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;
61 static DEFINE_MUTEX(rtnl_mutex);
65 mutex_lock(&rtnl_mutex);
67 EXPORT_SYMBOL(rtnl_lock);
69 void __rtnl_unlock(void)
71 mutex_unlock(&rtnl_mutex);
74 void rtnl_unlock(void)
76 /* This fellow will unlock it for us. */
79 EXPORT_SYMBOL(rtnl_unlock);
81 int rtnl_trylock(void)
83 return mutex_trylock(&rtnl_mutex);
85 EXPORT_SYMBOL(rtnl_trylock);
87 int rtnl_is_locked(void)
89 return mutex_is_locked(&rtnl_mutex);
91 EXPORT_SYMBOL(rtnl_is_locked);
93 #ifdef CONFIG_PROVE_LOCKING
94 int lockdep_rtnl_is_held(void)
96 return lockdep_is_held(&rtnl_mutex);
98 EXPORT_SYMBOL(lockdep_rtnl_is_held);
99 #endif /* #ifdef CONFIG_PROVE_LOCKING */
101 static struct rtnl_link *rtnl_msg_handlers[NPROTO];
103 static inline int rtm_msgindex(int msgtype)
105 int msgindex = msgtype - RTM_BASE;
108 * msgindex < 0 implies someone tried to register a netlink
109 * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
110 * the message type has not been added to linux/rtnetlink.h
112 BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
117 static rtnl_doit_func rtnl_get_doit(int protocol, int msgindex)
119 struct rtnl_link *tab;
121 if (protocol < NPROTO)
122 tab = rtnl_msg_handlers[protocol];
126 if (tab == NULL || tab[msgindex].doit == NULL)
127 tab = rtnl_msg_handlers[PF_UNSPEC];
129 return tab ? tab[msgindex].doit : NULL;
132 static rtnl_dumpit_func rtnl_get_dumpit(int protocol, int msgindex)
134 struct rtnl_link *tab;
136 if (protocol < NPROTO)
137 tab = rtnl_msg_handlers[protocol];
141 if (tab == NULL || tab[msgindex].dumpit == NULL)
142 tab = rtnl_msg_handlers[PF_UNSPEC];
144 return tab ? tab[msgindex].dumpit : NULL;
148 * __rtnl_register - Register a rtnetlink message type
149 * @protocol: Protocol family or PF_UNSPEC
150 * @msgtype: rtnetlink message type
151 * @doit: Function pointer called for each request message
152 * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
154 * Registers the specified function pointers (at least one of them has
155 * to be non-NULL) to be called whenever a request message for the
156 * specified protocol family and message type is received.
158 * The special protocol family PF_UNSPEC may be used to define fallback
159 * function pointers for the case when no entry for the specific protocol
162 * Returns 0 on success or a negative error code.
164 int __rtnl_register(int protocol, int msgtype,
165 rtnl_doit_func doit, rtnl_dumpit_func dumpit)
167 struct rtnl_link *tab;
170 BUG_ON(protocol < 0 || protocol >= NPROTO);
171 msgindex = rtm_msgindex(msgtype);
173 tab = rtnl_msg_handlers[protocol];
175 tab = kcalloc(RTM_NR_MSGTYPES, sizeof(*tab), GFP_KERNEL);
179 rtnl_msg_handlers[protocol] = tab;
183 tab[msgindex].doit = doit;
186 tab[msgindex].dumpit = dumpit;
190 EXPORT_SYMBOL_GPL(__rtnl_register);
193 * rtnl_register - Register a rtnetlink message type
195 * Identical to __rtnl_register() but panics on failure. This is useful
196 * as failure of this function is very unlikely, it can only happen due
197 * to lack of memory when allocating the chain to store all message
198 * handlers for a protocol. Meant for use in init functions where lack
199 * of memory implies no sense in continueing.
201 void rtnl_register(int protocol, int msgtype,
202 rtnl_doit_func doit, rtnl_dumpit_func dumpit)
204 if (__rtnl_register(protocol, msgtype, doit, dumpit) < 0)
205 panic("Unable to register rtnetlink message handler, "
206 "protocol = %d, message type = %d\n",
209 EXPORT_SYMBOL_GPL(rtnl_register);
212 * rtnl_unregister - Unregister a rtnetlink message type
213 * @protocol: Protocol family or PF_UNSPEC
214 * @msgtype: rtnetlink message type
216 * Returns 0 on success or a negative error code.
218 int rtnl_unregister(int protocol, int msgtype)
222 BUG_ON(protocol < 0 || protocol >= NPROTO);
223 msgindex = rtm_msgindex(msgtype);
225 if (rtnl_msg_handlers[protocol] == NULL)
228 rtnl_msg_handlers[protocol][msgindex].doit = NULL;
229 rtnl_msg_handlers[protocol][msgindex].dumpit = NULL;
233 EXPORT_SYMBOL_GPL(rtnl_unregister);
236 * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
237 * @protocol : Protocol family or PF_UNSPEC
239 * Identical to calling rtnl_unregster() for all registered message types
240 * of a certain protocol family.
242 void rtnl_unregister_all(int protocol)
244 BUG_ON(protocol < 0 || protocol >= NPROTO);
246 kfree(rtnl_msg_handlers[protocol]);
247 rtnl_msg_handlers[protocol] = NULL;
249 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
251 static LIST_HEAD(link_ops);
254 * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
255 * @ops: struct rtnl_link_ops * to register
257 * The caller must hold the rtnl_mutex. This function should be used
258 * by drivers that create devices during module initialization. It
259 * must be called before registering the devices.
261 * Returns 0 on success or a negative error code.
263 int __rtnl_link_register(struct rtnl_link_ops *ops)
266 ops->dellink = unregister_netdevice_queue;
268 list_add_tail(&ops->list, &link_ops);
271 EXPORT_SYMBOL_GPL(__rtnl_link_register);
274 * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
275 * @ops: struct rtnl_link_ops * to register
277 * Returns 0 on success or a negative error code.
279 int rtnl_link_register(struct rtnl_link_ops *ops)
284 err = __rtnl_link_register(ops);
288 EXPORT_SYMBOL_GPL(rtnl_link_register);
290 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
292 struct net_device *dev;
293 LIST_HEAD(list_kill);
295 for_each_netdev(net, dev) {
296 if (dev->rtnl_link_ops == ops)
297 ops->dellink(dev, &list_kill);
299 unregister_netdevice_many(&list_kill);
302 void rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
305 __rtnl_kill_links(net, ops);
308 EXPORT_SYMBOL_GPL(rtnl_kill_links);
311 * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
312 * @ops: struct rtnl_link_ops * to unregister
314 * The caller must hold the rtnl_mutex.
316 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
321 __rtnl_kill_links(net, ops);
323 list_del(&ops->list);
325 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
328 * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
329 * @ops: struct rtnl_link_ops * to unregister
331 void rtnl_link_unregister(struct rtnl_link_ops *ops)
334 __rtnl_link_unregister(ops);
337 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
339 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
341 const struct rtnl_link_ops *ops;
343 list_for_each_entry(ops, &link_ops, list) {
344 if (!strcmp(ops->kind, kind))
350 static size_t rtnl_link_get_size(const struct net_device *dev)
352 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
358 size = nlmsg_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
359 nlmsg_total_size(strlen(ops->kind) + 1); /* IFLA_INFO_KIND */
362 /* IFLA_INFO_DATA + nested data */
363 size += nlmsg_total_size(sizeof(struct nlattr)) +
366 if (ops->get_xstats_size)
367 size += ops->get_xstats_size(dev); /* IFLA_INFO_XSTATS */
372 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
374 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
375 struct nlattr *linkinfo, *data;
378 linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
379 if (linkinfo == NULL)
382 if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
383 goto err_cancel_link;
384 if (ops->fill_xstats) {
385 err = ops->fill_xstats(skb, dev);
387 goto err_cancel_link;
389 if (ops->fill_info) {
390 data = nla_nest_start(skb, IFLA_INFO_DATA);
392 goto err_cancel_link;
393 err = ops->fill_info(skb, dev);
395 goto err_cancel_data;
396 nla_nest_end(skb, data);
399 nla_nest_end(skb, linkinfo);
403 nla_nest_cancel(skb, data);
405 nla_nest_cancel(skb, linkinfo);
410 static const int rtm_min[RTM_NR_FAMILIES] =
412 [RTM_FAM(RTM_NEWLINK)] = NLMSG_LENGTH(sizeof(struct ifinfomsg)),
413 [RTM_FAM(RTM_NEWADDR)] = NLMSG_LENGTH(sizeof(struct ifaddrmsg)),
414 [RTM_FAM(RTM_NEWROUTE)] = NLMSG_LENGTH(sizeof(struct rtmsg)),
415 [RTM_FAM(RTM_NEWRULE)] = NLMSG_LENGTH(sizeof(struct fib_rule_hdr)),
416 [RTM_FAM(RTM_NEWQDISC)] = NLMSG_LENGTH(sizeof(struct tcmsg)),
417 [RTM_FAM(RTM_NEWTCLASS)] = NLMSG_LENGTH(sizeof(struct tcmsg)),
418 [RTM_FAM(RTM_NEWTFILTER)] = NLMSG_LENGTH(sizeof(struct tcmsg)),
419 [RTM_FAM(RTM_NEWACTION)] = NLMSG_LENGTH(sizeof(struct tcamsg)),
420 [RTM_FAM(RTM_GETMULTICAST)] = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
421 [RTM_FAM(RTM_GETANYCAST)] = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
424 static const int rta_max[RTM_NR_FAMILIES] =
426 [RTM_FAM(RTM_NEWLINK)] = IFLA_MAX,
427 [RTM_FAM(RTM_NEWADDR)] = IFA_MAX,
428 [RTM_FAM(RTM_NEWROUTE)] = RTA_MAX,
429 [RTM_FAM(RTM_NEWRULE)] = FRA_MAX,
430 [RTM_FAM(RTM_NEWQDISC)] = TCA_MAX,
431 [RTM_FAM(RTM_NEWTCLASS)] = TCA_MAX,
432 [RTM_FAM(RTM_NEWTFILTER)] = TCA_MAX,
433 [RTM_FAM(RTM_NEWACTION)] = TCAA_MAX,
436 void __rta_fill(struct sk_buff *skb, int attrtype, int attrlen, const void *data)
439 int size = RTA_LENGTH(attrlen);
441 rta = (struct rtattr *)skb_put(skb, RTA_ALIGN(size));
442 rta->rta_type = attrtype;
444 memcpy(RTA_DATA(rta), data, attrlen);
445 memset(RTA_DATA(rta) + attrlen, 0, RTA_ALIGN(size) - size);
447 EXPORT_SYMBOL(__rta_fill);
449 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned group, int echo)
451 struct sock *rtnl = net->rtnl;
454 NETLINK_CB(skb).dst_group = group;
456 atomic_inc(&skb->users);
457 netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
459 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
463 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
465 struct sock *rtnl = net->rtnl;
467 return nlmsg_unicast(rtnl, skb, pid);
469 EXPORT_SYMBOL(rtnl_unicast);
471 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
472 struct nlmsghdr *nlh, gfp_t flags)
474 struct sock *rtnl = net->rtnl;
478 report = nlmsg_report(nlh);
480 nlmsg_notify(rtnl, skb, pid, group, report, flags);
482 EXPORT_SYMBOL(rtnl_notify);
484 void rtnl_set_sk_err(struct net *net, u32 group, int error)
486 struct sock *rtnl = net->rtnl;
488 netlink_set_err(rtnl, 0, group, error);
490 EXPORT_SYMBOL(rtnl_set_sk_err);
492 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
497 mx = nla_nest_start(skb, RTA_METRICS);
501 for (i = 0; i < RTAX_MAX; i++) {
504 NLA_PUT_U32(skb, i+1, metrics[i]);
509 nla_nest_cancel(skb, mx);
513 return nla_nest_end(skb, mx);
516 nla_nest_cancel(skb, mx);
519 EXPORT_SYMBOL(rtnetlink_put_metrics);
521 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
522 u32 ts, u32 tsage, long expires, u32 error)
524 struct rta_cacheinfo ci = {
525 .rta_lastuse = jiffies_to_clock_t(jiffies - dst->lastuse),
526 .rta_used = dst->__use,
527 .rta_clntref = atomic_read(&(dst->__refcnt)),
535 ci.rta_expires = jiffies_to_clock_t(expires);
537 return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
539 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
541 static void set_operstate(struct net_device *dev, unsigned char transition)
543 unsigned char operstate = dev->operstate;
545 switch (transition) {
547 if ((operstate == IF_OPER_DORMANT ||
548 operstate == IF_OPER_UNKNOWN) &&
550 operstate = IF_OPER_UP;
553 case IF_OPER_DORMANT:
554 if (operstate == IF_OPER_UP ||
555 operstate == IF_OPER_UNKNOWN)
556 operstate = IF_OPER_DORMANT;
560 if (dev->operstate != operstate) {
561 write_lock_bh(&dev_base_lock);
562 dev->operstate = operstate;
563 write_unlock_bh(&dev_base_lock);
564 netdev_state_change(dev);
568 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
569 const struct ifinfomsg *ifm)
571 unsigned int flags = ifm->ifi_flags;
573 /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
575 flags = (flags & ifm->ifi_change) |
576 (dev->flags & ~ifm->ifi_change);
581 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
582 const struct net_device_stats *b)
584 a->rx_packets = b->rx_packets;
585 a->tx_packets = b->tx_packets;
586 a->rx_bytes = b->rx_bytes;
587 a->tx_bytes = b->tx_bytes;
588 a->rx_errors = b->rx_errors;
589 a->tx_errors = b->tx_errors;
590 a->rx_dropped = b->rx_dropped;
591 a->tx_dropped = b->tx_dropped;
593 a->multicast = b->multicast;
594 a->collisions = b->collisions;
596 a->rx_length_errors = b->rx_length_errors;
597 a->rx_over_errors = b->rx_over_errors;
598 a->rx_crc_errors = b->rx_crc_errors;
599 a->rx_frame_errors = b->rx_frame_errors;
600 a->rx_fifo_errors = b->rx_fifo_errors;
601 a->rx_missed_errors = b->rx_missed_errors;
603 a->tx_aborted_errors = b->tx_aborted_errors;
604 a->tx_carrier_errors = b->tx_carrier_errors;
605 a->tx_fifo_errors = b->tx_fifo_errors;
606 a->tx_heartbeat_errors = b->tx_heartbeat_errors;
607 a->tx_window_errors = b->tx_window_errors;
609 a->rx_compressed = b->rx_compressed;
610 a->tx_compressed = b->tx_compressed;
613 static void copy_rtnl_link_stats64(void *v, const struct net_device_stats *b)
615 struct rtnl_link_stats64 a;
617 a.rx_packets = b->rx_packets;
618 a.tx_packets = b->tx_packets;
619 a.rx_bytes = b->rx_bytes;
620 a.tx_bytes = b->tx_bytes;
621 a.rx_errors = b->rx_errors;
622 a.tx_errors = b->tx_errors;
623 a.rx_dropped = b->rx_dropped;
624 a.tx_dropped = b->tx_dropped;
626 a.multicast = b->multicast;
627 a.collisions = b->collisions;
629 a.rx_length_errors = b->rx_length_errors;
630 a.rx_over_errors = b->rx_over_errors;
631 a.rx_crc_errors = b->rx_crc_errors;
632 a.rx_frame_errors = b->rx_frame_errors;
633 a.rx_fifo_errors = b->rx_fifo_errors;
634 a.rx_missed_errors = b->rx_missed_errors;
636 a.tx_aborted_errors = b->tx_aborted_errors;
637 a.tx_carrier_errors = b->tx_carrier_errors;
638 a.tx_fifo_errors = b->tx_fifo_errors;
639 a.tx_heartbeat_errors = b->tx_heartbeat_errors;
640 a.tx_window_errors = b->tx_window_errors;
642 a.rx_compressed = b->rx_compressed;
643 a.tx_compressed = b->tx_compressed;
644 memcpy(v, &a, sizeof(a));
647 static inline int rtnl_vfinfo_size(const struct net_device *dev)
649 if (dev->dev.parent && dev_is_pci(dev->dev.parent))
650 return dev_num_vf(dev->dev.parent) *
651 sizeof(struct ifla_vf_info);
656 static inline size_t if_nlmsg_size(const struct net_device *dev)
658 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
659 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
660 + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
661 + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
662 + nla_total_size(sizeof(struct rtnl_link_ifmap))
663 + nla_total_size(sizeof(struct rtnl_link_stats))
664 + nla_total_size(sizeof(struct rtnl_link_stats64))
665 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
666 + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
667 + nla_total_size(4) /* IFLA_TXQLEN */
668 + nla_total_size(4) /* IFLA_WEIGHT */
669 + nla_total_size(4) /* IFLA_MTU */
670 + nla_total_size(4) /* IFLA_LINK */
671 + nla_total_size(4) /* IFLA_MASTER */
672 + nla_total_size(1) /* IFLA_OPERSTATE */
673 + nla_total_size(1) /* IFLA_LINKMODE */
674 + nla_total_size(4) /* IFLA_NUM_VF */
675 + nla_total_size(rtnl_vfinfo_size(dev)) /* IFLA_VFINFO */
676 + rtnl_link_get_size(dev); /* IFLA_LINKINFO */
679 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
680 int type, u32 pid, u32 seq, u32 change,
683 struct ifinfomsg *ifm;
684 struct nlmsghdr *nlh;
685 const struct net_device_stats *stats;
688 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
692 ifm = nlmsg_data(nlh);
693 ifm->ifi_family = AF_UNSPEC;
695 ifm->ifi_type = dev->type;
696 ifm->ifi_index = dev->ifindex;
697 ifm->ifi_flags = dev_get_flags(dev);
698 ifm->ifi_change = change;
700 NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
701 NLA_PUT_U32(skb, IFLA_TXQLEN, dev->tx_queue_len);
702 NLA_PUT_U8(skb, IFLA_OPERSTATE,
703 netif_running(dev) ? dev->operstate : IF_OPER_DOWN);
704 NLA_PUT_U8(skb, IFLA_LINKMODE, dev->link_mode);
705 NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
707 if (dev->ifindex != dev->iflink)
708 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
711 NLA_PUT_U32(skb, IFLA_MASTER, dev->master->ifindex);
714 NLA_PUT_STRING(skb, IFLA_QDISC, dev->qdisc->ops->id);
717 NLA_PUT_STRING(skb, IFLA_IFALIAS, dev->ifalias);
720 struct rtnl_link_ifmap map = {
721 .mem_start = dev->mem_start,
722 .mem_end = dev->mem_end,
723 .base_addr = dev->base_addr,
726 .port = dev->if_port,
728 NLA_PUT(skb, IFLA_MAP, sizeof(map), &map);
732 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
733 NLA_PUT(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast);
736 attr = nla_reserve(skb, IFLA_STATS,
737 sizeof(struct rtnl_link_stats));
739 goto nla_put_failure;
741 stats = dev_get_stats(dev);
742 copy_rtnl_link_stats(nla_data(attr), stats);
744 attr = nla_reserve(skb, IFLA_STATS64,
745 sizeof(struct rtnl_link_stats64));
747 goto nla_put_failure;
748 copy_rtnl_link_stats64(nla_data(attr), stats);
750 if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent) {
752 struct ifla_vf_info ivi;
754 NLA_PUT_U32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent));
755 for (i = 0; i < dev_num_vf(dev->dev.parent); i++) {
756 if (dev->netdev_ops->ndo_get_vf_config(dev, i, &ivi))
758 NLA_PUT(skb, IFLA_VFINFO, sizeof(ivi), &ivi);
761 if (dev->rtnl_link_ops) {
762 if (rtnl_link_fill(skb, dev) < 0)
763 goto nla_put_failure;
766 return nlmsg_end(skb, nlh);
769 nlmsg_cancel(skb, nlh);
773 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
775 struct net *net = sock_net(skb->sk);
778 struct net_device *dev;
779 struct hlist_head *head;
780 struct hlist_node *node;
785 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
787 head = &net->dev_index_head[h];
788 hlist_for_each_entry(dev, node, head, index_hlist) {
791 if (rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
792 NETLINK_CB(cb->skb).pid,
793 cb->nlh->nlmsg_seq, 0,
807 const struct nla_policy ifla_policy[IFLA_MAX+1] = {
808 [IFLA_IFNAME] = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
809 [IFLA_ADDRESS] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
810 [IFLA_BROADCAST] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
811 [IFLA_MAP] = { .len = sizeof(struct rtnl_link_ifmap) },
812 [IFLA_MTU] = { .type = NLA_U32 },
813 [IFLA_LINK] = { .type = NLA_U32 },
814 [IFLA_TXQLEN] = { .type = NLA_U32 },
815 [IFLA_WEIGHT] = { .type = NLA_U32 },
816 [IFLA_OPERSTATE] = { .type = NLA_U8 },
817 [IFLA_LINKMODE] = { .type = NLA_U8 },
818 [IFLA_LINKINFO] = { .type = NLA_NESTED },
819 [IFLA_NET_NS_PID] = { .type = NLA_U32 },
820 [IFLA_IFALIAS] = { .type = NLA_STRING, .len = IFALIASZ-1 },
821 [IFLA_VF_MAC] = { .type = NLA_BINARY,
822 .len = sizeof(struct ifla_vf_mac) },
823 [IFLA_VF_VLAN] = { .type = NLA_BINARY,
824 .len = sizeof(struct ifla_vf_vlan) },
825 [IFLA_VF_TX_RATE] = { .type = NLA_BINARY,
826 .len = sizeof(struct ifla_vf_tx_rate) },
828 EXPORT_SYMBOL(ifla_policy);
830 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
831 [IFLA_INFO_KIND] = { .type = NLA_STRING },
832 [IFLA_INFO_DATA] = { .type = NLA_NESTED },
835 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
838 /* Examine the link attributes and figure out which
839 * network namespace we are talking about.
841 if (tb[IFLA_NET_NS_PID])
842 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
844 net = get_net(src_net);
847 EXPORT_SYMBOL(rtnl_link_get_net);
849 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
852 if (tb[IFLA_ADDRESS] &&
853 nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
856 if (tb[IFLA_BROADCAST] &&
857 nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
864 static int do_setlink(struct net_device *dev, struct ifinfomsg *ifm,
865 struct nlattr **tb, char *ifname, int modified)
867 const struct net_device_ops *ops = dev->netdev_ops;
868 int send_addr_notify = 0;
871 if (tb[IFLA_NET_NS_PID]) {
872 struct net *net = rtnl_link_get_net(dev_net(dev), tb);
877 err = dev_change_net_namespace(dev, net, ifname);
885 struct rtnl_link_ifmap *u_map;
888 if (!ops->ndo_set_config) {
893 if (!netif_device_present(dev)) {
898 u_map = nla_data(tb[IFLA_MAP]);
899 k_map.mem_start = (unsigned long) u_map->mem_start;
900 k_map.mem_end = (unsigned long) u_map->mem_end;
901 k_map.base_addr = (unsigned short) u_map->base_addr;
902 k_map.irq = (unsigned char) u_map->irq;
903 k_map.dma = (unsigned char) u_map->dma;
904 k_map.port = (unsigned char) u_map->port;
906 err = ops->ndo_set_config(dev, &k_map);
913 if (tb[IFLA_ADDRESS]) {
917 if (!ops->ndo_set_mac_address) {
922 if (!netif_device_present(dev)) {
927 len = sizeof(sa_family_t) + dev->addr_len;
928 sa = kmalloc(len, GFP_KERNEL);
933 sa->sa_family = dev->type;
934 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
936 err = ops->ndo_set_mac_address(dev, sa);
940 send_addr_notify = 1;
945 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
952 * Interface selected by interface index but interface
953 * name provided implies that a name change has been
956 if (ifm->ifi_index > 0 && ifname[0]) {
957 err = dev_change_name(dev, ifname);
963 if (tb[IFLA_IFALIAS]) {
964 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
965 nla_len(tb[IFLA_IFALIAS]));
971 if (tb[IFLA_BROADCAST]) {
972 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
973 send_addr_notify = 1;
976 if (ifm->ifi_flags || ifm->ifi_change) {
977 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
983 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
985 if (tb[IFLA_OPERSTATE])
986 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
988 if (tb[IFLA_LINKMODE]) {
989 write_lock_bh(&dev_base_lock);
990 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
991 write_unlock_bh(&dev_base_lock);
994 if (tb[IFLA_VF_MAC]) {
995 struct ifla_vf_mac *ivm;
996 ivm = nla_data(tb[IFLA_VF_MAC]);
998 if (ops->ndo_set_vf_mac)
999 err = ops->ndo_set_vf_mac(dev, ivm->vf, ivm->mac);
1005 if (tb[IFLA_VF_VLAN]) {
1006 struct ifla_vf_vlan *ivv;
1007 ivv = nla_data(tb[IFLA_VF_VLAN]);
1009 if (ops->ndo_set_vf_vlan)
1010 err = ops->ndo_set_vf_vlan(dev, ivv->vf,
1019 if (tb[IFLA_VF_TX_RATE]) {
1020 struct ifla_vf_tx_rate *ivt;
1021 ivt = nla_data(tb[IFLA_VF_TX_RATE]);
1023 if (ops->ndo_set_vf_tx_rate)
1024 err = ops->ndo_set_vf_tx_rate(dev, ivt->vf, ivt->rate);
1032 if (err < 0 && modified && net_ratelimit())
1033 printk(KERN_WARNING "A link change request failed with "
1034 "some changes comitted already. Interface %s may "
1035 "have been left with an inconsistent configuration, "
1036 "please check.\n", dev->name);
1038 if (send_addr_notify)
1039 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
1043 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1045 struct net *net = sock_net(skb->sk);
1046 struct ifinfomsg *ifm;
1047 struct net_device *dev;
1049 struct nlattr *tb[IFLA_MAX+1];
1050 char ifname[IFNAMSIZ];
1052 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1056 if (tb[IFLA_IFNAME])
1057 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1062 ifm = nlmsg_data(nlh);
1063 if (ifm->ifi_index > 0)
1064 dev = __dev_get_by_index(net, ifm->ifi_index);
1065 else if (tb[IFLA_IFNAME])
1066 dev = __dev_get_by_name(net, ifname);
1075 err = validate_linkmsg(dev, tb);
1079 err = do_setlink(dev, ifm, tb, ifname, 0);
1084 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1086 struct net *net = sock_net(skb->sk);
1087 const struct rtnl_link_ops *ops;
1088 struct net_device *dev;
1089 struct ifinfomsg *ifm;
1090 char ifname[IFNAMSIZ];
1091 struct nlattr *tb[IFLA_MAX+1];
1094 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1098 if (tb[IFLA_IFNAME])
1099 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1101 ifm = nlmsg_data(nlh);
1102 if (ifm->ifi_index > 0)
1103 dev = __dev_get_by_index(net, ifm->ifi_index);
1104 else if (tb[IFLA_IFNAME])
1105 dev = __dev_get_by_name(net, ifname);
1112 ops = dev->rtnl_link_ops;
1116 ops->dellink(dev, NULL);
1120 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
1122 unsigned int old_flags;
1125 old_flags = dev->flags;
1126 if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
1127 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1132 dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
1133 rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U);
1135 __dev_notify_flags(dev, old_flags);
1138 EXPORT_SYMBOL(rtnl_configure_link);
1140 struct net_device *rtnl_create_link(struct net *src_net, struct net *net,
1141 char *ifname, const struct rtnl_link_ops *ops, struct nlattr *tb[])
1144 struct net_device *dev;
1145 unsigned int num_queues = 1;
1146 unsigned int real_num_queues = 1;
1148 if (ops->get_tx_queues) {
1149 err = ops->get_tx_queues(src_net, tb, &num_queues,
1155 dev = alloc_netdev_mq(ops->priv_size, ifname, ops->setup, num_queues);
1159 dev_net_set(dev, net);
1160 dev->rtnl_link_ops = ops;
1161 dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
1162 dev->real_num_tx_queues = real_num_queues;
1164 if (strchr(dev->name, '%')) {
1165 err = dev_alloc_name(dev, dev->name);
1171 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
1172 if (tb[IFLA_ADDRESS])
1173 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
1174 nla_len(tb[IFLA_ADDRESS]));
1175 if (tb[IFLA_BROADCAST])
1176 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
1177 nla_len(tb[IFLA_BROADCAST]));
1178 if (tb[IFLA_TXQLEN])
1179 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1180 if (tb[IFLA_OPERSTATE])
1181 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1182 if (tb[IFLA_LINKMODE])
1183 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1190 return ERR_PTR(err);
1192 EXPORT_SYMBOL(rtnl_create_link);
1194 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1196 struct net *net = sock_net(skb->sk);
1197 const struct rtnl_link_ops *ops;
1198 struct net_device *dev;
1199 struct ifinfomsg *ifm;
1200 char kind[MODULE_NAME_LEN];
1201 char ifname[IFNAMSIZ];
1202 struct nlattr *tb[IFLA_MAX+1];
1203 struct nlattr *linkinfo[IFLA_INFO_MAX+1];
1206 #ifdef CONFIG_MODULES
1209 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1213 if (tb[IFLA_IFNAME])
1214 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1218 ifm = nlmsg_data(nlh);
1219 if (ifm->ifi_index > 0)
1220 dev = __dev_get_by_index(net, ifm->ifi_index);
1222 dev = __dev_get_by_name(net, ifname);
1226 err = validate_linkmsg(dev, tb);
1230 if (tb[IFLA_LINKINFO]) {
1231 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
1232 tb[IFLA_LINKINFO], ifla_info_policy);
1236 memset(linkinfo, 0, sizeof(linkinfo));
1238 if (linkinfo[IFLA_INFO_KIND]) {
1239 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
1240 ops = rtnl_link_ops_get(kind);
1247 struct nlattr *attr[ops ? ops->maxtype + 1 : 0], **data = NULL;
1248 struct net *dest_net;
1251 if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
1252 err = nla_parse_nested(attr, ops->maxtype,
1253 linkinfo[IFLA_INFO_DATA],
1259 if (ops->validate) {
1260 err = ops->validate(tb, data);
1269 if (nlh->nlmsg_flags & NLM_F_EXCL)
1271 if (nlh->nlmsg_flags & NLM_F_REPLACE)
1274 if (linkinfo[IFLA_INFO_DATA]) {
1275 if (!ops || ops != dev->rtnl_link_ops ||
1279 err = ops->changelink(dev, tb, data);
1285 return do_setlink(dev, ifm, tb, ifname, modified);
1288 if (!(nlh->nlmsg_flags & NLM_F_CREATE))
1293 if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
1297 #ifdef CONFIG_MODULES
1300 request_module("rtnl-link-%s", kind);
1302 ops = rtnl_link_ops_get(kind);
1311 snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
1313 dest_net = rtnl_link_get_net(net, tb);
1314 dev = rtnl_create_link(net, dest_net, ifname, ops, tb);
1318 else if (ops->newlink)
1319 err = ops->newlink(net, dev, tb, data);
1321 err = register_netdevice(dev);
1323 if (err < 0 && !IS_ERR(dev))
1328 err = rtnl_configure_link(dev, ifm);
1330 unregister_netdevice(dev);
1337 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
1339 struct net *net = sock_net(skb->sk);
1340 struct ifinfomsg *ifm;
1341 char ifname[IFNAMSIZ];
1342 struct nlattr *tb[IFLA_MAX+1];
1343 struct net_device *dev = NULL;
1344 struct sk_buff *nskb;
1347 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1351 if (tb[IFLA_IFNAME])
1352 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1354 ifm = nlmsg_data(nlh);
1355 if (ifm->ifi_index > 0)
1356 dev = __dev_get_by_index(net, ifm->ifi_index);
1357 else if (tb[IFLA_IFNAME])
1358 dev = __dev_get_by_name(net, ifname);
1365 nskb = nlmsg_new(if_nlmsg_size(dev), GFP_KERNEL);
1369 err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).pid,
1370 nlh->nlmsg_seq, 0, 0);
1372 /* -EMSGSIZE implies BUG in if_nlmsg_size */
1373 WARN_ON(err == -EMSGSIZE);
1376 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).pid);
1381 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
1384 int s_idx = cb->family;
1388 for (idx = 1; idx < NPROTO; idx++) {
1389 int type = cb->nlh->nlmsg_type-RTM_BASE;
1390 if (idx < s_idx || idx == PF_PACKET)
1392 if (rtnl_msg_handlers[idx] == NULL ||
1393 rtnl_msg_handlers[idx][type].dumpit == NULL)
1396 memset(&cb->args[0], 0, sizeof(cb->args));
1397 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
1405 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned change)
1407 struct net *net = dev_net(dev);
1408 struct sk_buff *skb;
1411 skb = nlmsg_new(if_nlmsg_size(dev), GFP_KERNEL);
1415 err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0);
1417 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
1418 WARN_ON(err == -EMSGSIZE);
1422 rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_KERNEL);
1426 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
1429 /* Protected by RTNL sempahore. */
1430 static struct rtattr **rta_buf;
1431 static int rtattr_max;
1433 /* Process one rtnetlink message. */
1435 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
1437 struct net *net = sock_net(skb->sk);
1438 rtnl_doit_func doit;
1445 type = nlh->nlmsg_type;
1451 /* All the messages must have at least 1 byte length */
1452 if (nlh->nlmsg_len < NLMSG_LENGTH(sizeof(struct rtgenmsg)))
1455 family = ((struct rtgenmsg *)NLMSG_DATA(nlh))->rtgen_family;
1459 if (kind != 2 && security_netlink_recv(skb, CAP_NET_ADMIN))
1462 if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
1464 rtnl_dumpit_func dumpit;
1466 dumpit = rtnl_get_dumpit(family, type);
1472 err = netlink_dump_start(rtnl, skb, nlh, dumpit, NULL);
1477 memset(rta_buf, 0, (rtattr_max * sizeof(struct rtattr *)));
1479 min_len = rtm_min[sz_idx];
1480 if (nlh->nlmsg_len < min_len)
1483 if (nlh->nlmsg_len > min_len) {
1484 int attrlen = nlh->nlmsg_len - NLMSG_ALIGN(min_len);
1485 struct rtattr *attr = (void *)nlh + NLMSG_ALIGN(min_len);
1487 while (RTA_OK(attr, attrlen)) {
1488 unsigned flavor = attr->rta_type;
1490 if (flavor > rta_max[sz_idx])
1492 rta_buf[flavor-1] = attr;
1494 attr = RTA_NEXT(attr, attrlen);
1498 doit = rtnl_get_doit(family, type);
1502 return doit(skb, nlh, (void *)&rta_buf[0]);
1505 static void rtnetlink_rcv(struct sk_buff *skb)
1508 netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
1512 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
1514 struct net_device *dev = ptr;
1520 case NETDEV_POST_INIT:
1521 case NETDEV_REGISTER:
1523 case NETDEV_PRE_TYPE_CHANGE:
1524 case NETDEV_GOING_DOWN:
1525 case NETDEV_UNREGISTER:
1526 case NETDEV_UNREGISTER_BATCH:
1529 rtmsg_ifinfo(RTM_NEWLINK, dev, 0);
1535 static struct notifier_block rtnetlink_dev_notifier = {
1536 .notifier_call = rtnetlink_event,
1540 static int __net_init rtnetlink_net_init(struct net *net)
1543 sk = netlink_kernel_create(net, NETLINK_ROUTE, RTNLGRP_MAX,
1544 rtnetlink_rcv, &rtnl_mutex, THIS_MODULE);
1551 static void __net_exit rtnetlink_net_exit(struct net *net)
1553 netlink_kernel_release(net->rtnl);
1557 static struct pernet_operations rtnetlink_net_ops = {
1558 .init = rtnetlink_net_init,
1559 .exit = rtnetlink_net_exit,
1562 void __init rtnetlink_init(void)
1567 for (i = 0; i < ARRAY_SIZE(rta_max); i++)
1568 if (rta_max[i] > rtattr_max)
1569 rtattr_max = rta_max[i];
1570 rta_buf = kmalloc(rtattr_max * sizeof(struct rtattr *), GFP_KERNEL);
1572 panic("rtnetlink_init: cannot allocate rta_buf\n");
1574 if (register_pernet_subsys(&rtnetlink_net_ops))
1575 panic("rtnetlink_init: cannot initialize rtnetlink\n");
1577 netlink_set_nonroot(NETLINK_ROUTE, NL_NONROOT_RECV);
1578 register_netdevice_notifier(&rtnetlink_dev_notifier);
1580 rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink, rtnl_dump_ifinfo);
1581 rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL);
1582 rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL);
1583 rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL);
1585 rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all);
1586 rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all);