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/pci.h>
40 #include <asm/uaccess.h>
41 #include <asm/system.h>
43 #include <linux/inet.h>
44 #include <linux/netdevice.h>
46 #include <net/protocol.h>
48 #include <net/route.h>
51 #include <net/pkt_sched.h>
52 #include <net/fib_rules.h>
53 #include <net/rtnetlink.h>
54 #include <net/net_namespace.h>
58 rtnl_dumpit_func dumpit;
59 rtnl_calcit_func calcit;
62 static DEFINE_MUTEX(rtnl_mutex);
63 static u16 min_ifinfo_dump_size;
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 ? tab[msgindex].doit : NULL;
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 ? tab[msgindex].dumpit : NULL;
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 ? tab[msgindex].calcit : NULL;
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);
277 * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
278 * @ops: struct rtnl_link_ops * to register
280 * The caller must hold the rtnl_mutex. This function should be used
281 * by drivers that create devices during module initialization. It
282 * must be called before registering the devices.
284 * Returns 0 on success or a negative error code.
286 int __rtnl_link_register(struct rtnl_link_ops *ops)
289 ops->dellink = unregister_netdevice_queue;
291 list_add_tail(&ops->list, &link_ops);
294 EXPORT_SYMBOL_GPL(__rtnl_link_register);
297 * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
298 * @ops: struct rtnl_link_ops * to register
300 * Returns 0 on success or a negative error code.
302 int rtnl_link_register(struct rtnl_link_ops *ops)
307 err = __rtnl_link_register(ops);
311 EXPORT_SYMBOL_GPL(rtnl_link_register);
313 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
315 struct net_device *dev;
316 LIST_HEAD(list_kill);
318 for_each_netdev(net, dev) {
319 if (dev->rtnl_link_ops == ops)
320 ops->dellink(dev, &list_kill);
322 unregister_netdevice_many(&list_kill);
326 * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
327 * @ops: struct rtnl_link_ops * to unregister
329 * The caller must hold the rtnl_mutex.
331 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
336 __rtnl_kill_links(net, ops);
338 list_del(&ops->list);
340 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
343 * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
344 * @ops: struct rtnl_link_ops * to unregister
346 void rtnl_link_unregister(struct rtnl_link_ops *ops)
349 __rtnl_link_unregister(ops);
352 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
354 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
356 const struct rtnl_link_ops *ops;
358 list_for_each_entry(ops, &link_ops, list) {
359 if (!strcmp(ops->kind, kind))
365 static size_t rtnl_link_get_size(const struct net_device *dev)
367 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
373 size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
374 nla_total_size(strlen(ops->kind) + 1); /* IFLA_INFO_KIND */
377 /* IFLA_INFO_DATA + nested data */
378 size += nla_total_size(sizeof(struct nlattr)) +
381 if (ops->get_xstats_size)
382 /* IFLA_INFO_XSTATS */
383 size += nla_total_size(ops->get_xstats_size(dev));
388 static LIST_HEAD(rtnl_af_ops);
390 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
392 const struct rtnl_af_ops *ops;
394 list_for_each_entry(ops, &rtnl_af_ops, list) {
395 if (ops->family == family)
403 * __rtnl_af_register - Register rtnl_af_ops with rtnetlink.
404 * @ops: struct rtnl_af_ops * to register
406 * The caller must hold the rtnl_mutex.
408 * Returns 0 on success or a negative error code.
410 int __rtnl_af_register(struct rtnl_af_ops *ops)
412 list_add_tail(&ops->list, &rtnl_af_ops);
415 EXPORT_SYMBOL_GPL(__rtnl_af_register);
418 * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
419 * @ops: struct rtnl_af_ops * to register
421 * Returns 0 on success or a negative error code.
423 int rtnl_af_register(struct rtnl_af_ops *ops)
428 err = __rtnl_af_register(ops);
432 EXPORT_SYMBOL_GPL(rtnl_af_register);
435 * __rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
436 * @ops: struct rtnl_af_ops * to unregister
438 * The caller must hold the rtnl_mutex.
440 void __rtnl_af_unregister(struct rtnl_af_ops *ops)
442 list_del(&ops->list);
444 EXPORT_SYMBOL_GPL(__rtnl_af_unregister);
447 * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
448 * @ops: struct rtnl_af_ops * to unregister
450 void rtnl_af_unregister(struct rtnl_af_ops *ops)
453 __rtnl_af_unregister(ops);
456 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
458 static size_t rtnl_link_get_af_size(const struct net_device *dev)
460 struct rtnl_af_ops *af_ops;
464 size = nla_total_size(sizeof(struct nlattr));
466 list_for_each_entry(af_ops, &rtnl_af_ops, list) {
467 if (af_ops->get_link_af_size) {
468 /* AF_* + nested data */
469 size += nla_total_size(sizeof(struct nlattr)) +
470 af_ops->get_link_af_size(dev);
477 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
479 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
480 struct nlattr *linkinfo, *data;
483 linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
484 if (linkinfo == NULL)
487 if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
488 goto err_cancel_link;
489 if (ops->fill_xstats) {
490 err = ops->fill_xstats(skb, dev);
492 goto err_cancel_link;
494 if (ops->fill_info) {
495 data = nla_nest_start(skb, IFLA_INFO_DATA);
497 goto err_cancel_link;
498 err = ops->fill_info(skb, dev);
500 goto err_cancel_data;
501 nla_nest_end(skb, data);
504 nla_nest_end(skb, linkinfo);
508 nla_nest_cancel(skb, data);
510 nla_nest_cancel(skb, linkinfo);
515 static const int rtm_min[RTM_NR_FAMILIES] =
517 [RTM_FAM(RTM_NEWLINK)] = NLMSG_LENGTH(sizeof(struct ifinfomsg)),
518 [RTM_FAM(RTM_NEWADDR)] = NLMSG_LENGTH(sizeof(struct ifaddrmsg)),
519 [RTM_FAM(RTM_NEWROUTE)] = NLMSG_LENGTH(sizeof(struct rtmsg)),
520 [RTM_FAM(RTM_NEWRULE)] = NLMSG_LENGTH(sizeof(struct fib_rule_hdr)),
521 [RTM_FAM(RTM_NEWQDISC)] = NLMSG_LENGTH(sizeof(struct tcmsg)),
522 [RTM_FAM(RTM_NEWTCLASS)] = NLMSG_LENGTH(sizeof(struct tcmsg)),
523 [RTM_FAM(RTM_NEWTFILTER)] = NLMSG_LENGTH(sizeof(struct tcmsg)),
524 [RTM_FAM(RTM_NEWACTION)] = NLMSG_LENGTH(sizeof(struct tcamsg)),
525 [RTM_FAM(RTM_GETMULTICAST)] = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
526 [RTM_FAM(RTM_GETANYCAST)] = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
529 static const int rta_max[RTM_NR_FAMILIES] =
531 [RTM_FAM(RTM_NEWLINK)] = IFLA_MAX,
532 [RTM_FAM(RTM_NEWADDR)] = IFA_MAX,
533 [RTM_FAM(RTM_NEWROUTE)] = RTA_MAX,
534 [RTM_FAM(RTM_NEWRULE)] = FRA_MAX,
535 [RTM_FAM(RTM_NEWQDISC)] = TCA_MAX,
536 [RTM_FAM(RTM_NEWTCLASS)] = TCA_MAX,
537 [RTM_FAM(RTM_NEWTFILTER)] = TCA_MAX,
538 [RTM_FAM(RTM_NEWACTION)] = TCAA_MAX,
541 void __rta_fill(struct sk_buff *skb, int attrtype, int attrlen, const void *data)
544 int size = RTA_LENGTH(attrlen);
546 rta = (struct rtattr *)skb_put(skb, RTA_ALIGN(size));
547 rta->rta_type = attrtype;
549 memcpy(RTA_DATA(rta), data, attrlen);
550 memset(RTA_DATA(rta) + attrlen, 0, RTA_ALIGN(size) - size);
552 EXPORT_SYMBOL(__rta_fill);
554 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned group, int echo)
556 struct sock *rtnl = net->rtnl;
559 NETLINK_CB(skb).dst_group = group;
561 atomic_inc(&skb->users);
562 netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
564 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
568 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
570 struct sock *rtnl = net->rtnl;
572 return nlmsg_unicast(rtnl, skb, pid);
574 EXPORT_SYMBOL(rtnl_unicast);
576 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
577 struct nlmsghdr *nlh, gfp_t flags)
579 struct sock *rtnl = net->rtnl;
583 report = nlmsg_report(nlh);
585 nlmsg_notify(rtnl, skb, pid, group, report, flags);
587 EXPORT_SYMBOL(rtnl_notify);
589 void rtnl_set_sk_err(struct net *net, u32 group, int error)
591 struct sock *rtnl = net->rtnl;
593 netlink_set_err(rtnl, 0, group, error);
595 EXPORT_SYMBOL(rtnl_set_sk_err);
597 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
602 mx = nla_nest_start(skb, RTA_METRICS);
606 for (i = 0; i < RTAX_MAX; i++) {
609 NLA_PUT_U32(skb, i+1, metrics[i]);
614 nla_nest_cancel(skb, mx);
618 return nla_nest_end(skb, mx);
621 nla_nest_cancel(skb, mx);
624 EXPORT_SYMBOL(rtnetlink_put_metrics);
626 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
627 u32 ts, u32 tsage, long expires, u32 error)
629 struct rta_cacheinfo ci = {
630 .rta_lastuse = jiffies_to_clock_t(jiffies - dst->lastuse),
631 .rta_used = dst->__use,
632 .rta_clntref = atomic_read(&(dst->__refcnt)),
640 ci.rta_expires = jiffies_to_clock_t(expires);
642 return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
644 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
646 static void set_operstate(struct net_device *dev, unsigned char transition)
648 unsigned char operstate = dev->operstate;
650 switch (transition) {
652 if ((operstate == IF_OPER_DORMANT ||
653 operstate == IF_OPER_UNKNOWN) &&
655 operstate = IF_OPER_UP;
658 case IF_OPER_DORMANT:
659 if (operstate == IF_OPER_UP ||
660 operstate == IF_OPER_UNKNOWN)
661 operstate = IF_OPER_DORMANT;
665 if (dev->operstate != operstate) {
666 write_lock_bh(&dev_base_lock);
667 dev->operstate = operstate;
668 write_unlock_bh(&dev_base_lock);
669 netdev_state_change(dev);
673 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
675 return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
676 (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
679 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
680 const struct ifinfomsg *ifm)
682 unsigned int flags = ifm->ifi_flags;
684 /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
686 flags = (flags & ifm->ifi_change) |
687 (rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
692 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
693 const struct rtnl_link_stats64 *b)
695 a->rx_packets = b->rx_packets;
696 a->tx_packets = b->tx_packets;
697 a->rx_bytes = b->rx_bytes;
698 a->tx_bytes = b->tx_bytes;
699 a->rx_errors = b->rx_errors;
700 a->tx_errors = b->tx_errors;
701 a->rx_dropped = b->rx_dropped;
702 a->tx_dropped = b->tx_dropped;
704 a->multicast = b->multicast;
705 a->collisions = b->collisions;
707 a->rx_length_errors = b->rx_length_errors;
708 a->rx_over_errors = b->rx_over_errors;
709 a->rx_crc_errors = b->rx_crc_errors;
710 a->rx_frame_errors = b->rx_frame_errors;
711 a->rx_fifo_errors = b->rx_fifo_errors;
712 a->rx_missed_errors = b->rx_missed_errors;
714 a->tx_aborted_errors = b->tx_aborted_errors;
715 a->tx_carrier_errors = b->tx_carrier_errors;
716 a->tx_fifo_errors = b->tx_fifo_errors;
717 a->tx_heartbeat_errors = b->tx_heartbeat_errors;
718 a->tx_window_errors = b->tx_window_errors;
720 a->rx_compressed = b->rx_compressed;
721 a->tx_compressed = b->tx_compressed;
724 static void copy_rtnl_link_stats64(void *v, const struct rtnl_link_stats64 *b)
726 memcpy(v, b, sizeof(*b));
730 static inline int rtnl_vfinfo_size(const struct net_device *dev)
732 if (dev->dev.parent && dev_is_pci(dev->dev.parent)) {
734 int num_vfs = dev_num_vf(dev->dev.parent);
735 size_t size = nla_total_size(sizeof(struct nlattr));
736 size += nla_total_size(num_vfs * sizeof(struct nlattr));
738 (nla_total_size(sizeof(struct ifla_vf_mac)) +
739 nla_total_size(sizeof(struct ifla_vf_vlan)) +
740 nla_total_size(sizeof(struct ifla_vf_tx_rate)) +
741 nla_total_size(sizeof(struct ifla_vf_spoofchk)));
747 static size_t rtnl_port_size(const struct net_device *dev)
749 size_t port_size = nla_total_size(4) /* PORT_VF */
750 + nla_total_size(PORT_PROFILE_MAX) /* PORT_PROFILE */
751 + nla_total_size(sizeof(struct ifla_port_vsi))
753 + nla_total_size(PORT_UUID_MAX) /* PORT_INSTANCE_UUID */
754 + nla_total_size(PORT_UUID_MAX) /* PORT_HOST_UUID */
755 + nla_total_size(1) /* PROT_VDP_REQUEST */
756 + nla_total_size(2); /* PORT_VDP_RESPONSE */
757 size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
758 size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
760 size_t port_self_size = nla_total_size(sizeof(struct nlattr))
763 if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent)
765 if (dev_num_vf(dev->dev.parent))
766 return port_self_size + vf_ports_size +
767 vf_port_size * dev_num_vf(dev->dev.parent);
769 return port_self_size;
772 static noinline size_t if_nlmsg_size(const struct net_device *dev)
774 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
775 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
776 + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
777 + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
778 + nla_total_size(sizeof(struct rtnl_link_ifmap))
779 + nla_total_size(sizeof(struct rtnl_link_stats))
780 + nla_total_size(sizeof(struct rtnl_link_stats64))
781 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
782 + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
783 + nla_total_size(4) /* IFLA_TXQLEN */
784 + nla_total_size(4) /* IFLA_WEIGHT */
785 + nla_total_size(4) /* IFLA_MTU */
786 + nla_total_size(4) /* IFLA_LINK */
787 + nla_total_size(4) /* IFLA_MASTER */
788 + nla_total_size(1) /* IFLA_OPERSTATE */
789 + nla_total_size(1) /* IFLA_LINKMODE */
790 + nla_total_size(4) /* IFLA_NUM_VF */
791 + rtnl_vfinfo_size(dev) /* IFLA_VFINFO_LIST */
792 + rtnl_port_size(dev) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
793 + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
794 + rtnl_link_get_af_size(dev); /* IFLA_AF_SPEC */
797 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
799 struct nlattr *vf_ports;
800 struct nlattr *vf_port;
804 vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
808 for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
809 vf_port = nla_nest_start(skb, IFLA_VF_PORT);
811 goto nla_put_failure;
812 NLA_PUT_U32(skb, IFLA_PORT_VF, vf);
813 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
814 if (err == -EMSGSIZE)
815 goto nla_put_failure;
817 nla_nest_cancel(skb, vf_port);
820 nla_nest_end(skb, vf_port);
823 nla_nest_end(skb, vf_ports);
828 nla_nest_cancel(skb, vf_ports);
832 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
834 struct nlattr *port_self;
837 port_self = nla_nest_start(skb, IFLA_PORT_SELF);
841 err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
843 nla_nest_cancel(skb, port_self);
844 return (err == -EMSGSIZE) ? err : 0;
847 nla_nest_end(skb, port_self);
852 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev)
856 if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent)
859 err = rtnl_port_self_fill(skb, dev);
863 if (dev_num_vf(dev->dev.parent)) {
864 err = rtnl_vf_ports_fill(skb, dev);
872 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
873 int type, u32 pid, u32 seq, u32 change,
876 struct ifinfomsg *ifm;
877 struct nlmsghdr *nlh;
878 struct rtnl_link_stats64 temp;
879 const struct rtnl_link_stats64 *stats;
880 struct nlattr *attr, *af_spec;
881 struct rtnl_af_ops *af_ops;
884 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
888 ifm = nlmsg_data(nlh);
889 ifm->ifi_family = AF_UNSPEC;
891 ifm->ifi_type = dev->type;
892 ifm->ifi_index = dev->ifindex;
893 ifm->ifi_flags = dev_get_flags(dev);
894 ifm->ifi_change = change;
896 NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
897 NLA_PUT_U32(skb, IFLA_TXQLEN, dev->tx_queue_len);
898 NLA_PUT_U8(skb, IFLA_OPERSTATE,
899 netif_running(dev) ? dev->operstate : IF_OPER_DOWN);
900 NLA_PUT_U8(skb, IFLA_LINKMODE, dev->link_mode);
901 NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
902 NLA_PUT_U32(skb, IFLA_GROUP, dev->group);
904 if (dev->ifindex != dev->iflink)
905 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
908 NLA_PUT_U32(skb, IFLA_MASTER, dev->master->ifindex);
911 NLA_PUT_STRING(skb, IFLA_QDISC, dev->qdisc->ops->id);
914 NLA_PUT_STRING(skb, IFLA_IFALIAS, dev->ifalias);
917 struct rtnl_link_ifmap map = {
918 .mem_start = dev->mem_start,
919 .mem_end = dev->mem_end,
920 .base_addr = dev->base_addr,
923 .port = dev->if_port,
925 NLA_PUT(skb, IFLA_MAP, sizeof(map), &map);
929 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
930 NLA_PUT(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast);
933 attr = nla_reserve(skb, IFLA_STATS,
934 sizeof(struct rtnl_link_stats));
936 goto nla_put_failure;
938 stats = dev_get_stats(dev, &temp);
939 copy_rtnl_link_stats(nla_data(attr), stats);
941 attr = nla_reserve(skb, IFLA_STATS64,
942 sizeof(struct rtnl_link_stats64));
944 goto nla_put_failure;
945 copy_rtnl_link_stats64(nla_data(attr), stats);
948 NLA_PUT_U32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent));
950 if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent) {
953 struct nlattr *vfinfo, *vf;
954 int num_vfs = dev_num_vf(dev->dev.parent);
956 vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
958 goto nla_put_failure;
959 for (i = 0; i < num_vfs; i++) {
960 struct ifla_vf_info ivi;
961 struct ifla_vf_mac vf_mac;
962 struct ifla_vf_vlan vf_vlan;
963 struct ifla_vf_tx_rate vf_tx_rate;
964 struct ifla_vf_spoofchk vf_spoofchk;
967 * Not all SR-IOV capable drivers support the
968 * spoofcheck query. Preset to -1 so the user
969 * space tool can detect that the driver didn't
973 if (dev->netdev_ops->ndo_get_vf_config(dev, i, &ivi))
978 vf_spoofchk.vf = ivi.vf;
980 memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
981 vf_vlan.vlan = ivi.vlan;
982 vf_vlan.qos = ivi.qos;
983 vf_tx_rate.rate = ivi.tx_rate;
984 vf_spoofchk.setting = ivi.spoofchk;
985 vf = nla_nest_start(skb, IFLA_VF_INFO);
987 nla_nest_cancel(skb, vfinfo);
988 goto nla_put_failure;
990 NLA_PUT(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac);
991 NLA_PUT(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan);
992 NLA_PUT(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
994 NLA_PUT(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
996 nla_nest_end(skb, vf);
998 nla_nest_end(skb, vfinfo);
1001 if (rtnl_port_fill(skb, dev))
1002 goto nla_put_failure;
1004 if (dev->rtnl_link_ops) {
1005 if (rtnl_link_fill(skb, dev) < 0)
1006 goto nla_put_failure;
1009 if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1010 goto nla_put_failure;
1012 list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1013 if (af_ops->fill_link_af) {
1017 if (!(af = nla_nest_start(skb, af_ops->family)))
1018 goto nla_put_failure;
1020 err = af_ops->fill_link_af(skb, dev);
1023 * Caller may return ENODATA to indicate that there
1024 * was no data to be dumped. This is not an error, it
1025 * means we should trim the attribute header and
1028 if (err == -ENODATA)
1029 nla_nest_cancel(skb, af);
1031 goto nla_put_failure;
1033 nla_nest_end(skb, af);
1037 nla_nest_end(skb, af_spec);
1039 return nlmsg_end(skb, nlh);
1042 nlmsg_cancel(skb, nlh);
1046 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1048 struct net *net = sock_net(skb->sk);
1051 struct net_device *dev;
1052 struct hlist_head *head;
1053 struct hlist_node *node;
1056 s_idx = cb->args[1];
1059 cb->seq = net->dev_base_seq;
1061 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1063 head = &net->dev_index_head[h];
1064 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
1067 if (rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1068 NETLINK_CB(cb->skb).pid,
1069 cb->nlh->nlmsg_seq, 0,
1073 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1086 const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1087 [IFLA_IFNAME] = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1088 [IFLA_ADDRESS] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1089 [IFLA_BROADCAST] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1090 [IFLA_MAP] = { .len = sizeof(struct rtnl_link_ifmap) },
1091 [IFLA_MTU] = { .type = NLA_U32 },
1092 [IFLA_LINK] = { .type = NLA_U32 },
1093 [IFLA_MASTER] = { .type = NLA_U32 },
1094 [IFLA_TXQLEN] = { .type = NLA_U32 },
1095 [IFLA_WEIGHT] = { .type = NLA_U32 },
1096 [IFLA_OPERSTATE] = { .type = NLA_U8 },
1097 [IFLA_LINKMODE] = { .type = NLA_U8 },
1098 [IFLA_LINKINFO] = { .type = NLA_NESTED },
1099 [IFLA_NET_NS_PID] = { .type = NLA_U32 },
1100 [IFLA_NET_NS_FD] = { .type = NLA_U32 },
1101 [IFLA_IFALIAS] = { .type = NLA_STRING, .len = IFALIASZ-1 },
1102 [IFLA_VFINFO_LIST] = {. type = NLA_NESTED },
1103 [IFLA_VF_PORTS] = { .type = NLA_NESTED },
1104 [IFLA_PORT_SELF] = { .type = NLA_NESTED },
1105 [IFLA_AF_SPEC] = { .type = NLA_NESTED },
1107 EXPORT_SYMBOL(ifla_policy);
1109 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1110 [IFLA_INFO_KIND] = { .type = NLA_STRING },
1111 [IFLA_INFO_DATA] = { .type = NLA_NESTED },
1114 static const struct nla_policy ifla_vfinfo_policy[IFLA_VF_INFO_MAX+1] = {
1115 [IFLA_VF_INFO] = { .type = NLA_NESTED },
1118 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1119 [IFLA_VF_MAC] = { .type = NLA_BINARY,
1120 .len = sizeof(struct ifla_vf_mac) },
1121 [IFLA_VF_VLAN] = { .type = NLA_BINARY,
1122 .len = sizeof(struct ifla_vf_vlan) },
1123 [IFLA_VF_TX_RATE] = { .type = NLA_BINARY,
1124 .len = sizeof(struct ifla_vf_tx_rate) },
1125 [IFLA_VF_SPOOFCHK] = { .type = NLA_BINARY,
1126 .len = sizeof(struct ifla_vf_spoofchk) },
1129 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1130 [IFLA_PORT_VF] = { .type = NLA_U32 },
1131 [IFLA_PORT_PROFILE] = { .type = NLA_STRING,
1132 .len = PORT_PROFILE_MAX },
1133 [IFLA_PORT_VSI_TYPE] = { .type = NLA_BINARY,
1134 .len = sizeof(struct ifla_port_vsi)},
1135 [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1136 .len = PORT_UUID_MAX },
1137 [IFLA_PORT_HOST_UUID] = { .type = NLA_STRING,
1138 .len = PORT_UUID_MAX },
1139 [IFLA_PORT_REQUEST] = { .type = NLA_U8, },
1140 [IFLA_PORT_RESPONSE] = { .type = NLA_U16, },
1143 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1146 /* Examine the link attributes and figure out which
1147 * network namespace we are talking about.
1149 if (tb[IFLA_NET_NS_PID])
1150 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1151 else if (tb[IFLA_NET_NS_FD])
1152 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1154 net = get_net(src_net);
1157 EXPORT_SYMBOL(rtnl_link_get_net);
1159 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1162 if (tb[IFLA_ADDRESS] &&
1163 nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1166 if (tb[IFLA_BROADCAST] &&
1167 nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1171 if (tb[IFLA_AF_SPEC]) {
1175 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1176 const struct rtnl_af_ops *af_ops;
1178 if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1179 return -EAFNOSUPPORT;
1181 if (!af_ops->set_link_af)
1184 if (af_ops->validate_link_af) {
1185 err = af_ops->validate_link_af(dev, af);
1195 static int do_setvfinfo(struct net_device *dev, struct nlattr *attr)
1197 int rem, err = -EINVAL;
1199 const struct net_device_ops *ops = dev->netdev_ops;
1201 nla_for_each_nested(vf, attr, rem) {
1202 switch (nla_type(vf)) {
1204 struct ifla_vf_mac *ivm;
1207 if (ops->ndo_set_vf_mac)
1208 err = ops->ndo_set_vf_mac(dev, ivm->vf,
1212 case IFLA_VF_VLAN: {
1213 struct ifla_vf_vlan *ivv;
1216 if (ops->ndo_set_vf_vlan)
1217 err = ops->ndo_set_vf_vlan(dev, ivv->vf,
1222 case IFLA_VF_TX_RATE: {
1223 struct ifla_vf_tx_rate *ivt;
1226 if (ops->ndo_set_vf_tx_rate)
1227 err = ops->ndo_set_vf_tx_rate(dev, ivt->vf,
1231 case IFLA_VF_SPOOFCHK: {
1232 struct ifla_vf_spoofchk *ivs;
1235 if (ops->ndo_set_vf_spoofchk)
1236 err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1250 static int do_set_master(struct net_device *dev, int ifindex)
1252 struct net_device *master_dev;
1253 const struct net_device_ops *ops;
1257 if (dev->master->ifindex == ifindex)
1259 ops = dev->master->netdev_ops;
1260 if (ops->ndo_del_slave) {
1261 err = ops->ndo_del_slave(dev->master, dev);
1270 master_dev = __dev_get_by_index(dev_net(dev), ifindex);
1273 ops = master_dev->netdev_ops;
1274 if (ops->ndo_add_slave) {
1275 err = ops->ndo_add_slave(master_dev, dev);
1285 static int do_setlink(struct net_device *dev, struct ifinfomsg *ifm,
1286 struct nlattr **tb, char *ifname, int modified)
1288 const struct net_device_ops *ops = dev->netdev_ops;
1289 int send_addr_notify = 0;
1292 if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1293 struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1298 err = dev_change_net_namespace(dev, net, ifname);
1306 struct rtnl_link_ifmap *u_map;
1309 if (!ops->ndo_set_config) {
1314 if (!netif_device_present(dev)) {
1319 u_map = nla_data(tb[IFLA_MAP]);
1320 k_map.mem_start = (unsigned long) u_map->mem_start;
1321 k_map.mem_end = (unsigned long) u_map->mem_end;
1322 k_map.base_addr = (unsigned short) u_map->base_addr;
1323 k_map.irq = (unsigned char) u_map->irq;
1324 k_map.dma = (unsigned char) u_map->dma;
1325 k_map.port = (unsigned char) u_map->port;
1327 err = ops->ndo_set_config(dev, &k_map);
1334 if (tb[IFLA_ADDRESS]) {
1335 struct sockaddr *sa;
1338 if (!ops->ndo_set_mac_address) {
1343 if (!netif_device_present(dev)) {
1348 len = sizeof(sa_family_t) + dev->addr_len;
1349 sa = kmalloc(len, GFP_KERNEL);
1354 sa->sa_family = dev->type;
1355 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1357 err = ops->ndo_set_mac_address(dev, sa);
1361 send_addr_notify = 1;
1363 add_device_randomness(dev->dev_addr, dev->addr_len);
1367 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1373 if (tb[IFLA_GROUP]) {
1374 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1379 * Interface selected by interface index but interface
1380 * name provided implies that a name change has been
1383 if (ifm->ifi_index > 0 && ifname[0]) {
1384 err = dev_change_name(dev, ifname);
1390 if (tb[IFLA_IFALIAS]) {
1391 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
1392 nla_len(tb[IFLA_IFALIAS]));
1398 if (tb[IFLA_BROADCAST]) {
1399 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
1400 send_addr_notify = 1;
1403 if (ifm->ifi_flags || ifm->ifi_change) {
1404 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1409 if (tb[IFLA_MASTER]) {
1410 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
1416 if (tb[IFLA_TXQLEN])
1417 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1419 if (tb[IFLA_OPERSTATE])
1420 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1422 if (tb[IFLA_LINKMODE]) {
1423 write_lock_bh(&dev_base_lock);
1424 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1425 write_unlock_bh(&dev_base_lock);
1428 if (tb[IFLA_VFINFO_LIST]) {
1429 struct nlattr *attr;
1431 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
1432 if (nla_type(attr) != IFLA_VF_INFO) {
1436 err = do_setvfinfo(dev, attr);
1444 if (tb[IFLA_VF_PORTS]) {
1445 struct nlattr *port[IFLA_PORT_MAX+1];
1446 struct nlattr *attr;
1451 if (!ops->ndo_set_vf_port)
1454 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
1455 if (nla_type(attr) != IFLA_VF_PORT)
1457 err = nla_parse_nested(port, IFLA_PORT_MAX,
1458 attr, ifla_port_policy);
1461 if (!port[IFLA_PORT_VF]) {
1465 vf = nla_get_u32(port[IFLA_PORT_VF]);
1466 err = ops->ndo_set_vf_port(dev, vf, port);
1474 if (tb[IFLA_PORT_SELF]) {
1475 struct nlattr *port[IFLA_PORT_MAX+1];
1477 err = nla_parse_nested(port, IFLA_PORT_MAX,
1478 tb[IFLA_PORT_SELF], ifla_port_policy);
1483 if (ops->ndo_set_vf_port)
1484 err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
1490 if (tb[IFLA_AF_SPEC]) {
1494 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1495 const struct rtnl_af_ops *af_ops;
1497 if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1500 err = af_ops->set_link_af(dev, af);
1510 if (err < 0 && modified && net_ratelimit())
1511 printk(KERN_WARNING "A link change request failed with "
1512 "some changes committed already. Interface %s may "
1513 "have been left with an inconsistent configuration, "
1514 "please check.\n", dev->name);
1516 if (send_addr_notify)
1517 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
1521 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1523 struct net *net = sock_net(skb->sk);
1524 struct ifinfomsg *ifm;
1525 struct net_device *dev;
1527 struct nlattr *tb[IFLA_MAX+1];
1528 char ifname[IFNAMSIZ];
1530 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1534 if (tb[IFLA_IFNAME])
1535 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1540 ifm = nlmsg_data(nlh);
1541 if (ifm->ifi_index > 0)
1542 dev = __dev_get_by_index(net, ifm->ifi_index);
1543 else if (tb[IFLA_IFNAME])
1544 dev = __dev_get_by_name(net, ifname);
1553 err = validate_linkmsg(dev, tb);
1557 err = do_setlink(dev, ifm, tb, ifname, 0);
1562 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1564 struct net *net = sock_net(skb->sk);
1565 const struct rtnl_link_ops *ops;
1566 struct net_device *dev;
1567 struct ifinfomsg *ifm;
1568 char ifname[IFNAMSIZ];
1569 struct nlattr *tb[IFLA_MAX+1];
1571 LIST_HEAD(list_kill);
1573 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1577 if (tb[IFLA_IFNAME])
1578 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1580 ifm = nlmsg_data(nlh);
1581 if (ifm->ifi_index > 0)
1582 dev = __dev_get_by_index(net, ifm->ifi_index);
1583 else if (tb[IFLA_IFNAME])
1584 dev = __dev_get_by_name(net, ifname);
1591 ops = dev->rtnl_link_ops;
1595 ops->dellink(dev, &list_kill);
1596 unregister_netdevice_many(&list_kill);
1597 list_del(&list_kill);
1601 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
1603 unsigned int old_flags;
1606 old_flags = dev->flags;
1607 if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
1608 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1613 dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
1614 rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U);
1616 __dev_notify_flags(dev, old_flags);
1619 EXPORT_SYMBOL(rtnl_configure_link);
1621 struct net_device *rtnl_create_link(struct net *src_net, struct net *net,
1622 char *ifname, const struct rtnl_link_ops *ops, struct nlattr *tb[])
1625 struct net_device *dev;
1626 unsigned int num_queues = 1;
1627 unsigned int real_num_queues = 1;
1629 if (ops->get_tx_queues) {
1630 err = ops->get_tx_queues(src_net, tb, &num_queues,
1636 dev = alloc_netdev_mq(ops->priv_size, ifname, ops->setup, num_queues);
1640 dev_net_set(dev, net);
1641 dev->rtnl_link_ops = ops;
1642 dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
1645 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
1646 if (tb[IFLA_ADDRESS])
1647 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
1648 nla_len(tb[IFLA_ADDRESS]));
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, void *arg)
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]),
1792 if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
1796 #ifdef CONFIG_MODULES
1799 request_module("rtnl-link-%s", kind);
1801 ops = rtnl_link_ops_get(kind);
1810 snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
1812 dest_net = rtnl_link_get_net(net, tb);
1813 if (IS_ERR(dest_net))
1814 return PTR_ERR(dest_net);
1816 dev = rtnl_create_link(net, dest_net, ifname, ops, tb);
1820 else if (ops->newlink)
1821 err = ops->newlink(net, dev, tb, data);
1823 err = register_netdevice(dev);
1825 if (err < 0 && !IS_ERR(dev))
1830 err = rtnl_configure_link(dev, ifm);
1832 unregister_netdevice(dev);
1839 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
1841 struct net *net = sock_net(skb->sk);
1842 struct ifinfomsg *ifm;
1843 char ifname[IFNAMSIZ];
1844 struct nlattr *tb[IFLA_MAX+1];
1845 struct net_device *dev = NULL;
1846 struct sk_buff *nskb;
1849 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1853 if (tb[IFLA_IFNAME])
1854 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1856 ifm = nlmsg_data(nlh);
1857 if (ifm->ifi_index > 0)
1858 dev = __dev_get_by_index(net, ifm->ifi_index);
1859 else if (tb[IFLA_IFNAME])
1860 dev = __dev_get_by_name(net, ifname);
1867 nskb = nlmsg_new(if_nlmsg_size(dev), GFP_KERNEL);
1871 err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).pid,
1872 nlh->nlmsg_seq, 0, 0);
1874 /* -EMSGSIZE implies BUG in if_nlmsg_size */
1875 WARN_ON(err == -EMSGSIZE);
1878 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).pid);
1883 static u16 rtnl_calcit(struct sk_buff *skb)
1885 return min_ifinfo_dump_size;
1888 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
1891 int s_idx = cb->family;
1895 for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
1896 int type = cb->nlh->nlmsg_type-RTM_BASE;
1897 if (idx < s_idx || idx == PF_PACKET)
1899 if (rtnl_msg_handlers[idx] == NULL ||
1900 rtnl_msg_handlers[idx][type].dumpit == NULL)
1903 memset(&cb->args[0], 0, sizeof(cb->args));
1904 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
1912 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned change)
1914 struct net *net = dev_net(dev);
1915 struct sk_buff *skb;
1917 size_t if_info_size;
1919 skb = nlmsg_new((if_info_size = if_nlmsg_size(dev)), GFP_KERNEL);
1923 min_ifinfo_dump_size = max_t(u16, if_info_size, min_ifinfo_dump_size);
1925 err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0);
1927 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
1928 WARN_ON(err == -EMSGSIZE);
1932 rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_KERNEL);
1936 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
1939 /* Protected by RTNL sempahore. */
1940 static struct rtattr **rta_buf;
1941 static int rtattr_max;
1943 /* Process one rtnetlink message. */
1945 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
1947 struct net *net = sock_net(skb->sk);
1948 rtnl_doit_func doit;
1955 type = nlh->nlmsg_type;
1961 /* All the messages must have at least 1 byte length */
1962 if (nlh->nlmsg_len < NLMSG_LENGTH(sizeof(struct rtgenmsg)))
1965 family = ((struct rtgenmsg *)NLMSG_DATA(nlh))->rtgen_family;
1969 if (kind != 2 && security_netlink_recv(skb, CAP_NET_ADMIN))
1972 if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
1974 rtnl_dumpit_func dumpit;
1975 rtnl_calcit_func calcit;
1976 u16 min_dump_alloc = 0;
1978 dumpit = rtnl_get_dumpit(family, type);
1981 calcit = rtnl_get_calcit(family, type);
1983 min_dump_alloc = calcit(skb);
1987 err = netlink_dump_start(rtnl, skb, nlh, dumpit,
1988 NULL, min_dump_alloc);
1993 memset(rta_buf, 0, (rtattr_max * sizeof(struct rtattr *)));
1995 min_len = rtm_min[sz_idx];
1996 if (nlh->nlmsg_len < min_len)
1999 if (nlh->nlmsg_len > min_len) {
2000 int attrlen = nlh->nlmsg_len - NLMSG_ALIGN(min_len);
2001 struct rtattr *attr = (void *)nlh + NLMSG_ALIGN(min_len);
2003 while (RTA_OK(attr, attrlen)) {
2004 unsigned flavor = attr->rta_type;
2006 if (flavor > rta_max[sz_idx])
2008 rta_buf[flavor-1] = attr;
2010 attr = RTA_NEXT(attr, attrlen);
2014 doit = rtnl_get_doit(family, type);
2018 return doit(skb, nlh, (void *)&rta_buf[0]);
2021 static void rtnetlink_rcv(struct sk_buff *skb)
2024 netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
2028 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
2030 struct net_device *dev = ptr;
2036 case NETDEV_POST_INIT:
2037 case NETDEV_REGISTER:
2039 case NETDEV_PRE_TYPE_CHANGE:
2040 case NETDEV_GOING_DOWN:
2041 case NETDEV_UNREGISTER:
2042 case NETDEV_UNREGISTER_BATCH:
2043 case NETDEV_RELEASE:
2047 rtmsg_ifinfo(RTM_NEWLINK, dev, 0);
2053 static struct notifier_block rtnetlink_dev_notifier = {
2054 .notifier_call = rtnetlink_event,
2058 static int __net_init rtnetlink_net_init(struct net *net)
2061 sk = netlink_kernel_create(net, NETLINK_ROUTE, RTNLGRP_MAX,
2062 rtnetlink_rcv, &rtnl_mutex, THIS_MODULE);
2069 static void __net_exit rtnetlink_net_exit(struct net *net)
2071 netlink_kernel_release(net->rtnl);
2075 static struct pernet_operations rtnetlink_net_ops = {
2076 .init = rtnetlink_net_init,
2077 .exit = rtnetlink_net_exit,
2080 void __init rtnetlink_init(void)
2085 for (i = 0; i < ARRAY_SIZE(rta_max); i++)
2086 if (rta_max[i] > rtattr_max)
2087 rtattr_max = rta_max[i];
2088 rta_buf = kmalloc(rtattr_max * sizeof(struct rtattr *), GFP_KERNEL);
2090 panic("rtnetlink_init: cannot allocate rta_buf\n");
2092 if (register_pernet_subsys(&rtnetlink_net_ops))
2093 panic("rtnetlink_init: cannot initialize rtnetlink\n");
2095 netlink_set_nonroot(NETLINK_ROUTE, NL_NONROOT_RECV);
2096 register_netdevice_notifier(&rtnetlink_dev_notifier);
2098 rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
2099 rtnl_dump_ifinfo, rtnl_calcit);
2100 rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
2101 rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
2102 rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
2104 rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
2105 rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);