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);
276 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
278 const struct rtnl_link_ops *ops;
280 list_for_each_entry(ops, &link_ops, list) {
281 if (!strcmp(ops->kind, kind))
288 * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
289 * @ops: struct rtnl_link_ops * to register
291 * The caller must hold the rtnl_mutex. This function should be used
292 * by drivers that create devices during module initialization. It
293 * must be called before registering the devices.
295 * Returns 0 on success or a negative error code.
297 int __rtnl_link_register(struct rtnl_link_ops *ops)
299 if (rtnl_link_ops_get(ops->kind))
303 ops->dellink = unregister_netdevice_queue;
305 list_add_tail(&ops->list, &link_ops);
308 EXPORT_SYMBOL_GPL(__rtnl_link_register);
311 * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
312 * @ops: struct rtnl_link_ops * to register
314 * Returns 0 on success or a negative error code.
316 int rtnl_link_register(struct rtnl_link_ops *ops)
321 err = __rtnl_link_register(ops);
325 EXPORT_SYMBOL_GPL(rtnl_link_register);
327 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
329 struct net_device *dev;
330 LIST_HEAD(list_kill);
332 for_each_netdev(net, dev) {
333 if (dev->rtnl_link_ops == ops)
334 ops->dellink(dev, &list_kill);
336 unregister_netdevice_many(&list_kill);
340 * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
341 * @ops: struct rtnl_link_ops * to unregister
343 * The caller must hold the rtnl_mutex.
345 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
350 __rtnl_kill_links(net, ops);
352 list_del(&ops->list);
354 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
357 * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
358 * @ops: struct rtnl_link_ops * to unregister
360 void rtnl_link_unregister(struct rtnl_link_ops *ops)
363 __rtnl_link_unregister(ops);
366 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
368 static size_t rtnl_link_get_size(const struct net_device *dev)
370 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
376 size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
377 nla_total_size(strlen(ops->kind) + 1); /* IFLA_INFO_KIND */
380 /* IFLA_INFO_DATA + nested data */
381 size += nla_total_size(sizeof(struct nlattr)) +
384 if (ops->get_xstats_size)
385 /* IFLA_INFO_XSTATS */
386 size += nla_total_size(ops->get_xstats_size(dev));
391 static LIST_HEAD(rtnl_af_ops);
393 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
395 const struct rtnl_af_ops *ops;
397 list_for_each_entry(ops, &rtnl_af_ops, list) {
398 if (ops->family == family)
406 * __rtnl_af_register - Register rtnl_af_ops with rtnetlink.
407 * @ops: struct rtnl_af_ops * to register
409 * The caller must hold the rtnl_mutex.
411 * Returns 0 on success or a negative error code.
413 int __rtnl_af_register(struct rtnl_af_ops *ops)
415 list_add_tail(&ops->list, &rtnl_af_ops);
418 EXPORT_SYMBOL_GPL(__rtnl_af_register);
421 * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
422 * @ops: struct rtnl_af_ops * to register
424 * Returns 0 on success or a negative error code.
426 int rtnl_af_register(struct rtnl_af_ops *ops)
431 err = __rtnl_af_register(ops);
435 EXPORT_SYMBOL_GPL(rtnl_af_register);
438 * __rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
439 * @ops: struct rtnl_af_ops * to unregister
441 * The caller must hold the rtnl_mutex.
443 void __rtnl_af_unregister(struct rtnl_af_ops *ops)
445 list_del(&ops->list);
447 EXPORT_SYMBOL_GPL(__rtnl_af_unregister);
450 * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
451 * @ops: struct rtnl_af_ops * to unregister
453 void rtnl_af_unregister(struct rtnl_af_ops *ops)
456 __rtnl_af_unregister(ops);
459 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
461 static size_t rtnl_link_get_af_size(const struct net_device *dev)
463 struct rtnl_af_ops *af_ops;
467 size = nla_total_size(sizeof(struct nlattr));
469 list_for_each_entry(af_ops, &rtnl_af_ops, list) {
470 if (af_ops->get_link_af_size) {
471 /* AF_* + nested data */
472 size += nla_total_size(sizeof(struct nlattr)) +
473 af_ops->get_link_af_size(dev);
480 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
482 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
483 struct nlattr *linkinfo, *data;
486 linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
487 if (linkinfo == NULL)
490 if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
491 goto err_cancel_link;
492 if (ops->fill_xstats) {
493 err = ops->fill_xstats(skb, dev);
495 goto err_cancel_link;
497 if (ops->fill_info) {
498 data = nla_nest_start(skb, IFLA_INFO_DATA);
500 goto err_cancel_link;
501 err = ops->fill_info(skb, dev);
503 goto err_cancel_data;
504 nla_nest_end(skb, data);
507 nla_nest_end(skb, linkinfo);
511 nla_nest_cancel(skb, data);
513 nla_nest_cancel(skb, linkinfo);
518 static const int rtm_min[RTM_NR_FAMILIES] =
520 [RTM_FAM(RTM_NEWLINK)] = NLMSG_LENGTH(sizeof(struct ifinfomsg)),
521 [RTM_FAM(RTM_NEWADDR)] = NLMSG_LENGTH(sizeof(struct ifaddrmsg)),
522 [RTM_FAM(RTM_NEWROUTE)] = NLMSG_LENGTH(sizeof(struct rtmsg)),
523 [RTM_FAM(RTM_NEWRULE)] = NLMSG_LENGTH(sizeof(struct fib_rule_hdr)),
524 [RTM_FAM(RTM_NEWQDISC)] = NLMSG_LENGTH(sizeof(struct tcmsg)),
525 [RTM_FAM(RTM_NEWTCLASS)] = NLMSG_LENGTH(sizeof(struct tcmsg)),
526 [RTM_FAM(RTM_NEWTFILTER)] = NLMSG_LENGTH(sizeof(struct tcmsg)),
527 [RTM_FAM(RTM_NEWACTION)] = NLMSG_LENGTH(sizeof(struct tcamsg)),
528 [RTM_FAM(RTM_GETMULTICAST)] = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
529 [RTM_FAM(RTM_GETANYCAST)] = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
532 static const int rta_max[RTM_NR_FAMILIES] =
534 [RTM_FAM(RTM_NEWLINK)] = IFLA_MAX,
535 [RTM_FAM(RTM_NEWADDR)] = IFA_MAX,
536 [RTM_FAM(RTM_NEWROUTE)] = RTA_MAX,
537 [RTM_FAM(RTM_NEWRULE)] = FRA_MAX,
538 [RTM_FAM(RTM_NEWQDISC)] = TCA_MAX,
539 [RTM_FAM(RTM_NEWTCLASS)] = TCA_MAX,
540 [RTM_FAM(RTM_NEWTFILTER)] = TCA_MAX,
541 [RTM_FAM(RTM_NEWACTION)] = TCAA_MAX,
544 void __rta_fill(struct sk_buff *skb, int attrtype, int attrlen, const void *data)
547 int size = RTA_LENGTH(attrlen);
549 rta = (struct rtattr *)skb_put(skb, RTA_ALIGN(size));
550 rta->rta_type = attrtype;
552 memcpy(RTA_DATA(rta), data, attrlen);
553 memset(RTA_DATA(rta) + attrlen, 0, RTA_ALIGN(size) - size);
555 EXPORT_SYMBOL(__rta_fill);
557 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned group, int echo)
559 struct sock *rtnl = net->rtnl;
562 NETLINK_CB(skb).dst_group = group;
564 atomic_inc(&skb->users);
565 netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
567 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
571 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
573 struct sock *rtnl = net->rtnl;
575 return nlmsg_unicast(rtnl, skb, pid);
577 EXPORT_SYMBOL(rtnl_unicast);
579 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
580 struct nlmsghdr *nlh, gfp_t flags)
582 struct sock *rtnl = net->rtnl;
586 report = nlmsg_report(nlh);
588 nlmsg_notify(rtnl, skb, pid, group, report, flags);
590 EXPORT_SYMBOL(rtnl_notify);
592 void rtnl_set_sk_err(struct net *net, u32 group, int error)
594 struct sock *rtnl = net->rtnl;
596 netlink_set_err(rtnl, 0, group, error);
598 EXPORT_SYMBOL(rtnl_set_sk_err);
600 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
605 mx = nla_nest_start(skb, RTA_METRICS);
609 for (i = 0; i < RTAX_MAX; i++) {
612 NLA_PUT_U32(skb, i+1, metrics[i]);
617 nla_nest_cancel(skb, mx);
621 return nla_nest_end(skb, mx);
624 nla_nest_cancel(skb, mx);
627 EXPORT_SYMBOL(rtnetlink_put_metrics);
629 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
630 u32 ts, u32 tsage, long expires, u32 error)
632 struct rta_cacheinfo ci = {
633 .rta_lastuse = jiffies_to_clock_t(jiffies - dst->lastuse),
634 .rta_used = dst->__use,
635 .rta_clntref = atomic_read(&(dst->__refcnt)),
643 ci.rta_expires = jiffies_to_clock_t(expires);
645 return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
647 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
649 static void set_operstate(struct net_device *dev, unsigned char transition)
651 unsigned char operstate = dev->operstate;
653 switch (transition) {
655 if ((operstate == IF_OPER_DORMANT ||
656 operstate == IF_OPER_UNKNOWN) &&
658 operstate = IF_OPER_UP;
661 case IF_OPER_DORMANT:
662 if (operstate == IF_OPER_UP ||
663 operstate == IF_OPER_UNKNOWN)
664 operstate = IF_OPER_DORMANT;
668 if (dev->operstate != operstate) {
669 write_lock_bh(&dev_base_lock);
670 dev->operstate = operstate;
671 write_unlock_bh(&dev_base_lock);
672 netdev_state_change(dev);
676 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
677 const struct ifinfomsg *ifm)
679 unsigned int flags = ifm->ifi_flags;
681 /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
683 flags = (flags & ifm->ifi_change) |
684 (dev->flags & ~ifm->ifi_change);
689 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
690 const struct rtnl_link_stats64 *b)
692 a->rx_packets = b->rx_packets;
693 a->tx_packets = b->tx_packets;
694 a->rx_bytes = b->rx_bytes;
695 a->tx_bytes = b->tx_bytes;
696 a->rx_errors = b->rx_errors;
697 a->tx_errors = b->tx_errors;
698 a->rx_dropped = b->rx_dropped;
699 a->tx_dropped = b->tx_dropped;
701 a->multicast = b->multicast;
702 a->collisions = b->collisions;
704 a->rx_length_errors = b->rx_length_errors;
705 a->rx_over_errors = b->rx_over_errors;
706 a->rx_crc_errors = b->rx_crc_errors;
707 a->rx_frame_errors = b->rx_frame_errors;
708 a->rx_fifo_errors = b->rx_fifo_errors;
709 a->rx_missed_errors = b->rx_missed_errors;
711 a->tx_aborted_errors = b->tx_aborted_errors;
712 a->tx_carrier_errors = b->tx_carrier_errors;
713 a->tx_fifo_errors = b->tx_fifo_errors;
714 a->tx_heartbeat_errors = b->tx_heartbeat_errors;
715 a->tx_window_errors = b->tx_window_errors;
717 a->rx_compressed = b->rx_compressed;
718 a->tx_compressed = b->tx_compressed;
721 static void copy_rtnl_link_stats64(void *v, const struct rtnl_link_stats64 *b)
723 memcpy(v, b, sizeof(*b));
727 static inline int rtnl_vfinfo_size(const struct net_device *dev)
729 if (dev->dev.parent && dev_is_pci(dev->dev.parent)) {
731 int num_vfs = dev_num_vf(dev->dev.parent);
732 size_t size = nla_total_size(sizeof(struct nlattr));
733 size += nla_total_size(num_vfs * sizeof(struct nlattr));
735 (nla_total_size(sizeof(struct ifla_vf_mac)) +
736 nla_total_size(sizeof(struct ifla_vf_vlan)) +
737 nla_total_size(sizeof(struct ifla_vf_tx_rate)) +
738 nla_total_size(sizeof(struct ifla_vf_spoofchk)));
744 static size_t rtnl_port_size(const struct net_device *dev)
746 size_t port_size = nla_total_size(4) /* PORT_VF */
747 + nla_total_size(PORT_PROFILE_MAX) /* PORT_PROFILE */
748 + nla_total_size(sizeof(struct ifla_port_vsi))
750 + nla_total_size(PORT_UUID_MAX) /* PORT_INSTANCE_UUID */
751 + nla_total_size(PORT_UUID_MAX) /* PORT_HOST_UUID */
752 + nla_total_size(1) /* PROT_VDP_REQUEST */
753 + nla_total_size(2); /* PORT_VDP_RESPONSE */
754 size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
755 size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
757 size_t port_self_size = nla_total_size(sizeof(struct nlattr))
760 if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent)
762 if (dev_num_vf(dev->dev.parent))
763 return port_self_size + vf_ports_size +
764 vf_port_size * dev_num_vf(dev->dev.parent);
766 return port_self_size;
769 static noinline size_t if_nlmsg_size(const struct net_device *dev)
771 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
772 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
773 + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
774 + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
775 + nla_total_size(sizeof(struct rtnl_link_ifmap))
776 + nla_total_size(sizeof(struct rtnl_link_stats))
777 + nla_total_size(sizeof(struct rtnl_link_stats64))
778 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
779 + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
780 + nla_total_size(4) /* IFLA_TXQLEN */
781 + nla_total_size(4) /* IFLA_WEIGHT */
782 + nla_total_size(4) /* IFLA_MTU */
783 + nla_total_size(4) /* IFLA_LINK */
784 + nla_total_size(4) /* IFLA_MASTER */
785 + nla_total_size(1) /* IFLA_OPERSTATE */
786 + nla_total_size(1) /* IFLA_LINKMODE */
787 + nla_total_size(4) /* IFLA_NUM_VF */
788 + rtnl_vfinfo_size(dev) /* IFLA_VFINFO_LIST */
789 + rtnl_port_size(dev) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
790 + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
791 + rtnl_link_get_af_size(dev); /* IFLA_AF_SPEC */
794 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
796 struct nlattr *vf_ports;
797 struct nlattr *vf_port;
801 vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
805 for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
806 vf_port = nla_nest_start(skb, IFLA_VF_PORT);
808 goto nla_put_failure;
809 NLA_PUT_U32(skb, IFLA_PORT_VF, vf);
810 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
811 if (err == -EMSGSIZE)
812 goto nla_put_failure;
814 nla_nest_cancel(skb, vf_port);
817 nla_nest_end(skb, vf_port);
820 nla_nest_end(skb, vf_ports);
825 nla_nest_cancel(skb, vf_ports);
829 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
831 struct nlattr *port_self;
834 port_self = nla_nest_start(skb, IFLA_PORT_SELF);
838 err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
840 nla_nest_cancel(skb, port_self);
841 return (err == -EMSGSIZE) ? err : 0;
844 nla_nest_end(skb, port_self);
849 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev)
853 if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent)
856 err = rtnl_port_self_fill(skb, dev);
860 if (dev_num_vf(dev->dev.parent)) {
861 err = rtnl_vf_ports_fill(skb, dev);
869 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
870 int type, u32 pid, u32 seq, u32 change,
873 struct ifinfomsg *ifm;
874 struct nlmsghdr *nlh;
875 struct rtnl_link_stats64 temp;
876 const struct rtnl_link_stats64 *stats;
877 struct nlattr *attr, *af_spec;
878 struct rtnl_af_ops *af_ops;
881 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
885 ifm = nlmsg_data(nlh);
886 ifm->ifi_family = AF_UNSPEC;
888 ifm->ifi_type = dev->type;
889 ifm->ifi_index = dev->ifindex;
890 ifm->ifi_flags = dev_get_flags(dev);
891 ifm->ifi_change = change;
893 NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
894 NLA_PUT_U32(skb, IFLA_TXQLEN, dev->tx_queue_len);
895 NLA_PUT_U8(skb, IFLA_OPERSTATE,
896 netif_running(dev) ? dev->operstate : IF_OPER_DOWN);
897 NLA_PUT_U8(skb, IFLA_LINKMODE, dev->link_mode);
898 NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
899 NLA_PUT_U32(skb, IFLA_GROUP, dev->group);
901 if (dev->ifindex != dev->iflink)
902 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
905 NLA_PUT_U32(skb, IFLA_MASTER, dev->master->ifindex);
908 NLA_PUT_STRING(skb, IFLA_QDISC, dev->qdisc->ops->id);
911 NLA_PUT_STRING(skb, IFLA_IFALIAS, dev->ifalias);
914 struct rtnl_link_ifmap map = {
915 .mem_start = dev->mem_start,
916 .mem_end = dev->mem_end,
917 .base_addr = dev->base_addr,
920 .port = dev->if_port,
922 NLA_PUT(skb, IFLA_MAP, sizeof(map), &map);
926 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
927 NLA_PUT(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast);
930 attr = nla_reserve(skb, IFLA_STATS,
931 sizeof(struct rtnl_link_stats));
933 goto nla_put_failure;
935 stats = dev_get_stats(dev, &temp);
936 copy_rtnl_link_stats(nla_data(attr), stats);
938 attr = nla_reserve(skb, IFLA_STATS64,
939 sizeof(struct rtnl_link_stats64));
941 goto nla_put_failure;
942 copy_rtnl_link_stats64(nla_data(attr), stats);
945 NLA_PUT_U32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent));
947 if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent) {
950 struct nlattr *vfinfo, *vf;
951 int num_vfs = dev_num_vf(dev->dev.parent);
953 vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
955 goto nla_put_failure;
956 for (i = 0; i < num_vfs; i++) {
957 struct ifla_vf_info ivi;
958 struct ifla_vf_mac vf_mac;
959 struct ifla_vf_vlan vf_vlan;
960 struct ifla_vf_tx_rate vf_tx_rate;
961 struct ifla_vf_spoofchk vf_spoofchk;
964 * Not all SR-IOV capable drivers support the
965 * spoofcheck query. Preset to -1 so the user
966 * space tool can detect that the driver didn't
970 if (dev->netdev_ops->ndo_get_vf_config(dev, i, &ivi))
975 vf_spoofchk.vf = ivi.vf;
977 memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
978 vf_vlan.vlan = ivi.vlan;
979 vf_vlan.qos = ivi.qos;
980 vf_tx_rate.rate = ivi.tx_rate;
981 vf_spoofchk.setting = ivi.spoofchk;
982 vf = nla_nest_start(skb, IFLA_VF_INFO);
984 nla_nest_cancel(skb, vfinfo);
985 goto nla_put_failure;
987 NLA_PUT(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac);
988 NLA_PUT(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan);
989 NLA_PUT(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
991 NLA_PUT(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
993 nla_nest_end(skb, vf);
995 nla_nest_end(skb, vfinfo);
998 if (rtnl_port_fill(skb, dev))
999 goto nla_put_failure;
1001 if (dev->rtnl_link_ops) {
1002 if (rtnl_link_fill(skb, dev) < 0)
1003 goto nla_put_failure;
1006 if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1007 goto nla_put_failure;
1009 list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1010 if (af_ops->fill_link_af) {
1014 if (!(af = nla_nest_start(skb, af_ops->family)))
1015 goto nla_put_failure;
1017 err = af_ops->fill_link_af(skb, dev);
1020 * Caller may return ENODATA to indicate that there
1021 * was no data to be dumped. This is not an error, it
1022 * means we should trim the attribute header and
1025 if (err == -ENODATA)
1026 nla_nest_cancel(skb, af);
1028 goto nla_put_failure;
1030 nla_nest_end(skb, af);
1034 nla_nest_end(skb, af_spec);
1036 return nlmsg_end(skb, nlh);
1039 nlmsg_cancel(skb, nlh);
1043 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1045 struct net *net = sock_net(skb->sk);
1048 struct net_device *dev;
1049 struct hlist_head *head;
1050 struct hlist_node *node;
1053 s_idx = cb->args[1];
1056 cb->seq = net->dev_base_seq;
1058 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1060 head = &net->dev_index_head[h];
1061 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
1064 if (rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1065 NETLINK_CB(cb->skb).pid,
1066 cb->nlh->nlmsg_seq, 0,
1070 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1083 const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1084 [IFLA_IFNAME] = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1085 [IFLA_ADDRESS] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1086 [IFLA_BROADCAST] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1087 [IFLA_MAP] = { .len = sizeof(struct rtnl_link_ifmap) },
1088 [IFLA_MTU] = { .type = NLA_U32 },
1089 [IFLA_LINK] = { .type = NLA_U32 },
1090 [IFLA_MASTER] = { .type = NLA_U32 },
1091 [IFLA_TXQLEN] = { .type = NLA_U32 },
1092 [IFLA_WEIGHT] = { .type = NLA_U32 },
1093 [IFLA_OPERSTATE] = { .type = NLA_U8 },
1094 [IFLA_LINKMODE] = { .type = NLA_U8 },
1095 [IFLA_LINKINFO] = { .type = NLA_NESTED },
1096 [IFLA_NET_NS_PID] = { .type = NLA_U32 },
1097 [IFLA_NET_NS_FD] = { .type = NLA_U32 },
1098 [IFLA_IFALIAS] = { .type = NLA_STRING, .len = IFALIASZ-1 },
1099 [IFLA_VFINFO_LIST] = {. type = NLA_NESTED },
1100 [IFLA_VF_PORTS] = { .type = NLA_NESTED },
1101 [IFLA_PORT_SELF] = { .type = NLA_NESTED },
1102 [IFLA_AF_SPEC] = { .type = NLA_NESTED },
1104 EXPORT_SYMBOL(ifla_policy);
1106 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1107 [IFLA_INFO_KIND] = { .type = NLA_STRING },
1108 [IFLA_INFO_DATA] = { .type = NLA_NESTED },
1111 static const struct nla_policy ifla_vfinfo_policy[IFLA_VF_INFO_MAX+1] = {
1112 [IFLA_VF_INFO] = { .type = NLA_NESTED },
1115 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1116 [IFLA_VF_MAC] = { .type = NLA_BINARY,
1117 .len = sizeof(struct ifla_vf_mac) },
1118 [IFLA_VF_VLAN] = { .type = NLA_BINARY,
1119 .len = sizeof(struct ifla_vf_vlan) },
1120 [IFLA_VF_TX_RATE] = { .type = NLA_BINARY,
1121 .len = sizeof(struct ifla_vf_tx_rate) },
1124 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1125 [IFLA_PORT_VF] = { .type = NLA_U32 },
1126 [IFLA_PORT_PROFILE] = { .type = NLA_STRING,
1127 .len = PORT_PROFILE_MAX },
1128 [IFLA_PORT_VSI_TYPE] = { .type = NLA_BINARY,
1129 .len = sizeof(struct ifla_port_vsi)},
1130 [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1131 .len = PORT_UUID_MAX },
1132 [IFLA_PORT_HOST_UUID] = { .type = NLA_STRING,
1133 .len = PORT_UUID_MAX },
1134 [IFLA_PORT_REQUEST] = { .type = NLA_U8, },
1135 [IFLA_PORT_RESPONSE] = { .type = NLA_U16, },
1138 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1141 /* Examine the link attributes and figure out which
1142 * network namespace we are talking about.
1144 if (tb[IFLA_NET_NS_PID])
1145 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1146 else if (tb[IFLA_NET_NS_FD])
1147 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1149 net = get_net(src_net);
1152 EXPORT_SYMBOL(rtnl_link_get_net);
1154 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1157 if (tb[IFLA_ADDRESS] &&
1158 nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1161 if (tb[IFLA_BROADCAST] &&
1162 nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1166 if (tb[IFLA_AF_SPEC]) {
1170 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1171 const struct rtnl_af_ops *af_ops;
1173 if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1174 return -EAFNOSUPPORT;
1176 if (!af_ops->set_link_af)
1179 if (af_ops->validate_link_af) {
1180 err = af_ops->validate_link_af(dev, af);
1190 static int do_setvfinfo(struct net_device *dev, struct nlattr *attr)
1192 int rem, err = -EINVAL;
1194 const struct net_device_ops *ops = dev->netdev_ops;
1196 nla_for_each_nested(vf, attr, rem) {
1197 switch (nla_type(vf)) {
1199 struct ifla_vf_mac *ivm;
1202 if (ops->ndo_set_vf_mac)
1203 err = ops->ndo_set_vf_mac(dev, ivm->vf,
1207 case IFLA_VF_VLAN: {
1208 struct ifla_vf_vlan *ivv;
1211 if (ops->ndo_set_vf_vlan)
1212 err = ops->ndo_set_vf_vlan(dev, ivv->vf,
1217 case IFLA_VF_TX_RATE: {
1218 struct ifla_vf_tx_rate *ivt;
1221 if (ops->ndo_set_vf_tx_rate)
1222 err = ops->ndo_set_vf_tx_rate(dev, ivt->vf,
1226 case IFLA_VF_SPOOFCHK: {
1227 struct ifla_vf_spoofchk *ivs;
1230 if (ops->ndo_set_vf_spoofchk)
1231 err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1245 static int do_set_master(struct net_device *dev, int ifindex)
1247 struct net_device *master_dev;
1248 const struct net_device_ops *ops;
1252 if (dev->master->ifindex == ifindex)
1254 ops = dev->master->netdev_ops;
1255 if (ops->ndo_del_slave) {
1256 err = ops->ndo_del_slave(dev->master, dev);
1265 master_dev = __dev_get_by_index(dev_net(dev), ifindex);
1268 ops = master_dev->netdev_ops;
1269 if (ops->ndo_add_slave) {
1270 err = ops->ndo_add_slave(master_dev, dev);
1280 static int do_setlink(struct net_device *dev, struct ifinfomsg *ifm,
1281 struct nlattr **tb, char *ifname, int modified)
1283 const struct net_device_ops *ops = dev->netdev_ops;
1284 int send_addr_notify = 0;
1287 if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1288 struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1293 err = dev_change_net_namespace(dev, net, ifname);
1301 struct rtnl_link_ifmap *u_map;
1304 if (!ops->ndo_set_config) {
1309 if (!netif_device_present(dev)) {
1314 u_map = nla_data(tb[IFLA_MAP]);
1315 k_map.mem_start = (unsigned long) u_map->mem_start;
1316 k_map.mem_end = (unsigned long) u_map->mem_end;
1317 k_map.base_addr = (unsigned short) u_map->base_addr;
1318 k_map.irq = (unsigned char) u_map->irq;
1319 k_map.dma = (unsigned char) u_map->dma;
1320 k_map.port = (unsigned char) u_map->port;
1322 err = ops->ndo_set_config(dev, &k_map);
1329 if (tb[IFLA_ADDRESS]) {
1330 struct sockaddr *sa;
1333 if (!ops->ndo_set_mac_address) {
1338 if (!netif_device_present(dev)) {
1343 len = sizeof(sa_family_t) + dev->addr_len;
1344 sa = kmalloc(len, GFP_KERNEL);
1349 sa->sa_family = dev->type;
1350 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1352 err = ops->ndo_set_mac_address(dev, sa);
1356 send_addr_notify = 1;
1361 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1367 if (tb[IFLA_GROUP]) {
1368 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1373 * Interface selected by interface index but interface
1374 * name provided implies that a name change has been
1377 if (ifm->ifi_index > 0 && ifname[0]) {
1378 err = dev_change_name(dev, ifname);
1384 if (tb[IFLA_IFALIAS]) {
1385 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
1386 nla_len(tb[IFLA_IFALIAS]));
1392 if (tb[IFLA_BROADCAST]) {
1393 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
1394 send_addr_notify = 1;
1397 if (ifm->ifi_flags || ifm->ifi_change) {
1398 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1403 if (tb[IFLA_MASTER]) {
1404 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
1410 if (tb[IFLA_TXQLEN])
1411 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1413 if (tb[IFLA_OPERSTATE])
1414 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1416 if (tb[IFLA_LINKMODE]) {
1417 write_lock_bh(&dev_base_lock);
1418 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1419 write_unlock_bh(&dev_base_lock);
1422 if (tb[IFLA_VFINFO_LIST]) {
1423 struct nlattr *attr;
1425 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
1426 if (nla_type(attr) != IFLA_VF_INFO) {
1430 err = do_setvfinfo(dev, attr);
1438 if (tb[IFLA_VF_PORTS]) {
1439 struct nlattr *port[IFLA_PORT_MAX+1];
1440 struct nlattr *attr;
1445 if (!ops->ndo_set_vf_port)
1448 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
1449 if (nla_type(attr) != IFLA_VF_PORT)
1451 err = nla_parse_nested(port, IFLA_PORT_MAX,
1452 attr, ifla_port_policy);
1455 if (!port[IFLA_PORT_VF]) {
1459 vf = nla_get_u32(port[IFLA_PORT_VF]);
1460 err = ops->ndo_set_vf_port(dev, vf, port);
1468 if (tb[IFLA_PORT_SELF]) {
1469 struct nlattr *port[IFLA_PORT_MAX+1];
1471 err = nla_parse_nested(port, IFLA_PORT_MAX,
1472 tb[IFLA_PORT_SELF], ifla_port_policy);
1477 if (ops->ndo_set_vf_port)
1478 err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
1484 if (tb[IFLA_AF_SPEC]) {
1488 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1489 const struct rtnl_af_ops *af_ops;
1491 if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1494 err = af_ops->set_link_af(dev, af);
1504 if (err < 0 && modified && net_ratelimit())
1505 printk(KERN_WARNING "A link change request failed with "
1506 "some changes committed already. Interface %s may "
1507 "have been left with an inconsistent configuration, "
1508 "please check.\n", dev->name);
1510 if (send_addr_notify)
1511 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
1515 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1517 struct net *net = sock_net(skb->sk);
1518 struct ifinfomsg *ifm;
1519 struct net_device *dev;
1521 struct nlattr *tb[IFLA_MAX+1];
1522 char ifname[IFNAMSIZ];
1524 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1528 if (tb[IFLA_IFNAME])
1529 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1534 ifm = nlmsg_data(nlh);
1535 if (ifm->ifi_index > 0)
1536 dev = __dev_get_by_index(net, ifm->ifi_index);
1537 else if (tb[IFLA_IFNAME])
1538 dev = __dev_get_by_name(net, ifname);
1547 err = validate_linkmsg(dev, tb);
1551 err = do_setlink(dev, ifm, tb, ifname, 0);
1556 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1558 struct net *net = sock_net(skb->sk);
1559 const struct rtnl_link_ops *ops;
1560 struct net_device *dev;
1561 struct ifinfomsg *ifm;
1562 char ifname[IFNAMSIZ];
1563 struct nlattr *tb[IFLA_MAX+1];
1565 LIST_HEAD(list_kill);
1567 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1571 if (tb[IFLA_IFNAME])
1572 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1574 ifm = nlmsg_data(nlh);
1575 if (ifm->ifi_index > 0)
1576 dev = __dev_get_by_index(net, ifm->ifi_index);
1577 else if (tb[IFLA_IFNAME])
1578 dev = __dev_get_by_name(net, ifname);
1585 ops = dev->rtnl_link_ops;
1589 ops->dellink(dev, &list_kill);
1590 unregister_netdevice_many(&list_kill);
1591 list_del(&list_kill);
1595 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
1597 unsigned int old_flags;
1600 old_flags = dev->flags;
1601 if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
1602 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1607 dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
1608 rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U);
1610 __dev_notify_flags(dev, old_flags);
1613 EXPORT_SYMBOL(rtnl_configure_link);
1615 struct net_device *rtnl_create_link(struct net *src_net, struct net *net,
1616 char *ifname, const struct rtnl_link_ops *ops, struct nlattr *tb[])
1619 struct net_device *dev;
1620 unsigned int num_queues = 1;
1621 unsigned int real_num_queues = 1;
1623 if (ops->get_tx_queues) {
1624 err = ops->get_tx_queues(src_net, tb, &num_queues,
1630 dev = alloc_netdev_mq(ops->priv_size, ifname, ops->setup, num_queues);
1634 dev_net_set(dev, net);
1635 dev->rtnl_link_ops = ops;
1636 dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
1639 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
1640 if (tb[IFLA_ADDRESS])
1641 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
1642 nla_len(tb[IFLA_ADDRESS]));
1643 if (tb[IFLA_BROADCAST])
1644 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
1645 nla_len(tb[IFLA_BROADCAST]));
1646 if (tb[IFLA_TXQLEN])
1647 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1648 if (tb[IFLA_OPERSTATE])
1649 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1650 if (tb[IFLA_LINKMODE])
1651 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1653 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1658 return ERR_PTR(err);
1660 EXPORT_SYMBOL(rtnl_create_link);
1662 static int rtnl_group_changelink(struct net *net, int group,
1663 struct ifinfomsg *ifm,
1666 struct net_device *dev;
1669 for_each_netdev(net, dev) {
1670 if (dev->group == group) {
1671 err = do_setlink(dev, ifm, tb, NULL, 0);
1680 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1682 struct net *net = sock_net(skb->sk);
1683 const struct rtnl_link_ops *ops;
1684 struct net_device *dev;
1685 struct ifinfomsg *ifm;
1686 char kind[MODULE_NAME_LEN];
1687 char ifname[IFNAMSIZ];
1688 struct nlattr *tb[IFLA_MAX+1];
1689 struct nlattr *linkinfo[IFLA_INFO_MAX+1];
1692 #ifdef CONFIG_MODULES
1695 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1699 if (tb[IFLA_IFNAME])
1700 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1704 ifm = nlmsg_data(nlh);
1705 if (ifm->ifi_index > 0)
1706 dev = __dev_get_by_index(net, ifm->ifi_index);
1709 dev = __dev_get_by_name(net, ifname);
1714 err = validate_linkmsg(dev, tb);
1718 if (tb[IFLA_LINKINFO]) {
1719 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
1720 tb[IFLA_LINKINFO], ifla_info_policy);
1724 memset(linkinfo, 0, sizeof(linkinfo));
1726 if (linkinfo[IFLA_INFO_KIND]) {
1727 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
1728 ops = rtnl_link_ops_get(kind);
1735 struct nlattr *attr[ops ? ops->maxtype + 1 : 0], **data = NULL;
1736 struct net *dest_net;
1739 if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
1740 err = nla_parse_nested(attr, ops->maxtype,
1741 linkinfo[IFLA_INFO_DATA],
1747 if (ops->validate) {
1748 err = ops->validate(tb, data);
1757 if (nlh->nlmsg_flags & NLM_F_EXCL)
1759 if (nlh->nlmsg_flags & NLM_F_REPLACE)
1762 if (linkinfo[IFLA_INFO_DATA]) {
1763 if (!ops || ops != dev->rtnl_link_ops ||
1767 err = ops->changelink(dev, tb, data);
1773 return do_setlink(dev, ifm, tb, ifname, modified);
1776 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
1777 if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
1778 return rtnl_group_changelink(net,
1779 nla_get_u32(tb[IFLA_GROUP]),
1786 if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
1790 #ifdef CONFIG_MODULES
1793 request_module("rtnl-link-%s", kind);
1795 ops = rtnl_link_ops_get(kind);
1804 snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
1806 dest_net = rtnl_link_get_net(net, tb);
1807 if (IS_ERR(dest_net))
1808 return PTR_ERR(dest_net);
1810 dev = rtnl_create_link(net, dest_net, ifname, ops, tb);
1814 else if (ops->newlink)
1815 err = ops->newlink(net, dev, tb, data);
1817 err = register_netdevice(dev);
1819 if (err < 0 && !IS_ERR(dev))
1824 err = rtnl_configure_link(dev, ifm);
1826 unregister_netdevice(dev);
1833 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
1835 struct net *net = sock_net(skb->sk);
1836 struct ifinfomsg *ifm;
1837 char ifname[IFNAMSIZ];
1838 struct nlattr *tb[IFLA_MAX+1];
1839 struct net_device *dev = NULL;
1840 struct sk_buff *nskb;
1843 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1847 if (tb[IFLA_IFNAME])
1848 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1850 ifm = nlmsg_data(nlh);
1851 if (ifm->ifi_index > 0)
1852 dev = __dev_get_by_index(net, ifm->ifi_index);
1853 else if (tb[IFLA_IFNAME])
1854 dev = __dev_get_by_name(net, ifname);
1861 nskb = nlmsg_new(if_nlmsg_size(dev), GFP_KERNEL);
1865 err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).pid,
1866 nlh->nlmsg_seq, 0, 0);
1868 /* -EMSGSIZE implies BUG in if_nlmsg_size */
1869 WARN_ON(err == -EMSGSIZE);
1872 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).pid);
1877 static u16 rtnl_calcit(struct sk_buff *skb)
1879 return min_ifinfo_dump_size;
1882 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
1885 int s_idx = cb->family;
1889 for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
1890 int type = cb->nlh->nlmsg_type-RTM_BASE;
1891 if (idx < s_idx || idx == PF_PACKET)
1893 if (rtnl_msg_handlers[idx] == NULL ||
1894 rtnl_msg_handlers[idx][type].dumpit == NULL)
1897 memset(&cb->args[0], 0, sizeof(cb->args));
1898 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
1906 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned change)
1908 struct net *net = dev_net(dev);
1909 struct sk_buff *skb;
1911 size_t if_info_size;
1913 skb = nlmsg_new((if_info_size = if_nlmsg_size(dev)), GFP_KERNEL);
1917 min_ifinfo_dump_size = max_t(u16, if_info_size, min_ifinfo_dump_size);
1919 err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0);
1921 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
1922 WARN_ON(err == -EMSGSIZE);
1926 rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_KERNEL);
1930 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
1933 /* Protected by RTNL sempahore. */
1934 static struct rtattr **rta_buf;
1935 static int rtattr_max;
1937 /* Process one rtnetlink message. */
1939 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
1941 struct net *net = sock_net(skb->sk);
1942 rtnl_doit_func doit;
1949 type = nlh->nlmsg_type;
1955 /* All the messages must have at least 1 byte length */
1956 if (nlh->nlmsg_len < NLMSG_LENGTH(sizeof(struct rtgenmsg)))
1959 family = ((struct rtgenmsg *)NLMSG_DATA(nlh))->rtgen_family;
1963 if (kind != 2 && security_netlink_recv(skb, CAP_NET_ADMIN))
1966 if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
1968 rtnl_dumpit_func dumpit;
1969 rtnl_calcit_func calcit;
1970 u16 min_dump_alloc = 0;
1972 dumpit = rtnl_get_dumpit(family, type);
1975 calcit = rtnl_get_calcit(family, type);
1977 min_dump_alloc = calcit(skb);
1981 err = netlink_dump_start(rtnl, skb, nlh, dumpit,
1982 NULL, min_dump_alloc);
1987 memset(rta_buf, 0, (rtattr_max * sizeof(struct rtattr *)));
1989 min_len = rtm_min[sz_idx];
1990 if (nlh->nlmsg_len < min_len)
1993 if (nlh->nlmsg_len > min_len) {
1994 int attrlen = nlh->nlmsg_len - NLMSG_ALIGN(min_len);
1995 struct rtattr *attr = (void *)nlh + NLMSG_ALIGN(min_len);
1997 while (RTA_OK(attr, attrlen)) {
1998 unsigned flavor = attr->rta_type;
2000 if (flavor > rta_max[sz_idx])
2002 rta_buf[flavor-1] = attr;
2004 attr = RTA_NEXT(attr, attrlen);
2008 doit = rtnl_get_doit(family, type);
2012 return doit(skb, nlh, (void *)&rta_buf[0]);
2015 static void rtnetlink_rcv(struct sk_buff *skb)
2018 netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
2022 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
2024 struct net_device *dev = ptr;
2030 case NETDEV_POST_INIT:
2031 case NETDEV_REGISTER:
2033 case NETDEV_PRE_TYPE_CHANGE:
2034 case NETDEV_GOING_DOWN:
2035 case NETDEV_UNREGISTER:
2036 case NETDEV_UNREGISTER_BATCH:
2037 case NETDEV_RELEASE:
2041 rtmsg_ifinfo(RTM_NEWLINK, dev, 0);
2047 static struct notifier_block rtnetlink_dev_notifier = {
2048 .notifier_call = rtnetlink_event,
2052 static int __net_init rtnetlink_net_init(struct net *net)
2055 sk = netlink_kernel_create(net, NETLINK_ROUTE, RTNLGRP_MAX,
2056 rtnetlink_rcv, &rtnl_mutex, THIS_MODULE);
2063 static void __net_exit rtnetlink_net_exit(struct net *net)
2065 netlink_kernel_release(net->rtnl);
2069 static struct pernet_operations rtnetlink_net_ops = {
2070 .init = rtnetlink_net_init,
2071 .exit = rtnetlink_net_exit,
2074 void __init rtnetlink_init(void)
2079 for (i = 0; i < ARRAY_SIZE(rta_max); i++)
2080 if (rta_max[i] > rtattr_max)
2081 rtattr_max = rta_max[i];
2082 rta_buf = kmalloc(rtattr_max * sizeof(struct rtattr *), GFP_KERNEL);
2084 panic("rtnetlink_init: cannot allocate rta_buf\n");
2086 if (register_pernet_subsys(&rtnetlink_net_ops))
2087 panic("rtnetlink_init: cannot initialize rtnetlink\n");
2089 netlink_set_nonroot(NETLINK_ROUTE, NL_NONROOT_RECV);
2090 register_netdevice_notifier(&rtnetlink_dev_notifier);
2092 rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
2093 rtnl_dump_ifinfo, rtnl_calcit);
2094 rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
2095 rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
2096 rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
2098 rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
2099 rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);