Preset to -1 so the user
958 * space tool can detect that the driver didn't
962 memset(ivi.mac, 0, sizeof(ivi.mac));
963 if (dev->netdev_ops->ndo_get_vf_config(dev, i, &ivi))
968 vf_spoofchk.vf = ivi.vf;
970 memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
971 vf_vlan.vlan = ivi.vlan;
972 vf_vlan.qos = ivi.qos;
973 vf_tx_rate.rate = ivi.tx_rate;
974 vf_spoofchk.setting = ivi.spoofchk;
975 vf = nla_nest_start(skb, IFLA_VF_INFO);
977 nla_nest_cancel(skb, vfinfo);
978 goto nla_put_failure;
980 if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
981 nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
982 nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
984 nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
986 goto nla_put_failure;
987 nla_nest_end(skb, vf);
989 nla_nest_end(skb, vfinfo);
992 if (rtnl_port_fill(skb, dev, ext_filter_mask))
993 goto nla_put_failure;
995 if (dev->rtnl_link_ops) {
996 if (rtnl_link_fill(skb, dev) < 0)
997 goto nla_put_failure;
1000 if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1001 goto nla_put_failure;
1003 list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1004 if (af_ops->fill_link_af) {
1008 if (!(af = nla_nest_start(skb, af_ops->family)))
1009 goto nla_put_failure;
1011 err = af_ops->fill_link_af(skb, dev);
1014 * Caller may return ENODATA to indicate that there
1015 * was no data to be dumped. This is not an error, it
1016 * means we should trim the attribute header and
1019 if (err == -ENODATA)
1020 nla_nest_cancel(skb, af);
1022 goto nla_put_failure;
1024 nla_nest_end(skb, af);
1028 nla_nest_end(skb, af_spec);
1030 return nlmsg_end(skb, nlh);
1033 nlmsg_cancel(skb, nlh);
1037 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1039 struct net *net = sock_net(skb->sk);
1042 struct net_device *dev;
1043 struct hlist_head *head;
1044 struct nlattr *tb[IFLA_MAX+1];
1045 u32 ext_filter_mask = 0;
1050 s_idx = cb->args[1];
1053 cb->seq = net->dev_base_seq;
1055 /* A hack to preserve kernel<->userspace interface.
1056 * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
1057 * However, before Linux v3.9 the code here assumed rtgenmsg and that's
1058 * what iproute2 < v3.9.0 used.
1059 * We can detect the old iproute2. Even including the IFLA_EXT_MASK
1060 * attribute, its netlink message is shorter than struct ifinfomsg.
1062 hdrlen = nlmsg_len(cb->nlh) < sizeof(struct ifinfomsg) ?
1063 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
1065 if (nlmsg_parse(cb->nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
1067 if (tb[IFLA_EXT_MASK])
1068 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1071 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1073 head = &net->dev_index_head[h];
1074 hlist_for_each_entry_rcu(dev, head, index_hlist) {
1077 err = rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1078 NETLINK_CB(cb->skb).portid,
1079 cb->nlh->nlmsg_seq, 0,
1082 /* If we ran out of room on the first message,
1085 WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
1090 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1103 const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1104 [IFLA_IFNAME] = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1105 [IFLA_ADDRESS] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1106 [IFLA_BROADCAST] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1107 [IFLA_MAP] = { .len = sizeof(struct rtnl_link_ifmap) },
1108 [IFLA_MTU] = { .type = NLA_U32 },
1109 [IFLA_LINK] = { .type = NLA_U32 },
1110 [IFLA_MASTER] = { .type = NLA_U32 },
1111 [IFLA_CARRIER] = { .type = NLA_U8 },
1112 [IFLA_TXQLEN] = { .type = NLA_U32 },
1113 [IFLA_WEIGHT] = { .type = NLA_U32 },
1114 [IFLA_OPERSTATE] = { .type = NLA_U8 },
1115 [IFLA_LINKMODE] = { .type = NLA_U8 },
1116 [IFLA_LINKINFO] = { .type = NLA_NESTED },
1117 [IFLA_NET_NS_PID] = { .type = NLA_U32 },
1118 [IFLA_NET_NS_FD] = { .type = NLA_U32 },
1119 [IFLA_IFALIAS] = { .type = NLA_STRING, .len = IFALIASZ-1 },
1120 [IFLA_VFINFO_LIST] = {. type = NLA_NESTED },
1121 [IFLA_VF_PORTS] = { .type = NLA_NESTED },
1122 [IFLA_PORT_SELF] = { .type = NLA_NESTED },
1123 [IFLA_AF_SPEC] = { .type = NLA_NESTED },
1124 [IFLA_EXT_MASK] = { .type = NLA_U32 },
1125 [IFLA_PROMISCUITY] = { .type = NLA_U32 },
1126 [IFLA_NUM_TX_QUEUES] = { .type = NLA_U32 },
1127 [IFLA_NUM_RX_QUEUES] = { .type = NLA_U32 },
1129 EXPORT_SYMBOL(ifla_policy);
1131 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1132 [IFLA_INFO_KIND] = { .type = NLA_STRING },
1133 [IFLA_INFO_DATA] = { .type = NLA_NESTED },
1136 static const struct nla_policy ifla_vfinfo_policy[IFLA_VF_INFO_MAX+1] = {
1137 [IFLA_VF_INFO] = { .type = NLA_NESTED },
1140 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1141 [IFLA_VF_MAC] = { .type = NLA_BINARY,
1142 .len = sizeof(struct ifla_vf_mac) },
1143 [IFLA_VF_VLAN] = { .type = NLA_BINARY,
1144 .len = sizeof(struct ifla_vf_vlan) },
1145 [IFLA_VF_TX_RATE] = { .type = NLA_BINARY,
1146 .len = sizeof(struct ifla_vf_tx_rate) },
1147 [IFLA_VF_SPOOFCHK] = { .type = NLA_BINARY,
1148 .len = sizeof(struct ifla_vf_spoofchk) },
1151 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1152 [IFLA_PORT_VF] = { .type = NLA_U32 },
1153 [IFLA_PORT_PROFILE] = { .type = NLA_STRING,
1154 .len = PORT_PROFILE_MAX },
1155 [IFLA_PORT_VSI_TYPE] = { .type = NLA_BINARY,
1156 .len = sizeof(struct ifla_port_vsi)},
1157 [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1158 .len = PORT_UUID_MAX },
1159 [IFLA_PORT_HOST_UUID] = { .type = NLA_STRING,
1160 .len = PORT_UUID_MAX },
1161 [IFLA_PORT_REQUEST] = { .type = NLA_U8, },
1162 [IFLA_PORT_RESPONSE] = { .type = NLA_U16, },
1165 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1168 /* Examine the link attributes and figure out which
1169 * network namespace we are talking about.
1171 if (tb[IFLA_NET_NS_PID])
1172 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1173 else if (tb[IFLA_NET_NS_FD])
1174 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1176 net = get_net(src_net);
1179 EXPORT_SYMBOL(rtnl_link_get_net);
1181 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1184 if (tb[IFLA_ADDRESS] &&
1185 nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1188 if (tb[IFLA_BROADCAST] &&
1189 nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1193 if (tb[IFLA_AF_SPEC]) {
1197 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1198 const struct rtnl_af_ops *af_ops;
1200 if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1201 return -EAFNOSUPPORT;
1203 if (!af_ops->set_link_af)
1206 if (af_ops->validate_link_af) {
1207 err = af_ops->validate_link_af(dev, af);
1217 static int do_setvfinfo(struct net_device *dev, struct nlattr *attr)
1219 int rem, err = -EINVAL;
1221 const struct net_device_ops *ops = dev->netdev_ops;
1223 nla_for_each_nested(vf, attr, rem) {
1224 switch (nla_type(vf)) {
1226 struct ifla_vf_mac *ivm;
1229 if (ops->ndo_set_vf_mac)
1230 err = ops->ndo_set_vf_mac(dev, ivm->vf,
1234 case IFLA_VF_VLAN: {
1235 struct ifla_vf_vlan *ivv;
1238 if (ops->ndo_set_vf_vlan)
1239 err = ops->ndo_set_vf_vlan(dev, ivv->vf,
1244 case IFLA_VF_TX_RATE: {
1245 struct ifla_vf_tx_rate *ivt;
1248 if (ops->ndo_set_vf_tx_rate)
1249 err = ops->ndo_set_vf_tx_rate(dev, ivt->vf,
1253 case IFLA_VF_SPOOFCHK: {
1254 struct ifla_vf_spoofchk *ivs;
1257 if (ops->ndo_set_vf_spoofchk)
1258 err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1272 static int do_set_master(struct net_device *dev, int ifindex)
1274 struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1275 const struct net_device_ops *ops;
1279 if (upper_dev->ifindex == ifindex)
1281 ops = upper_dev->netdev_ops;
1282 if (ops->ndo_del_slave) {
1283 err = ops->ndo_del_slave(upper_dev, dev);
1292 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
1295 ops = upper_dev->netdev_ops;
1296 if (ops->ndo_add_slave) {
1297 err = ops->ndo_add_slave(upper_dev, dev);
1307 static int do_setlink(const struct sk_buff *skb,
1308 struct net_device *dev, struct ifinfomsg *ifm,
1309 struct nlattr **tb, char *ifname, int modified)
1311 const struct net_device_ops *ops = dev->netdev_ops;
1314 if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1315 struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1320 if (!netlink_ns_capable(skb, net->user_ns, CAP_NET_ADMIN)) {
1324 err = dev_change_net_namespace(dev, net, ifname);
1332 struct rtnl_link_ifmap *u_map;
1335 if (!ops->ndo_set_config) {
1340 if (!netif_device_present(dev)) {
1345 u_map = nla_data(tb[IFLA_MAP]);
1346 k_map.mem_start = (unsigned long) u_map->mem_start;
1347 k_map.mem_end = (unsigned long) u_map->mem_end;
1348 k_map.base_addr = (unsigned short) u_map->base_addr;
1349 k_map.irq = (unsigned char) u_map->irq;
1350 k_map.dma = (unsigned char) u_map->dma;
1351 k_map.port = (unsigned char) u_map->port;
1353 err = ops->ndo_set_config(dev, &k_map);
1360 if (tb[IFLA_ADDRESS]) {
1361 struct sockaddr *sa;
1364 len = sizeof(sa_family_t) + dev->addr_len;
1365 sa = kmalloc(len, GFP_KERNEL);
1370 sa->sa_family = dev->type;
1371 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1373 err = dev_set_mac_address(dev, sa);
1381 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1387 if (tb[IFLA_GROUP]) {
1388 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1393 * Interface selected by interface index but interface
1394 * name provided implies that a name change has been
1397 if (ifm->ifi_index > 0 && ifname[0]) {
1398 err = dev_change_name(dev, ifname);
1404 if (tb[IFLA_IFALIAS]) {
1405 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
1406 nla_len(tb[IFLA_IFALIAS]));
1412 if (tb[IFLA_BROADCAST]) {
1413 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
1414 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
1417 if (ifm->ifi_flags || ifm->ifi_change) {
1418 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1423 if (tb[IFLA_MASTER]) {
1424 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
1430 if (tb[IFLA_CARRIER]) {
1431 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
1437 if (tb[IFLA_TXQLEN])
1438 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1440 if (tb[IFLA_OPERSTATE])
1441 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1443 if (tb[IFLA_LINKMODE]) {
1444 write_lock_bh(&dev_base_lock);
1445 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1446 write_unlock_bh(&dev_base_lock);
1449 if (tb[IFLA_VFINFO_LIST]) {
1450 struct nlattr *attr;
1452 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
1453 if (nla_type(attr) != IFLA_VF_INFO) {
1457 err = do_setvfinfo(dev, attr);
1465 if (tb[IFLA_VF_PORTS]) {
1466 struct nlattr *port[IFLA_PORT_MAX+1];
1467 struct nlattr *attr;
1472 if (!ops->ndo_set_vf_port)
1475 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
1476 if (nla_type(attr) != IFLA_VF_PORT)
1478 err = nla_parse_nested(port, IFLA_PORT_MAX,
1479 attr, ifla_port_policy);
1482 if (!port[IFLA_PORT_VF]) {
1486 vf = nla_get_u32(port[IFLA_PORT_VF]);
1487 err = ops->ndo_set_vf_port(dev, vf, port);
1495 if (tb[IFLA_PORT_SELF]) {
1496 struct nlattr *port[IFLA_PORT_MAX+1];
1498 err = nla_parse_nested(port, IFLA_PORT_MAX,
1499 tb[IFLA_PORT_SELF], ifla_port_policy);
1504 if (ops->ndo_set_vf_port)
1505 err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
1511 if (tb[IFLA_AF_SPEC]) {
1515 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1516 const struct rtnl_af_ops *af_ops;
1518 if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1521 err = af_ops->set_link_af(dev, af);
1531 if (err < 0 && modified)
1532 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",
1538 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
1540 struct net *net = sock_net(skb->sk);
1541 struct ifinfomsg *ifm;
1542 struct net_device *dev;
1544 struct nlattr *tb[IFLA_MAX+1];
1545 char ifname[IFNAMSIZ];
1547 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1551 if (tb[IFLA_IFNAME])
1552 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1557 ifm = nlmsg_data(nlh);
1558 if (ifm->ifi_index > 0)
1559 dev = __dev_get_by_index(net, ifm->ifi_index);
1560 else if (tb[IFLA_IFNAME])
1561 dev = __dev_get_by_name(net, ifname);
1570 err = validate_linkmsg(dev, tb);
1574 err = do_setlink(skb, dev, ifm, tb, ifname, 0);
1579 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
1581 struct net *net = sock_net(skb->sk);
1582 const struct rtnl_link_ops *ops;
1583 struct net_device *dev;
1584 struct ifinfomsg *ifm;
1585 char ifname[IFNAMSIZ];
1586 struct nlattr *tb[IFLA_MAX+1];
1588 LIST_HEAD(list_kill);
1590 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1594 if (tb[IFLA_IFNAME])
1595 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1597 ifm = nlmsg_data(nlh);
1598 if (ifm->ifi_index > 0)
1599 dev = __dev_get_by_index(net, ifm->ifi_index);
1600 else if (tb[IFLA_IFNAME])
1601 dev = __dev_get_by_name(net, ifname);
1608 ops = dev->rtnl_link_ops;
1612 ops->dellink(dev, &list_kill);
1613 unregister_netdevice_many(&list_kill);
1617 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
1619 unsigned int old_flags;
1622 old_flags = dev->flags;
1623 if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
1624 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1629 dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
1630 rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U);
1632 __dev_notify_flags(dev, old_flags);
1635 EXPORT_SYMBOL(rtnl_configure_link);
1637 struct net_device *rtnl_create_link(struct net *net,
1638 char *ifname, const struct rtnl_link_ops *ops, struct nlattr *tb[])
1641 struct net_device *dev;
1642 unsigned int num_tx_queues = 1;
1643 unsigned int num_rx_queues = 1;
1645 if (tb[IFLA_NUM_TX_QUEUES])
1646 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
1647 else if (ops->get_num_tx_queues)
1648 num_tx_queues = ops->get_num_tx_queues();
1650 if (tb[IFLA_NUM_RX_QUEUES])
1651 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
1652 else if (ops->get_num_rx_queues)
1653 num_rx_queues = ops->get_num_rx_queues();
1656 dev = alloc_netdev_mqs(ops->priv_size, ifname, ops->setup,
1657 num_tx_queues, num_rx_queues);
1661 dev_net_set(dev, net);
1662 dev->rtnl_link_ops = ops;
1663 dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
1666 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
1667 if (tb[IFLA_ADDRESS]) {
1668 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
1669 nla_len(tb[IFLA_ADDRESS]));
1670 dev->addr_assign_type = NET_ADDR_SET;
1672 if (tb[IFLA_BROADCAST])
1673 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
1674 nla_len(tb[IFLA_BROADCAST]));
1675 if (tb[IFLA_TXQLEN])
1676 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1677 if (tb[IFLA_OPERSTATE])
1678 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1679 if (tb[IFLA_LINKMODE])
1680 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1682 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1687 return ERR_PTR(err);
1689 EXPORT_SYMBOL(rtnl_create_link);
1691 static int rtnl_group_changelink(const struct sk_buff *skb,
1692 struct net *net, int group,
1693 struct ifinfomsg *ifm,
1696 struct net_device *dev;
1699 for_each_netdev(net, dev) {
1700 if (dev->group == group) {
1701 err = do_setlink(skb, dev, ifm, tb, NULL, 0);
1710 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh)
1712 struct net *net = sock_net(skb->sk);
1713 const struct rtnl_link_ops *ops;
1714 struct net_device *dev;
1715 struct ifinfomsg *ifm;
1716 char kind[MODULE_NAME_LEN];
1717 char ifname[IFNAMSIZ];
1718 struct nlattr *tb[IFLA_MAX+1];
1719 struct nlattr *linkinfo[IFLA_INFO_MAX+1];
1722 #ifdef CONFIG_MODULES
1725 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1729 if (tb[IFLA_IFNAME])
1730 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1734 ifm = nlmsg_data(nlh);
1735 if (ifm->ifi_index > 0)
1736 dev = __dev_get_by_index(net, ifm->ifi_index);
1739 dev = __dev_get_by_name(net, ifname);
1744 err = validate_linkmsg(dev, tb);
1748 if (tb[IFLA_LINKINFO]) {
1749 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
1750 tb[IFLA_LINKINFO], ifla_info_policy);
1754 memset(linkinfo, 0, sizeof(linkinfo));
1756 if (linkinfo[IFLA_INFO_KIND]) {
1757 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
1758 ops = rtnl_link_ops_get(kind);
1765 struct nlattr *attr[ops ? ops->maxtype + 1 : 0], **data = NULL;
1766 struct net *dest_net;
1769 if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
1770 err = nla_parse_nested(attr, ops->maxtype,
1771 linkinfo[IFLA_INFO_DATA],
1777 if (ops->validate) {
1778 err = ops->validate(tb, data);
1787 if (nlh->nlmsg_flags & NLM_F_EXCL)
1789 if (nlh->nlmsg_flags & NLM_F_REPLACE)
1792 if (linkinfo[IFLA_INFO_DATA]) {
1793 if (!ops || ops != dev->rtnl_link_ops ||
1797 err = ops->changelink(dev, tb, data);
1803 return do_setlink(skb, dev, ifm, tb, ifname, modified);
1806 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
1807 if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
1808 return rtnl_group_changelink(skb, net,
1809 nla_get_u32(tb[IFLA_GROUP]),
1814 if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
1818 #ifdef CONFIG_MODULES
1821 request_module("rtnl-link-%s", kind);
1823 ops = rtnl_link_ops_get(kind);
1832 snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
1834 dest_net = rtnl_link_get_net(net, tb);
1835 if (IS_ERR(dest_net))
1836 return PTR_ERR(dest_net);
1838 dev = rtnl_create_link(dest_net, ifname, ops, tb);
1844 dev->ifindex = ifm->ifi_index;
1847 err = ops->newlink(net, dev, tb, data);
1849 err = register_netdevice(dev);
1851 if (err < 0 && !IS_ERR(dev))
1856 err = rtnl_configure_link(dev, ifm);
1858 unregister_netdevice(dev);
1865 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh)
1867 struct net *net = sock_net(skb->sk);
1868 struct ifinfomsg *ifm;
1869 char ifname[IFNAMSIZ];
1870 struct nlattr *tb[IFLA_MAX+1];
1871 struct net_device *dev = NULL;
1872 struct sk_buff *nskb;
1874 u32 ext_filter_mask = 0;
1876 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1880 if (tb[IFLA_IFNAME])
1881 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1883 if (tb[IFLA_EXT_MASK])
1884 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1886 ifm = nlmsg_data(nlh);
1887 if (ifm->ifi_index > 0)
1888 dev = __dev_get_by_index(net, ifm->ifi_index);
1889 else if (tb[IFLA_IFNAME])
1890 dev = __dev_get_by_name(net, ifname);
1897 nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
1901 err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).portid,
1902 nlh->nlmsg_seq, 0, 0, ext_filter_mask);
1904 /* -EMSGSIZE implies BUG in if_nlmsg_size */
1905 WARN_ON(err == -EMSGSIZE);
1908 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
1913 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
1915 struct net *net = sock_net(skb->sk);
1916 struct net_device *dev;
1917 struct nlattr *tb[IFLA_MAX+1];
1918 u32 ext_filter_mask = 0;
1919 u16 min_ifinfo_dump_size = 0;
1922 /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
1923 hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
1924 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
1926 if (nlmsg_parse(nlh, hdrlen, tb, IFLA_MAX, ifla_policy) >= 0) {
1927 if (tb[IFLA_EXT_MASK])
1928 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1931 if (!ext_filter_mask)
1932 return NLMSG_GOODSIZE;
1934 * traverse the list of net devices and compute the minimum
1935 * buffer size based upon the filter mask.
1937 list_for_each_entry(dev, &net->dev_base_head, dev_list) {
1938 min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
1943 return min_ifinfo_dump_size;
1946 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
1949 int s_idx = cb->family;
1953 for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
1954 int type = cb->nlh->nlmsg_type-RTM_BASE;
1955 if (idx < s_idx || idx == PF_PACKET)
1957 if (rtnl_msg_handlers[idx] == NULL ||
1958 rtnl_msg_handlers[idx][type].dumpit == NULL)
1961 memset(&cb->args[0], 0, sizeof(cb->args));
1965 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
1973 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change)
1975 struct net *net = dev_net(dev);
1976 struct sk_buff *skb;
1978 size_t if_info_size;
1980 skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), GFP_KERNEL);
1984 err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0);
1986 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
1987 WARN_ON(err == -EMSGSIZE);
1991 rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_KERNEL);
1995 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
1997 EXPORT_SYMBOL(rtmsg_ifinfo);
1999 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
2000 struct net_device *dev,
2001 u8 *addr, u32 pid, u32 seq,
2002 int type, unsigned int flags,
2005 struct nlmsghdr *nlh;
2008 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
2012 ndm = nlmsg_data(nlh);
2013 ndm->ndm_family = AF_BRIDGE;
2016 ndm->ndm_flags = flags;
2018 ndm->ndm_ifindex = dev->ifindex;
2019 ndm->ndm_state = NUD_PERMANENT;
2021 if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
2022 goto nla_put_failure;
2024 return nlmsg_end(skb, nlh);
2027 nlmsg_cancel(skb, nlh);
2031 static inline size_t rtnl_fdb_nlmsg_size(void)
2033 return NLMSG_ALIGN(sizeof(struct ndmsg)) + nla_total_size(ETH_ALEN);
2036 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, int type)
2038 struct net *net = dev_net(dev);
2039 struct sk_buff *skb;
2042 skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
2046 err = nlmsg_populate_fdb_fill(skb, dev, addr, 0, 0, type, NTF_SELF, 0);
2052 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2055 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
2059 * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
2061 int ndo_dflt_fdb_add(struct ndmsg *ndm,
2062 struct nlattr *tb[],
2063 struct net_device *dev,
2064 const unsigned char *addr,
2069 /* If aging addresses are supported device will need to
2070 * implement its own handler for this.
2072 if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
2073 pr_info("%s: FDB only supports static addresses\n", dev->name);
2077 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2078 err = dev_uc_add_excl(dev, addr);
2079 else if (is_multicast_ether_addr(addr))
2080 err = dev_mc_add_excl(dev, addr);
2082 /* Only return duplicate errors if NLM_F_EXCL is set */
2083 if (err == -EEXIST && !(flags & NLM_F_EXCL))
2088 EXPORT_SYMBOL(ndo_dflt_fdb_add);
2090 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh)
2092 struct net *net = sock_net(skb->sk);
2094 struct nlattr *tb[NDA_MAX+1];
2095 struct net_device *dev;
2099 err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2103 ndm = nlmsg_data(nlh);
2104 if (ndm->ndm_ifindex == 0) {
2105 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n");
2109 dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2111 pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n");
2115 if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2116 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n");
2120 addr = nla_data(tb[NDA_LLADDR]);
2121 if (is_zero_ether_addr(addr)) {
2122 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ether address\n");
2128 /* Support fdb on master device the net/bridge default case */
2129 if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2130 (dev->priv_flags & IFF_BRIDGE_PORT)) {
2131 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2132 const struct net_device_ops *ops = br_dev->netdev_ops;
2134 err = ops->ndo_fdb_add(ndm, tb, dev, addr, nlh->nlmsg_flags);
2138 ndm->ndm_flags &= ~NTF_MASTER;
2141 /* Embedded bridge, macvlan, and any other device support */
2142 if ((ndm->ndm_flags & NTF_SELF)) {
2143 if (dev->netdev_ops->ndo_fdb_add)
2144 err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
2147 err = ndo_dflt_fdb_add(ndm, tb, dev, addr,
2151 rtnl_fdb_notify(dev, addr, RTM_NEWNEIGH);
2152 ndm->ndm_flags &= ~NTF_SELF;
2160 * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
2162 int ndo_dflt_fdb_del(struct ndmsg *ndm,
2163 struct nlattr *tb[],
2164 struct net_device *dev,
2165 const unsigned char *addr)
2167 int err = -EOPNOTSUPP;
2169 /* If aging addresses are supported device will need to
2170 * implement its own handler for this.
2172 if (!(ndm->ndm_state & NUD_PERMANENT)) {
2173 pr_info("%s: FDB only supports static addresses\n", dev->name);
2177 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2178 err = dev_uc_del(dev, addr);
2179 else if (is_multicast_ether_addr(addr))
2180 err = dev_mc_del(dev, addr);
2186 EXPORT_SYMBOL(ndo_dflt_fdb_del);
2188 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh)
2190 struct net *net = sock_net(skb->sk);
2192 struct nlattr *tb[NDA_MAX+1];
2193 struct net_device *dev;
2197 if (!netlink_capable(skb, CAP_NET_ADMIN))
2200 err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL);
2204 ndm = nlmsg_data(nlh);
2205 if (ndm->ndm_ifindex == 0) {
2206 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n");
2210 dev = __dev_get_by_index(net, ndm->ndm_ifindex);
2212 pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n");
2216 if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
2217 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid address\n");
2221 addr = nla_data(tb[NDA_LLADDR]);
2222 if (is_zero_ether_addr(addr)) {
2223 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ether address\n");
2229 /* Support fdb on master device the net/bridge default case */
2230 if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
2231 (dev->priv_flags & IFF_BRIDGE_PORT)) {
2232 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2233 const struct net_device_ops *ops = br_dev->netdev_ops;
2235 if (ops->ndo_fdb_del)
2236 err = ops->ndo_fdb_del(ndm, tb, dev, addr);
2241 ndm->ndm_flags &= ~NTF_MASTER;
2244 /* Embedded bridge, macvlan, and any other device support */
2245 if (ndm->ndm_flags & NTF_SELF) {
2246 if (dev->netdev_ops->ndo_fdb_del)
2247 err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr);
2249 err = ndo_dflt_fdb_del(ndm, tb, dev, addr);
2252 rtnl_fdb_notify(dev, addr, RTM_DELNEIGH);
2253 ndm->ndm_flags &= ~NTF_SELF;
2260 static int nlmsg_populate_fdb(struct sk_buff *skb,
2261 struct netlink_callback *cb,
2262 struct net_device *dev,
2264 struct netdev_hw_addr_list *list)
2266 struct netdev_hw_addr *ha;
2270 portid = NETLINK_CB(cb->skb).portid;
2271 seq = cb->nlh->nlmsg_seq;
2273 list_for_each_entry(ha, &list->list, list) {
2274 if (*idx < cb->args[0])
2277 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr,
2279 RTM_NEWNEIGH, NTF_SELF,
2290 * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
2291 * @nlh: netlink message header
2294 * Default netdevice operation to dump the existing unicast address list.
2295 * Returns number of addresses from list put in skb.
2297 int ndo_dflt_fdb_dump(struct sk_buff *skb,
2298 struct netlink_callback *cb,
2299 struct net_device *dev,
2304 netif_addr_lock_bh(dev);
2305 err = nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->uc);
2308 nlmsg_populate_fdb(skb, cb, dev, &idx, &dev->mc);
2310 netif_addr_unlock_bh(dev);
2313 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
2315 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
2318 struct net *net = sock_net(skb->sk);
2319 struct net_device *dev;
2322 for_each_netdev_rcu(net, dev) {
2323 if (dev->priv_flags & IFF_BRIDGE_PORT) {
2324 struct net_device *br_dev;
2325 const struct net_device_ops *ops;
2327 br_dev = netdev_master_upper_dev_get(dev);
2328 ops = br_dev->netdev_ops;
2329 if (ops->ndo_fdb_dump)
2330 idx = ops->ndo_fdb_dump(skb, cb, dev, idx);
2333 if (dev->netdev_ops->ndo_fdb_dump)
2334 idx = dev->netdev_ops->ndo_fdb_dump(skb, cb, dev, idx);
2336 idx = ndo_dflt_fdb_dump(skb, cb, dev, idx);
2344 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
2345 struct net_device *dev, u16 mode)
2347 struct nlmsghdr *nlh;
2348 struct ifinfomsg *ifm;
2349 struct nlattr *br_afspec;
2350 u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
2351 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2353 nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), NLM_F_MULTI);
2357 ifm = nlmsg_data(nlh);
2358 ifm->ifi_family = AF_BRIDGE;
2360 ifm->ifi_type = dev->type;
2361 ifm->ifi_index = dev->ifindex;
2362 ifm->ifi_flags = dev_get_flags(dev);
2363 ifm->ifi_change = 0;
2366 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
2367 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
2368 nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
2370 nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
2372 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
2373 (dev->ifindex != dev->iflink &&
2374 nla_put_u32(skb, IFLA_LINK, dev->iflink)))
2375 goto nla_put_failure;
2377 br_afspec = nla_nest_start(skb, IFLA_AF_SPEC);
2379 goto nla_put_failure;
2381 if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF) ||
2382 nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
2383 nla_nest_cancel(skb, br_afspec);
2384 goto nla_put_failure;
2386 nla_nest_end(skb, br_afspec);
2388 return nlmsg_end(skb, nlh);
2390 nlmsg_cancel(skb, nlh);
2393 EXPORT_SYMBOL(ndo_dflt_bridge_getlink);
2395 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
2397 struct net *net = sock_net(skb->sk);
2398 struct net_device *dev;
2400 u32 portid = NETLINK_CB(cb->skb).portid;
2401 u32 seq = cb->nlh->nlmsg_seq;
2402 struct nlattr *extfilt;
2403 u32 filter_mask = 0;
2405 extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct ifinfomsg),
2408 filter_mask = nla_get_u32(extfilt);
2411 for_each_netdev_rcu(net, dev) {
2412 const struct net_device_ops *ops = dev->netdev_ops;
2413 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2415 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
2416 if (idx >= cb->args[0] &&
2417 br_dev->netdev_ops->ndo_bridge_getlink(
2418 skb, portid, seq, dev, filter_mask) < 0)
2423 if (ops->ndo_bridge_getlink) {
2424 if (idx >= cb->args[0] &&
2425 ops->ndo_bridge_getlink(skb, portid, seq, dev,
2437 static inline size_t bridge_nlmsg_size(void)
2439 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
2440 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
2441 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
2442 + nla_total_size(sizeof(u32)) /* IFLA_MASTER */
2443 + nla_total_size(sizeof(u32)) /* IFLA_MTU */
2444 + nla_total_size(sizeof(u32)) /* IFLA_LINK */
2445 + nla_total_size(sizeof(u32)) /* IFLA_OPERSTATE */
2446 + nla_total_size(sizeof(u8)) /* IFLA_PROTINFO */
2447 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
2448 + nla_total_size(sizeof(u16)) /* IFLA_BRIDGE_FLAGS */
2449 + nla_total_size(sizeof(u16)); /* IFLA_BRIDGE_MODE */
2452 static int rtnl_bridge_notify(struct net_device *dev, u16 flags)
2454 struct net *net = dev_net(dev);
2455 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2456 struct sk_buff *skb;
2457 int err = -EOPNOTSUPP;
2459 skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
2465 if ((!flags || (flags & BRIDGE_FLAGS_MASTER)) &&
2466 br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
2467 err = br_dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0);
2472 if ((flags & BRIDGE_FLAGS_SELF) &&
2473 dev->netdev_ops->ndo_bridge_getlink) {
2474 err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0);
2479 rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
2482 WARN_ON(err == -EMSGSIZE);
2484 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2488 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh)
2490 struct net *net = sock_net(skb->sk);
2491 struct ifinfomsg *ifm;
2492 struct net_device *dev;
2493 struct nlattr *br_spec, *attr = NULL;
2494 int rem, err = -EOPNOTSUPP;
2495 u16 oflags, flags = 0;
2496 bool have_flags = false;
2498 if (nlmsg_len(nlh) < sizeof(*ifm))
2501 ifm = nlmsg_data(nlh);
2502 if (ifm->ifi_family != AF_BRIDGE)
2503 return -EPFNOSUPPORT;
2505 dev = __dev_get_by_index(net, ifm->ifi_index);
2507 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
2511 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
2513 nla_for_each_nested(attr, br_spec, rem) {
2514 if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
2516 flags = nla_get_u16(attr);
2524 if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
2525 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2527 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
2532 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh);
2536 flags &= ~BRIDGE_FLAGS_MASTER;
2539 if ((flags & BRIDGE_FLAGS_SELF)) {
2540 if (!dev->netdev_ops->ndo_bridge_setlink)
2543 err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh);
2546 flags &= ~BRIDGE_FLAGS_SELF;
2550 memcpy(nla_data(attr), &flags, sizeof(flags));
2551 /* Generate event to notify upper layer of bridge change */
2553 err = rtnl_bridge_notify(dev, oflags);
2558 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh)
2560 struct net *net = sock_net(skb->sk);
2561 struct ifinfomsg *ifm;
2562 struct net_device *dev;
2563 struct nlattr *br_spec, *attr = NULL;
2564 int rem, err = -EOPNOTSUPP;
2565 u16 oflags, flags = 0;
2566 bool have_flags = false;
2568 if (nlmsg_len(nlh) < sizeof(*ifm))
2571 ifm = nlmsg_data(nlh);
2572 if (ifm->ifi_family != AF_BRIDGE)
2573 return -EPFNOSUPPORT;
2575 dev = __dev_get_by_index(net, ifm->ifi_index);
2577 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
2581 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
2583 nla_for_each_nested(attr, br_spec, rem) {
2584 if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
2586 flags = nla_get_u16(attr);
2594 if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
2595 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
2597 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
2602 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh);
2606 flags &= ~BRIDGE_FLAGS_MASTER;
2609 if ((flags & BRIDGE_FLAGS_SELF)) {
2610 if (!dev->netdev_ops->ndo_bridge_dellink)
2613 err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh);
2616 flags &= ~BRIDGE_FLAGS_SELF;
2620 memcpy(nla_data(attr), &flags, sizeof(flags));
2621 /* Generate event to notify upper layer of bridge change */
2623 err = rtnl_bridge_notify(dev, oflags);
2628 /* Process one rtnetlink message. */
2630 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
2632 struct net *net = sock_net(skb->sk);
2633 rtnl_doit_func doit;
2639 type = nlh->nlmsg_type;
2645 /* All the messages must have at least 1 byte length */
2646 if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
2649 family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
2653 if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
2656 if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
2658 rtnl_dumpit_func dumpit;
2659 rtnl_calcit_func calcit;
2660 u16 min_dump_alloc = 0;
2662 dumpit = rtnl_get_dumpit(family, type);
2665 calcit = rtnl_get_calcit(family, type);
2667 min_dump_alloc = calcit(skb, nlh);
2672 struct netlink_dump_control c = {
2674 .min_dump_alloc = min_dump_alloc,
2676 err = netlink_dump_start(rtnl, skb, nlh, &c);
2682 doit = rtnl_get_doit(family, type);
2686 return doit(skb, nlh);
2689 static void rtnetlink_rcv(struct sk_buff *skb)
2692 netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
2696 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
2698 struct net_device *dev = ptr;
2704 case NETDEV_POST_INIT:
2705 case NETDEV_REGISTER:
2707 case NETDEV_PRE_TYPE_CHANGE:
2708 case NETDEV_GOING_DOWN:
2709 case NETDEV_UNREGISTER:
2710 case NETDEV_UNREGISTER_FINAL:
2711 case NETDEV_RELEASE:
2715 rtmsg_ifinfo(RTM_NEWLINK, dev, 0);
2721 static struct notifier_block rtnetlink_dev_notifier = {
2722 .notifier_call = rtnetlink_event,
2726 static int __net_init rtnetlink_net_init(struct net *net)
2729 struct netlink_kernel_cfg cfg = {
2730 .groups = RTNLGRP_MAX,
2731 .input = rtnetlink_rcv,
2732 .cb_mutex = &rtnl_mutex,
2733 .flags = NL_CFG_F_NONROOT_RECV,
2736 sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
2743 static void __net_exit rtnetlink_net_exit(struct net *net)
2745 netlink_kernel_release(net->rtnl);
2749 static struct pernet_operations rtnetlink_net_ops = {
2750 .init = rtnetlink_net_init,
2751 .exit = rtnetlink_net_exit,
2754 void __init rtnetlink_init(void)
2756 if (register_pernet_subsys(&rtnetlink_net_ops))
2757 panic("rtnetlink_init: cannot initialize rtnetlink\n");
2759 register_netdevice_notifier(&rtnetlink_dev_notifier);
2761 rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
2762 rtnl_dump_ifinfo, rtnl_calcit);
2763 rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
2764 rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
2765 rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
2767 rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
2768 rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);
2770 rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, NULL);
2771 rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, NULL);
2772 rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, NULL);
2774 rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, NULL);
2775 rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, NULL);
2776 rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, NULL);