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/nsproxy.h>
40 #include <asm/uaccess.h>
41 #include <asm/system.h>
42 #include <asm/string.h>
44 #include <linux/inet.h>
45 #include <linux/netdevice.h>
47 #include <net/protocol.h>
49 #include <net/route.h>
52 #include <net/pkt_sched.h>
53 #include <net/fib_rules.h>
54 #include <net/rtnetlink.h>
59 rtnl_dumpit_func dumpit;
62 static DEFINE_MUTEX(rtnl_mutex);
66 mutex_lock(&rtnl_mutex);
69 void __rtnl_unlock(void)
71 mutex_unlock(&rtnl_mutex);
74 void rtnl_unlock(void)
76 /* This fellow will unlock it for us. */
80 int rtnl_trylock(void)
82 return mutex_trylock(&rtnl_mutex);
85 int rtnl_is_locked(void)
87 return mutex_is_locked(&rtnl_mutex);
90 static struct rtnl_link *rtnl_msg_handlers[NPROTO];
92 static inline int rtm_msgindex(int msgtype)
94 int msgindex = msgtype - RTM_BASE;
97 * msgindex < 0 implies someone tried to register a netlink
98 * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
99 * the message type has not been added to linux/rtnetlink.h
101 BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
106 static rtnl_doit_func rtnl_get_doit(int protocol, int msgindex)
108 struct rtnl_link *tab;
110 tab = rtnl_msg_handlers[protocol];
111 if (tab == NULL || tab[msgindex].doit == NULL)
112 tab = rtnl_msg_handlers[PF_UNSPEC];
114 return tab ? tab[msgindex].doit : NULL;
117 static rtnl_dumpit_func rtnl_get_dumpit(int protocol, int msgindex)
119 struct rtnl_link *tab;
121 tab = rtnl_msg_handlers[protocol];
122 if (tab == NULL || tab[msgindex].dumpit == NULL)
123 tab = rtnl_msg_handlers[PF_UNSPEC];
125 return tab ? tab[msgindex].dumpit : NULL;
129 * __rtnl_register - Register a rtnetlink message type
130 * @protocol: Protocol family or PF_UNSPEC
131 * @msgtype: rtnetlink message type
132 * @doit: Function pointer called for each request message
133 * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
135 * Registers the specified function pointers (at least one of them has
136 * to be non-NULL) to be called whenever a request message for the
137 * specified protocol family and message type is received.
139 * The special protocol family PF_UNSPEC may be used to define fallback
140 * function pointers for the case when no entry for the specific protocol
143 * Returns 0 on success or a negative error code.
145 int __rtnl_register(int protocol, int msgtype,
146 rtnl_doit_func doit, rtnl_dumpit_func dumpit)
148 struct rtnl_link *tab;
151 BUG_ON(protocol < 0 || protocol >= NPROTO);
152 msgindex = rtm_msgindex(msgtype);
154 tab = rtnl_msg_handlers[protocol];
156 tab = kcalloc(RTM_NR_MSGTYPES, sizeof(*tab), GFP_KERNEL);
160 rtnl_msg_handlers[protocol] = tab;
164 tab[msgindex].doit = doit;
167 tab[msgindex].dumpit = dumpit;
172 EXPORT_SYMBOL_GPL(__rtnl_register);
175 * rtnl_register - Register a rtnetlink message type
177 * Identical to __rtnl_register() but panics on failure. This is useful
178 * as failure of this function is very unlikely, it can only happen due
179 * to lack of memory when allocating the chain to store all message
180 * handlers for a protocol. Meant for use in init functions where lack
181 * of memory implies no sense in continueing.
183 void rtnl_register(int protocol, int msgtype,
184 rtnl_doit_func doit, rtnl_dumpit_func dumpit)
186 if (__rtnl_register(protocol, msgtype, doit, dumpit) < 0)
187 panic("Unable to register rtnetlink message handler, "
188 "protocol = %d, message type = %d\n",
192 EXPORT_SYMBOL_GPL(rtnl_register);
195 * rtnl_unregister - Unregister a rtnetlink message type
196 * @protocol: Protocol family or PF_UNSPEC
197 * @msgtype: rtnetlink message type
199 * Returns 0 on success or a negative error code.
201 int rtnl_unregister(int protocol, int msgtype)
205 BUG_ON(protocol < 0 || protocol >= NPROTO);
206 msgindex = rtm_msgindex(msgtype);
208 if (rtnl_msg_handlers[protocol] == NULL)
211 rtnl_msg_handlers[protocol][msgindex].doit = NULL;
212 rtnl_msg_handlers[protocol][msgindex].dumpit = NULL;
217 EXPORT_SYMBOL_GPL(rtnl_unregister);
220 * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
221 * @protocol : Protocol family or PF_UNSPEC
223 * Identical to calling rtnl_unregster() for all registered message types
224 * of a certain protocol family.
226 void rtnl_unregister_all(int protocol)
228 BUG_ON(protocol < 0 || protocol >= NPROTO);
230 kfree(rtnl_msg_handlers[protocol]);
231 rtnl_msg_handlers[protocol] = NULL;
234 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
236 static LIST_HEAD(link_ops);
239 * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
240 * @ops: struct rtnl_link_ops * to register
242 * The caller must hold the rtnl_mutex. This function should be used
243 * by drivers that create devices during module initialization. It
244 * must be called before registering the devices.
246 * Returns 0 on success or a negative error code.
248 int __rtnl_link_register(struct rtnl_link_ops *ops)
251 ops->dellink = unregister_netdevice;
253 list_add_tail(&ops->list, &link_ops);
257 EXPORT_SYMBOL_GPL(__rtnl_link_register);
260 * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
261 * @ops: struct rtnl_link_ops * to register
263 * Returns 0 on success or a negative error code.
265 int rtnl_link_register(struct rtnl_link_ops *ops)
270 err = __rtnl_link_register(ops);
275 EXPORT_SYMBOL_GPL(rtnl_link_register);
277 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
279 struct net_device *dev;
281 for_each_netdev(net, dev) {
282 if (dev->rtnl_link_ops == ops) {
289 void rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
292 __rtnl_kill_links(net, ops);
295 EXPORT_SYMBOL_GPL(rtnl_kill_links);
298 * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
299 * @ops: struct rtnl_link_ops * to unregister
301 * The caller must hold the rtnl_mutex.
303 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
308 __rtnl_kill_links(net, ops);
310 list_del(&ops->list);
313 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
316 * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
317 * @ops: struct rtnl_link_ops * to unregister
319 void rtnl_link_unregister(struct rtnl_link_ops *ops)
322 __rtnl_link_unregister(ops);
326 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
328 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
330 const struct rtnl_link_ops *ops;
332 list_for_each_entry(ops, &link_ops, list) {
333 if (!strcmp(ops->kind, kind))
339 static size_t rtnl_link_get_size(const struct net_device *dev)
341 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
347 size = nlmsg_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
348 nlmsg_total_size(strlen(ops->kind) + 1); /* IFLA_INFO_KIND */
351 /* IFLA_INFO_DATA + nested data */
352 size += nlmsg_total_size(sizeof(struct nlattr)) +
355 if (ops->get_xstats_size)
356 size += ops->get_xstats_size(dev); /* IFLA_INFO_XSTATS */
361 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
363 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
364 struct nlattr *linkinfo, *data;
367 linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
368 if (linkinfo == NULL)
371 if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
372 goto err_cancel_link;
373 if (ops->fill_xstats) {
374 err = ops->fill_xstats(skb, dev);
376 goto err_cancel_link;
378 if (ops->fill_info) {
379 data = nla_nest_start(skb, IFLA_INFO_DATA);
381 goto err_cancel_link;
382 err = ops->fill_info(skb, dev);
384 goto err_cancel_data;
385 nla_nest_end(skb, data);
388 nla_nest_end(skb, linkinfo);
392 nla_nest_cancel(skb, data);
394 nla_nest_cancel(skb, linkinfo);
399 static const int rtm_min[RTM_NR_FAMILIES] =
401 [RTM_FAM(RTM_NEWLINK)] = NLMSG_LENGTH(sizeof(struct ifinfomsg)),
402 [RTM_FAM(RTM_NEWADDR)] = NLMSG_LENGTH(sizeof(struct ifaddrmsg)),
403 [RTM_FAM(RTM_NEWROUTE)] = NLMSG_LENGTH(sizeof(struct rtmsg)),
404 [RTM_FAM(RTM_NEWRULE)] = NLMSG_LENGTH(sizeof(struct fib_rule_hdr)),
405 [RTM_FAM(RTM_NEWQDISC)] = NLMSG_LENGTH(sizeof(struct tcmsg)),
406 [RTM_FAM(RTM_NEWTCLASS)] = NLMSG_LENGTH(sizeof(struct tcmsg)),
407 [RTM_FAM(RTM_NEWTFILTER)] = NLMSG_LENGTH(sizeof(struct tcmsg)),
408 [RTM_FAM(RTM_NEWACTION)] = NLMSG_LENGTH(sizeof(struct tcamsg)),
409 [RTM_FAM(RTM_GETMULTICAST)] = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
410 [RTM_FAM(RTM_GETANYCAST)] = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
413 static const int rta_max[RTM_NR_FAMILIES] =
415 [RTM_FAM(RTM_NEWLINK)] = IFLA_MAX,
416 [RTM_FAM(RTM_NEWADDR)] = IFA_MAX,
417 [RTM_FAM(RTM_NEWROUTE)] = RTA_MAX,
418 [RTM_FAM(RTM_NEWRULE)] = FRA_MAX,
419 [RTM_FAM(RTM_NEWQDISC)] = TCA_MAX,
420 [RTM_FAM(RTM_NEWTCLASS)] = TCA_MAX,
421 [RTM_FAM(RTM_NEWTFILTER)] = TCA_MAX,
422 [RTM_FAM(RTM_NEWACTION)] = TCAA_MAX,
425 void __rta_fill(struct sk_buff *skb, int attrtype, int attrlen, const void *data)
428 int size = RTA_LENGTH(attrlen);
430 rta = (struct rtattr*)skb_put(skb, RTA_ALIGN(size));
431 rta->rta_type = attrtype;
433 memcpy(RTA_DATA(rta), data, attrlen);
434 memset(RTA_DATA(rta) + attrlen, 0, RTA_ALIGN(size) - size);
437 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned group, int echo)
439 struct sock *rtnl = net->rtnl;
442 NETLINK_CB(skb).dst_group = group;
444 atomic_inc(&skb->users);
445 netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
447 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
451 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
453 struct sock *rtnl = net->rtnl;
455 return nlmsg_unicast(rtnl, skb, pid);
458 int rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
459 struct nlmsghdr *nlh, gfp_t flags)
461 struct sock *rtnl = net->rtnl;
465 report = nlmsg_report(nlh);
467 return nlmsg_notify(rtnl, skb, pid, group, report, flags);
470 void rtnl_set_sk_err(struct net *net, u32 group, int error)
472 struct sock *rtnl = net->rtnl;
474 netlink_set_err(rtnl, 0, group, error);
477 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
482 mx = nla_nest_start(skb, RTA_METRICS);
486 for (i = 0; i < RTAX_MAX; i++) {
489 NLA_PUT_U32(skb, i+1, metrics[i]);
494 nla_nest_cancel(skb, mx);
498 return nla_nest_end(skb, mx);
501 nla_nest_cancel(skb, mx);
505 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
506 u32 ts, u32 tsage, long expires, u32 error)
508 struct rta_cacheinfo ci = {
509 .rta_lastuse = jiffies_to_clock_t(jiffies - dst->lastuse),
510 .rta_used = dst->__use,
511 .rta_clntref = atomic_read(&(dst->__refcnt)),
519 ci.rta_expires = jiffies_to_clock_t(expires);
521 return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
524 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
526 static void set_operstate(struct net_device *dev, unsigned char transition)
528 unsigned char operstate = dev->operstate;
532 if ((operstate == IF_OPER_DORMANT ||
533 operstate == IF_OPER_UNKNOWN) &&
535 operstate = IF_OPER_UP;
538 case IF_OPER_DORMANT:
539 if (operstate == IF_OPER_UP ||
540 operstate == IF_OPER_UNKNOWN)
541 operstate = IF_OPER_DORMANT;
545 if (dev->operstate != operstate) {
546 write_lock_bh(&dev_base_lock);
547 dev->operstate = operstate;
548 write_unlock_bh(&dev_base_lock);
549 netdev_state_change(dev);
553 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
554 struct net_device_stats *b)
556 a->rx_packets = b->rx_packets;
557 a->tx_packets = b->tx_packets;
558 a->rx_bytes = b->rx_bytes;
559 a->tx_bytes = b->tx_bytes;
560 a->rx_errors = b->rx_errors;
561 a->tx_errors = b->tx_errors;
562 a->rx_dropped = b->rx_dropped;
563 a->tx_dropped = b->tx_dropped;
565 a->multicast = b->multicast;
566 a->collisions = b->collisions;
568 a->rx_length_errors = b->rx_length_errors;
569 a->rx_over_errors = b->rx_over_errors;
570 a->rx_crc_errors = b->rx_crc_errors;
571 a->rx_frame_errors = b->rx_frame_errors;
572 a->rx_fifo_errors = b->rx_fifo_errors;
573 a->rx_missed_errors = b->rx_missed_errors;
575 a->tx_aborted_errors = b->tx_aborted_errors;
576 a->tx_carrier_errors = b->tx_carrier_errors;
577 a->tx_fifo_errors = b->tx_fifo_errors;
578 a->tx_heartbeat_errors = b->tx_heartbeat_errors;
579 a->tx_window_errors = b->tx_window_errors;
581 a->rx_compressed = b->rx_compressed;
582 a->tx_compressed = b->tx_compressed;
585 static inline size_t if_nlmsg_size(const struct net_device *dev)
587 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
588 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
589 + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
590 + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
591 + nla_total_size(sizeof(struct rtnl_link_ifmap))
592 + nla_total_size(sizeof(struct rtnl_link_stats))
593 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
594 + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
595 + nla_total_size(4) /* IFLA_TXQLEN */
596 + nla_total_size(4) /* IFLA_WEIGHT */
597 + nla_total_size(4) /* IFLA_MTU */
598 + nla_total_size(4) /* IFLA_LINK */
599 + nla_total_size(4) /* IFLA_MASTER */
600 + nla_total_size(1) /* IFLA_OPERSTATE */
601 + nla_total_size(1) /* IFLA_LINKMODE */
602 + rtnl_link_get_size(dev); /* IFLA_LINKINFO */
605 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
606 int type, u32 pid, u32 seq, u32 change,
609 struct netdev_queue *txq;
610 struct ifinfomsg *ifm;
611 struct nlmsghdr *nlh;
612 struct net_device_stats *stats;
615 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
619 ifm = nlmsg_data(nlh);
620 ifm->ifi_family = AF_UNSPEC;
622 ifm->ifi_type = dev->type;
623 ifm->ifi_index = dev->ifindex;
624 ifm->ifi_flags = dev_get_flags(dev);
625 ifm->ifi_change = change;
627 NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
628 NLA_PUT_U32(skb, IFLA_TXQLEN, dev->tx_queue_len);
629 NLA_PUT_U8(skb, IFLA_OPERSTATE,
630 netif_running(dev) ? dev->operstate : IF_OPER_DOWN);
631 NLA_PUT_U8(skb, IFLA_LINKMODE, dev->link_mode);
632 NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
634 if (dev->ifindex != dev->iflink)
635 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
638 NLA_PUT_U32(skb, IFLA_MASTER, dev->master->ifindex);
640 txq = netdev_get_tx_queue(dev, 0);
641 if (txq->qdisc_sleeping)
642 NLA_PUT_STRING(skb, IFLA_QDISC, txq->qdisc_sleeping->ops->id);
645 NLA_PUT_STRING(skb, IFLA_IFALIAS, dev->ifalias);
648 struct rtnl_link_ifmap map = {
649 .mem_start = dev->mem_start,
650 .mem_end = dev->mem_end,
651 .base_addr = dev->base_addr,
654 .port = dev->if_port,
656 NLA_PUT(skb, IFLA_MAP, sizeof(map), &map);
660 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
661 NLA_PUT(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast);
664 attr = nla_reserve(skb, IFLA_STATS,
665 sizeof(struct rtnl_link_stats));
667 goto nla_put_failure;
669 stats = dev->get_stats(dev);
670 copy_rtnl_link_stats(nla_data(attr), stats);
672 if (dev->rtnl_link_ops) {
673 if (rtnl_link_fill(skb, dev) < 0)
674 goto nla_put_failure;
677 return nlmsg_end(skb, nlh);
680 nlmsg_cancel(skb, nlh);
684 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
686 struct net *net = sock_net(skb->sk);
688 int s_idx = cb->args[0];
689 struct net_device *dev;
692 for_each_netdev(net, dev) {
695 if (rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
696 NETLINK_CB(cb->skb).pid,
697 cb->nlh->nlmsg_seq, 0, NLM_F_MULTI) <= 0)
707 const struct nla_policy ifla_policy[IFLA_MAX+1] = {
708 [IFLA_IFNAME] = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
709 [IFLA_ADDRESS] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
710 [IFLA_BROADCAST] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
711 [IFLA_MAP] = { .len = sizeof(struct rtnl_link_ifmap) },
712 [IFLA_MTU] = { .type = NLA_U32 },
713 [IFLA_LINK] = { .type = NLA_U32 },
714 [IFLA_TXQLEN] = { .type = NLA_U32 },
715 [IFLA_WEIGHT] = { .type = NLA_U32 },
716 [IFLA_OPERSTATE] = { .type = NLA_U8 },
717 [IFLA_LINKMODE] = { .type = NLA_U8 },
718 [IFLA_LINKINFO] = { .type = NLA_NESTED },
719 [IFLA_NET_NS_PID] = { .type = NLA_U32 },
720 [IFLA_IFALIAS] = { .type = NLA_STRING, .len = IFALIASZ-1 },
723 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
724 [IFLA_INFO_KIND] = { .type = NLA_STRING },
725 [IFLA_INFO_DATA] = { .type = NLA_NESTED },
728 static struct net *get_net_ns_by_pid(pid_t pid)
730 struct task_struct *tsk;
733 /* Lookup the network namespace */
734 net = ERR_PTR(-ESRCH);
736 tsk = find_task_by_vpid(pid);
738 struct nsproxy *nsproxy;
739 nsproxy = task_nsproxy(tsk);
741 net = get_net(nsproxy->net_ns);
747 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
750 if (tb[IFLA_ADDRESS] &&
751 nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
754 if (tb[IFLA_BROADCAST] &&
755 nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
762 static int do_setlink(struct net_device *dev, struct ifinfomsg *ifm,
763 struct nlattr **tb, char *ifname, int modified)
765 int send_addr_notify = 0;
768 if (tb[IFLA_NET_NS_PID]) {
770 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
775 err = dev_change_net_namespace(dev, net, ifname);
783 struct rtnl_link_ifmap *u_map;
786 if (!dev->set_config) {
791 if (!netif_device_present(dev)) {
796 u_map = nla_data(tb[IFLA_MAP]);
797 k_map.mem_start = (unsigned long) u_map->mem_start;
798 k_map.mem_end = (unsigned long) u_map->mem_end;
799 k_map.base_addr = (unsigned short) u_map->base_addr;
800 k_map.irq = (unsigned char) u_map->irq;
801 k_map.dma = (unsigned char) u_map->dma;
802 k_map.port = (unsigned char) u_map->port;
804 err = dev->set_config(dev, &k_map);
811 if (tb[IFLA_ADDRESS]) {
815 if (!dev->set_mac_address) {
820 if (!netif_device_present(dev)) {
825 len = sizeof(sa_family_t) + dev->addr_len;
826 sa = kmalloc(len, GFP_KERNEL);
831 sa->sa_family = dev->type;
832 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
834 err = dev->set_mac_address(dev, sa);
838 send_addr_notify = 1;
843 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
850 * Interface selected by interface index but interface
851 * name provided implies that a name change has been
854 if (ifm->ifi_index > 0 && ifname[0]) {
855 err = dev_change_name(dev, ifname);
861 if (tb[IFLA_IFALIAS]) {
862 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
863 nla_len(tb[IFLA_IFALIAS]));
869 if (tb[IFLA_BROADCAST]) {
870 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
871 send_addr_notify = 1;
874 if (ifm->ifi_flags || ifm->ifi_change) {
875 unsigned int flags = ifm->ifi_flags;
877 /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
879 flags = (flags & ifm->ifi_change) |
880 (dev->flags & ~ifm->ifi_change);
881 err = dev_change_flags(dev, flags);
887 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
889 if (tb[IFLA_OPERSTATE])
890 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
892 if (tb[IFLA_LINKMODE]) {
893 write_lock_bh(&dev_base_lock);
894 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
895 write_unlock_bh(&dev_base_lock);
901 if (err < 0 && modified && net_ratelimit())
902 printk(KERN_WARNING "A link change request failed with "
903 "some changes comitted already. Interface %s may "
904 "have been left with an inconsistent configuration, "
905 "please check.\n", dev->name);
907 if (send_addr_notify)
908 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
912 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
914 struct net *net = sock_net(skb->sk);
915 struct ifinfomsg *ifm;
916 struct net_device *dev;
918 struct nlattr *tb[IFLA_MAX+1];
919 char ifname[IFNAMSIZ];
921 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
926 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
931 ifm = nlmsg_data(nlh);
932 if (ifm->ifi_index > 0)
933 dev = dev_get_by_index(net, ifm->ifi_index);
934 else if (tb[IFLA_IFNAME])
935 dev = dev_get_by_name(net, ifname);
944 if ((err = validate_linkmsg(dev, tb)) < 0)
947 err = do_setlink(dev, ifm, tb, ifname, 0);
954 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
956 struct net *net = sock_net(skb->sk);
957 const struct rtnl_link_ops *ops;
958 struct net_device *dev;
959 struct ifinfomsg *ifm;
960 char ifname[IFNAMSIZ];
961 struct nlattr *tb[IFLA_MAX+1];
964 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
969 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
971 ifm = nlmsg_data(nlh);
972 if (ifm->ifi_index > 0)
973 dev = __dev_get_by_index(net, ifm->ifi_index);
974 else if (tb[IFLA_IFNAME])
975 dev = __dev_get_by_name(net, ifname);
982 ops = dev->rtnl_link_ops;
990 struct net_device *rtnl_create_link(struct net *net, char *ifname,
991 const struct rtnl_link_ops *ops, struct nlattr *tb[])
994 struct net_device *dev;
997 dev = alloc_netdev(ops->priv_size, ifname, ops->setup);
1001 if (strchr(dev->name, '%')) {
1002 err = dev_alloc_name(dev, dev->name);
1007 dev_net_set(dev, net);
1008 dev->rtnl_link_ops = ops;
1011 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
1012 if (tb[IFLA_ADDRESS])
1013 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
1014 nla_len(tb[IFLA_ADDRESS]));
1015 if (tb[IFLA_BROADCAST])
1016 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
1017 nla_len(tb[IFLA_BROADCAST]));
1018 if (tb[IFLA_TXQLEN])
1019 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1020 if (tb[IFLA_OPERSTATE])
1021 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1022 if (tb[IFLA_LINKMODE])
1023 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1030 return ERR_PTR(err);
1033 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1035 struct net *net = sock_net(skb->sk);
1036 const struct rtnl_link_ops *ops;
1037 struct net_device *dev;
1038 struct ifinfomsg *ifm;
1039 char kind[MODULE_NAME_LEN];
1040 char ifname[IFNAMSIZ];
1041 struct nlattr *tb[IFLA_MAX+1];
1042 struct nlattr *linkinfo[IFLA_INFO_MAX+1];
1045 #ifdef CONFIG_MODULES
1048 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1052 if (tb[IFLA_IFNAME])
1053 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1057 ifm = nlmsg_data(nlh);
1058 if (ifm->ifi_index > 0)
1059 dev = __dev_get_by_index(net, ifm->ifi_index);
1061 dev = __dev_get_by_name(net, ifname);
1065 if ((err = validate_linkmsg(dev, tb)) < 0)
1068 if (tb[IFLA_LINKINFO]) {
1069 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
1070 tb[IFLA_LINKINFO], ifla_info_policy);
1074 memset(linkinfo, 0, sizeof(linkinfo));
1076 if (linkinfo[IFLA_INFO_KIND]) {
1077 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
1078 ops = rtnl_link_ops_get(kind);
1085 struct nlattr *attr[ops ? ops->maxtype + 1 : 0], **data = NULL;
1088 if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
1089 err = nla_parse_nested(attr, ops->maxtype,
1090 linkinfo[IFLA_INFO_DATA],
1096 if (ops->validate) {
1097 err = ops->validate(tb, data);
1106 if (nlh->nlmsg_flags & NLM_F_EXCL)
1108 if (nlh->nlmsg_flags & NLM_F_REPLACE)
1111 if (linkinfo[IFLA_INFO_DATA]) {
1112 if (!ops || ops != dev->rtnl_link_ops ||
1116 err = ops->changelink(dev, tb, data);
1122 return do_setlink(dev, ifm, tb, ifname, modified);
1125 if (!(nlh->nlmsg_flags & NLM_F_CREATE))
1128 if (ifm->ifi_index || ifm->ifi_flags || ifm->ifi_change)
1130 if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
1134 #ifdef CONFIG_MODULES
1137 request_module("rtnl-link-%s", kind);
1139 ops = rtnl_link_ops_get(kind);
1148 snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
1150 dev = rtnl_create_link(net, ifname, ops, tb);
1154 else if (ops->newlink)
1155 err = ops->newlink(dev, tb, data);
1157 err = register_netdevice(dev);
1159 if (err < 0 && !IS_ERR(dev))
1165 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
1167 struct net *net = sock_net(skb->sk);
1168 struct ifinfomsg *ifm;
1169 struct nlattr *tb[IFLA_MAX+1];
1170 struct net_device *dev = NULL;
1171 struct sk_buff *nskb;
1174 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1178 ifm = nlmsg_data(nlh);
1179 if (ifm->ifi_index > 0) {
1180 dev = dev_get_by_index(net, ifm->ifi_index);
1186 nskb = nlmsg_new(if_nlmsg_size(dev), GFP_KERNEL);
1192 err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).pid,
1193 nlh->nlmsg_seq, 0, 0);
1195 /* -EMSGSIZE implies BUG in if_nlmsg_size */
1196 WARN_ON(err == -EMSGSIZE);
1200 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).pid);
1207 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
1210 int s_idx = cb->family;
1214 for (idx=1; idx<NPROTO; idx++) {
1215 int type = cb->nlh->nlmsg_type-RTM_BASE;
1216 if (idx < s_idx || idx == PF_PACKET)
1218 if (rtnl_msg_handlers[idx] == NULL ||
1219 rtnl_msg_handlers[idx][type].dumpit == NULL)
1222 memset(&cb->args[0], 0, sizeof(cb->args));
1223 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
1231 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned change)
1233 struct net *net = dev_net(dev);
1234 struct sk_buff *skb;
1237 skb = nlmsg_new(if_nlmsg_size(dev), GFP_KERNEL);
1241 err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0);
1243 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
1244 WARN_ON(err == -EMSGSIZE);
1248 err = rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_KERNEL);
1251 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
1254 /* Protected by RTNL sempahore. */
1255 static struct rtattr **rta_buf;
1256 static int rtattr_max;
1258 /* Process one rtnetlink message. */
1260 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
1262 struct net *net = sock_net(skb->sk);
1263 rtnl_doit_func doit;
1270 type = nlh->nlmsg_type;
1276 /* All the messages must have at least 1 byte length */
1277 if (nlh->nlmsg_len < NLMSG_LENGTH(sizeof(struct rtgenmsg)))
1280 family = ((struct rtgenmsg*)NLMSG_DATA(nlh))->rtgen_family;
1281 if (family >= NPROTO)
1282 return -EAFNOSUPPORT;
1287 if (kind != 2 && security_netlink_recv(skb, CAP_NET_ADMIN))
1290 if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
1292 rtnl_dumpit_func dumpit;
1294 dumpit = rtnl_get_dumpit(family, type);
1300 err = netlink_dump_start(rtnl, skb, nlh, dumpit, NULL);
1305 memset(rta_buf, 0, (rtattr_max * sizeof(struct rtattr *)));
1307 min_len = rtm_min[sz_idx];
1308 if (nlh->nlmsg_len < min_len)
1311 if (nlh->nlmsg_len > min_len) {
1312 int attrlen = nlh->nlmsg_len - NLMSG_ALIGN(min_len);
1313 struct rtattr *attr = (void*)nlh + NLMSG_ALIGN(min_len);
1315 while (RTA_OK(attr, attrlen)) {
1316 unsigned flavor = attr->rta_type;
1318 if (flavor > rta_max[sz_idx])
1320 rta_buf[flavor-1] = attr;
1322 attr = RTA_NEXT(attr, attrlen);
1326 doit = rtnl_get_doit(family, type);
1330 return doit(skb, nlh, (void *)&rta_buf[0]);
1333 static void rtnetlink_rcv(struct sk_buff *skb)
1336 netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
1340 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
1342 struct net_device *dev = ptr;
1345 case NETDEV_UNREGISTER:
1346 rtmsg_ifinfo(RTM_DELLINK, dev, ~0U);
1348 case NETDEV_REGISTER:
1349 rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U);
1353 rtmsg_ifinfo(RTM_NEWLINK, dev, IFF_UP|IFF_RUNNING);
1356 case NETDEV_GOING_DOWN:
1359 rtmsg_ifinfo(RTM_NEWLINK, dev, 0);
1365 static struct notifier_block rtnetlink_dev_notifier = {
1366 .notifier_call = rtnetlink_event,
1370 static int rtnetlink_net_init(struct net *net)
1373 sk = netlink_kernel_create(net, NETLINK_ROUTE, RTNLGRP_MAX,
1374 rtnetlink_rcv, &rtnl_mutex, THIS_MODULE);
1381 static void rtnetlink_net_exit(struct net *net)
1383 netlink_kernel_release(net->rtnl);
1387 static struct pernet_operations rtnetlink_net_ops = {
1388 .init = rtnetlink_net_init,
1389 .exit = rtnetlink_net_exit,
1392 void __init rtnetlink_init(void)
1397 for (i = 0; i < ARRAY_SIZE(rta_max); i++)
1398 if (rta_max[i] > rtattr_max)
1399 rtattr_max = rta_max[i];
1400 rta_buf = kmalloc(rtattr_max * sizeof(struct rtattr *), GFP_KERNEL);
1402 panic("rtnetlink_init: cannot allocate rta_buf\n");
1404 if (register_pernet_subsys(&rtnetlink_net_ops))
1405 panic("rtnetlink_init: cannot initialize rtnetlink\n");
1407 netlink_set_nonroot(NETLINK_ROUTE, NL_NONROOT_RECV);
1408 register_netdevice_notifier(&rtnetlink_dev_notifier);
1410 rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink, rtnl_dump_ifinfo);
1411 rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL);
1412 rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL);
1413 rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL);
1415 rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all);
1416 rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all);
1419 EXPORT_SYMBOL(__rta_fill);
1420 EXPORT_SYMBOL(rtnetlink_put_metrics);
1421 EXPORT_SYMBOL(rtnl_lock);
1422 EXPORT_SYMBOL(rtnl_trylock);
1423 EXPORT_SYMBOL(rtnl_unlock);
1424 EXPORT_SYMBOL(rtnl_is_locked);
1425 EXPORT_SYMBOL(rtnl_unicast);
1426 EXPORT_SYMBOL(rtnl_notify);
1427 EXPORT_SYMBOL(rtnl_set_sk_err);
1428 EXPORT_SYMBOL(rtnl_create_link);
1429 EXPORT_SYMBOL(ifla_policy);