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 mutex_unlock(&rtnl_mutex);
80 int rtnl_trylock(void)
82 return mutex_trylock(&rtnl_mutex);
85 static struct rtnl_link *rtnl_msg_handlers[NPROTO];
87 static inline int rtm_msgindex(int msgtype)
89 int msgindex = msgtype - RTM_BASE;
92 * msgindex < 0 implies someone tried to register a netlink
93 * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
94 * the message type has not been added to linux/rtnetlink.h
96 BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
101 static rtnl_doit_func rtnl_get_doit(int protocol, int msgindex)
103 struct rtnl_link *tab;
105 tab = rtnl_msg_handlers[protocol];
106 if (tab == NULL || tab[msgindex].doit == NULL)
107 tab = rtnl_msg_handlers[PF_UNSPEC];
109 return tab ? tab[msgindex].doit : NULL;
112 static rtnl_dumpit_func rtnl_get_dumpit(int protocol, int msgindex)
114 struct rtnl_link *tab;
116 tab = rtnl_msg_handlers[protocol];
117 if (tab == NULL || tab[msgindex].dumpit == NULL)
118 tab = rtnl_msg_handlers[PF_UNSPEC];
120 return tab ? tab[msgindex].dumpit : NULL;
124 * __rtnl_register - Register a rtnetlink message type
125 * @protocol: Protocol family or PF_UNSPEC
126 * @msgtype: rtnetlink message type
127 * @doit: Function pointer called for each request message
128 * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
130 * Registers the specified function pointers (at least one of them has
131 * to be non-NULL) to be called whenever a request message for the
132 * specified protocol family and message type is received.
134 * The special protocol family PF_UNSPEC may be used to define fallback
135 * function pointers for the case when no entry for the specific protocol
138 * Returns 0 on success or a negative error code.
140 int __rtnl_register(int protocol, int msgtype,
141 rtnl_doit_func doit, rtnl_dumpit_func dumpit)
143 struct rtnl_link *tab;
146 BUG_ON(protocol < 0 || protocol >= NPROTO);
147 msgindex = rtm_msgindex(msgtype);
149 tab = rtnl_msg_handlers[protocol];
151 tab = kcalloc(RTM_NR_MSGTYPES, sizeof(*tab), GFP_KERNEL);
155 rtnl_msg_handlers[protocol] = tab;
159 tab[msgindex].doit = doit;
162 tab[msgindex].dumpit = dumpit;
167 EXPORT_SYMBOL_GPL(__rtnl_register);
170 * rtnl_register - Register a rtnetlink message type
172 * Identical to __rtnl_register() but panics on failure. This is useful
173 * as failure of this function is very unlikely, it can only happen due
174 * to lack of memory when allocating the chain to store all message
175 * handlers for a protocol. Meant for use in init functions where lack
176 * of memory implies no sense in continueing.
178 void rtnl_register(int protocol, int msgtype,
179 rtnl_doit_func doit, rtnl_dumpit_func dumpit)
181 if (__rtnl_register(protocol, msgtype, doit, dumpit) < 0)
182 panic("Unable to register rtnetlink message handler, "
183 "protocol = %d, message type = %d\n",
187 EXPORT_SYMBOL_GPL(rtnl_register);
190 * rtnl_unregister - Unregister a rtnetlink message type
191 * @protocol: Protocol family or PF_UNSPEC
192 * @msgtype: rtnetlink message type
194 * Returns 0 on success or a negative error code.
196 int rtnl_unregister(int protocol, int msgtype)
200 BUG_ON(protocol < 0 || protocol >= NPROTO);
201 msgindex = rtm_msgindex(msgtype);
203 if (rtnl_msg_handlers[protocol] == NULL)
206 rtnl_msg_handlers[protocol][msgindex].doit = NULL;
207 rtnl_msg_handlers[protocol][msgindex].dumpit = NULL;
212 EXPORT_SYMBOL_GPL(rtnl_unregister);
215 * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
216 * @protocol : Protocol family or PF_UNSPEC
218 * Identical to calling rtnl_unregster() for all registered message types
219 * of a certain protocol family.
221 void rtnl_unregister_all(int protocol)
223 BUG_ON(protocol < 0 || protocol >= NPROTO);
225 kfree(rtnl_msg_handlers[protocol]);
226 rtnl_msg_handlers[protocol] = NULL;
229 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
231 static LIST_HEAD(link_ops);
234 * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
235 * @ops: struct rtnl_link_ops * to register
237 * The caller must hold the rtnl_mutex. This function should be used
238 * by drivers that create devices during module initialization. It
239 * must be called before registering the devices.
241 * Returns 0 on success or a negative error code.
243 int __rtnl_link_register(struct rtnl_link_ops *ops)
246 ops->dellink = unregister_netdevice;
248 list_add_tail(&ops->list, &link_ops);
252 EXPORT_SYMBOL_GPL(__rtnl_link_register);
255 * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
256 * @ops: struct rtnl_link_ops * to register
258 * Returns 0 on success or a negative error code.
260 int rtnl_link_register(struct rtnl_link_ops *ops)
265 err = __rtnl_link_register(ops);
270 EXPORT_SYMBOL_GPL(rtnl_link_register);
273 * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
274 * @ops: struct rtnl_link_ops * to unregister
276 * The caller must hold the rtnl_mutex.
278 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
280 struct net_device *dev, *n;
285 for_each_netdev_safe(net, dev, n) {
286 if (dev->rtnl_link_ops == ops) {
292 list_del(&ops->list);
295 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
298 * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
299 * @ops: struct rtnl_link_ops * to unregister
301 void rtnl_link_unregister(struct rtnl_link_ops *ops)
304 __rtnl_link_unregister(ops);
308 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
310 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
312 const struct rtnl_link_ops *ops;
314 list_for_each_entry(ops, &link_ops, list) {
315 if (!strcmp(ops->kind, kind))
321 static size_t rtnl_link_get_size(const struct net_device *dev)
323 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
329 size = nlmsg_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
330 nlmsg_total_size(strlen(ops->kind) + 1); /* IFLA_INFO_KIND */
333 /* IFLA_INFO_DATA + nested data */
334 size += nlmsg_total_size(sizeof(struct nlattr)) +
337 if (ops->get_xstats_size)
338 size += ops->get_xstats_size(dev); /* IFLA_INFO_XSTATS */
343 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
345 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
346 struct nlattr *linkinfo, *data;
349 linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
350 if (linkinfo == NULL)
353 if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
354 goto err_cancel_link;
355 if (ops->fill_xstats) {
356 err = ops->fill_xstats(skb, dev);
358 goto err_cancel_link;
360 if (ops->fill_info) {
361 data = nla_nest_start(skb, IFLA_INFO_DATA);
363 goto err_cancel_link;
364 err = ops->fill_info(skb, dev);
366 goto err_cancel_data;
367 nla_nest_end(skb, data);
370 nla_nest_end(skb, linkinfo);
374 nla_nest_cancel(skb, data);
376 nla_nest_cancel(skb, linkinfo);
381 static const int rtm_min[RTM_NR_FAMILIES] =
383 [RTM_FAM(RTM_NEWLINK)] = NLMSG_LENGTH(sizeof(struct ifinfomsg)),
384 [RTM_FAM(RTM_NEWADDR)] = NLMSG_LENGTH(sizeof(struct ifaddrmsg)),
385 [RTM_FAM(RTM_NEWROUTE)] = NLMSG_LENGTH(sizeof(struct rtmsg)),
386 [RTM_FAM(RTM_NEWRULE)] = NLMSG_LENGTH(sizeof(struct fib_rule_hdr)),
387 [RTM_FAM(RTM_NEWQDISC)] = NLMSG_LENGTH(sizeof(struct tcmsg)),
388 [RTM_FAM(RTM_NEWTCLASS)] = NLMSG_LENGTH(sizeof(struct tcmsg)),
389 [RTM_FAM(RTM_NEWTFILTER)] = NLMSG_LENGTH(sizeof(struct tcmsg)),
390 [RTM_FAM(RTM_NEWACTION)] = NLMSG_LENGTH(sizeof(struct tcamsg)),
391 [RTM_FAM(RTM_GETMULTICAST)] = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
392 [RTM_FAM(RTM_GETANYCAST)] = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
395 static const int rta_max[RTM_NR_FAMILIES] =
397 [RTM_FAM(RTM_NEWLINK)] = IFLA_MAX,
398 [RTM_FAM(RTM_NEWADDR)] = IFA_MAX,
399 [RTM_FAM(RTM_NEWROUTE)] = RTA_MAX,
400 [RTM_FAM(RTM_NEWRULE)] = FRA_MAX,
401 [RTM_FAM(RTM_NEWQDISC)] = TCA_MAX,
402 [RTM_FAM(RTM_NEWTCLASS)] = TCA_MAX,
403 [RTM_FAM(RTM_NEWTFILTER)] = TCA_MAX,
404 [RTM_FAM(RTM_NEWACTION)] = TCAA_MAX,
407 void __rta_fill(struct sk_buff *skb, int attrtype, int attrlen, const void *data)
410 int size = RTA_LENGTH(attrlen);
412 rta = (struct rtattr*)skb_put(skb, RTA_ALIGN(size));
413 rta->rta_type = attrtype;
415 memcpy(RTA_DATA(rta), data, attrlen);
416 memset(RTA_DATA(rta) + attrlen, 0, RTA_ALIGN(size) - size);
419 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned group, int echo)
421 struct sock *rtnl = net->rtnl;
424 NETLINK_CB(skb).dst_group = group;
426 atomic_inc(&skb->users);
427 netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
429 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
433 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
435 struct sock *rtnl = net->rtnl;
437 return nlmsg_unicast(rtnl, skb, pid);
440 int rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
441 struct nlmsghdr *nlh, gfp_t flags)
443 struct sock *rtnl = net->rtnl;
447 report = nlmsg_report(nlh);
449 return nlmsg_notify(rtnl, skb, pid, group, report, flags);
452 void rtnl_set_sk_err(struct net *net, u32 group, int error)
454 struct sock *rtnl = net->rtnl;
456 netlink_set_err(rtnl, 0, group, error);
459 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
464 mx = nla_nest_start(skb, RTA_METRICS);
468 for (i = 0; i < RTAX_MAX; i++) {
471 NLA_PUT_U32(skb, i+1, metrics[i]);
476 nla_nest_cancel(skb, mx);
480 return nla_nest_end(skb, mx);
483 return nla_nest_cancel(skb, mx);
486 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
487 u32 ts, u32 tsage, long expires, u32 error)
489 struct rta_cacheinfo ci = {
490 .rta_lastuse = jiffies_to_clock_t(jiffies - dst->lastuse),
491 .rta_used = dst->__use,
492 .rta_clntref = atomic_read(&(dst->__refcnt)),
500 ci.rta_expires = jiffies_to_clock_t(expires);
502 return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
505 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
507 static void set_operstate(struct net_device *dev, unsigned char transition)
509 unsigned char operstate = dev->operstate;
513 if ((operstate == IF_OPER_DORMANT ||
514 operstate == IF_OPER_UNKNOWN) &&
516 operstate = IF_OPER_UP;
519 case IF_OPER_DORMANT:
520 if (operstate == IF_OPER_UP ||
521 operstate == IF_OPER_UNKNOWN)
522 operstate = IF_OPER_DORMANT;
526 if (dev->operstate != operstate) {
527 write_lock_bh(&dev_base_lock);
528 dev->operstate = operstate;
529 write_unlock_bh(&dev_base_lock);
530 netdev_state_change(dev);
534 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
535 struct net_device_stats *b)
537 a->rx_packets = b->rx_packets;
538 a->tx_packets = b->tx_packets;
539 a->rx_bytes = b->rx_bytes;
540 a->tx_bytes = b->tx_bytes;
541 a->rx_errors = b->rx_errors;
542 a->tx_errors = b->tx_errors;
543 a->rx_dropped = b->rx_dropped;
544 a->tx_dropped = b->tx_dropped;
546 a->multicast = b->multicast;
547 a->collisions = b->collisions;
549 a->rx_length_errors = b->rx_length_errors;
550 a->rx_over_errors = b->rx_over_errors;
551 a->rx_crc_errors = b->rx_crc_errors;
552 a->rx_frame_errors = b->rx_frame_errors;
553 a->rx_fifo_errors = b->rx_fifo_errors;
554 a->rx_missed_errors = b->rx_missed_errors;
556 a->tx_aborted_errors = b->tx_aborted_errors;
557 a->tx_carrier_errors = b->tx_carrier_errors;
558 a->tx_fifo_errors = b->tx_fifo_errors;
559 a->tx_heartbeat_errors = b->tx_heartbeat_errors;
560 a->tx_window_errors = b->tx_window_errors;
562 a->rx_compressed = b->rx_compressed;
563 a->tx_compressed = b->tx_compressed;
566 static inline size_t if_nlmsg_size(const struct net_device *dev)
568 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
569 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
570 + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
571 + nla_total_size(sizeof(struct rtnl_link_ifmap))
572 + nla_total_size(sizeof(struct rtnl_link_stats))
573 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
574 + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
575 + nla_total_size(4) /* IFLA_TXQLEN */
576 + nla_total_size(4) /* IFLA_WEIGHT */
577 + nla_total_size(4) /* IFLA_MTU */
578 + nla_total_size(4) /* IFLA_LINK */
579 + nla_total_size(4) /* IFLA_MASTER */
580 + nla_total_size(1) /* IFLA_OPERSTATE */
581 + nla_total_size(1) /* IFLA_LINKMODE */
582 + rtnl_link_get_size(dev); /* IFLA_LINKINFO */
585 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
586 int type, u32 pid, u32 seq, u32 change,
589 struct ifinfomsg *ifm;
590 struct nlmsghdr *nlh;
592 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
596 ifm = nlmsg_data(nlh);
597 ifm->ifi_family = AF_UNSPEC;
599 ifm->ifi_type = dev->type;
600 ifm->ifi_index = dev->ifindex;
601 ifm->ifi_flags = dev_get_flags(dev);
602 ifm->ifi_change = change;
604 NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
605 NLA_PUT_U32(skb, IFLA_TXQLEN, dev->tx_queue_len);
606 NLA_PUT_U8(skb, IFLA_OPERSTATE,
607 netif_running(dev) ? dev->operstate : IF_OPER_DOWN);
608 NLA_PUT_U8(skb, IFLA_LINKMODE, dev->link_mode);
609 NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
611 if (dev->ifindex != dev->iflink)
612 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
615 NLA_PUT_U32(skb, IFLA_MASTER, dev->master->ifindex);
617 if (dev->qdisc_sleeping)
618 NLA_PUT_STRING(skb, IFLA_QDISC, dev->qdisc_sleeping->ops->id);
621 struct rtnl_link_ifmap map = {
622 .mem_start = dev->mem_start,
623 .mem_end = dev->mem_end,
624 .base_addr = dev->base_addr,
627 .port = dev->if_port,
629 NLA_PUT(skb, IFLA_MAP, sizeof(map), &map);
633 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
634 NLA_PUT(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast);
637 if (dev->get_stats) {
638 struct net_device_stats *stats = dev->get_stats(dev);
642 attr = nla_reserve(skb, IFLA_STATS,
643 sizeof(struct rtnl_link_stats));
645 goto nla_put_failure;
647 copy_rtnl_link_stats(nla_data(attr), stats);
651 if (dev->rtnl_link_ops) {
652 if (rtnl_link_fill(skb, dev) < 0)
653 goto nla_put_failure;
656 return nlmsg_end(skb, nlh);
659 nlmsg_cancel(skb, nlh);
663 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
665 struct net *net = skb->sk->sk_net;
667 int s_idx = cb->args[0];
668 struct net_device *dev;
671 for_each_netdev(net, dev) {
674 if (rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
675 NETLINK_CB(cb->skb).pid,
676 cb->nlh->nlmsg_seq, 0, NLM_F_MULTI) <= 0)
686 const struct nla_policy ifla_policy[IFLA_MAX+1] = {
687 [IFLA_IFNAME] = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
688 [IFLA_ADDRESS] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
689 [IFLA_BROADCAST] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
690 [IFLA_MAP] = { .len = sizeof(struct rtnl_link_ifmap) },
691 [IFLA_MTU] = { .type = NLA_U32 },
692 [IFLA_TXQLEN] = { .type = NLA_U32 },
693 [IFLA_WEIGHT] = { .type = NLA_U32 },
694 [IFLA_OPERSTATE] = { .type = NLA_U8 },
695 [IFLA_LINKMODE] = { .type = NLA_U8 },
696 [IFLA_NET_NS_PID] = { .type = NLA_U32 },
699 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
700 [IFLA_INFO_KIND] = { .type = NLA_STRING },
701 [IFLA_INFO_DATA] = { .type = NLA_NESTED },
704 static struct net *get_net_ns_by_pid(pid_t pid)
706 struct task_struct *tsk;
709 /* Lookup the network namespace */
710 net = ERR_PTR(-ESRCH);
712 tsk = find_task_by_vpid(pid);
714 struct nsproxy *nsproxy;
715 nsproxy = task_nsproxy(tsk);
717 net = get_net(nsproxy->net_ns);
723 static int do_setlink(struct net_device *dev, struct ifinfomsg *ifm,
724 struct nlattr **tb, char *ifname, int modified)
726 int send_addr_notify = 0;
729 if (tb[IFLA_NET_NS_PID]) {
731 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
736 err = dev_change_net_namespace(dev, net, ifname);
744 struct rtnl_link_ifmap *u_map;
747 if (!dev->set_config) {
752 if (!netif_device_present(dev)) {
757 u_map = nla_data(tb[IFLA_MAP]);
758 k_map.mem_start = (unsigned long) u_map->mem_start;
759 k_map.mem_end = (unsigned long) u_map->mem_end;
760 k_map.base_addr = (unsigned short) u_map->base_addr;
761 k_map.irq = (unsigned char) u_map->irq;
762 k_map.dma = (unsigned char) u_map->dma;
763 k_map.port = (unsigned char) u_map->port;
765 err = dev->set_config(dev, &k_map);
772 if (tb[IFLA_ADDRESS]) {
776 if (!dev->set_mac_address) {
781 if (!netif_device_present(dev)) {
786 len = sizeof(sa_family_t) + dev->addr_len;
787 sa = kmalloc(len, GFP_KERNEL);
792 sa->sa_family = dev->type;
793 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
795 err = dev->set_mac_address(dev, sa);
799 send_addr_notify = 1;
804 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
811 * Interface selected by interface index but interface
812 * name provided implies that a name change has been
815 if (ifm->ifi_index > 0 && ifname[0]) {
816 err = dev_change_name(dev, ifname);
822 if (tb[IFLA_BROADCAST]) {
823 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
824 send_addr_notify = 1;
827 if (ifm->ifi_flags || ifm->ifi_change) {
828 unsigned int flags = ifm->ifi_flags;
830 /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
832 flags = (flags & ifm->ifi_change) |
833 (dev->flags & ~ifm->ifi_change);
834 dev_change_flags(dev, flags);
838 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
840 if (tb[IFLA_OPERSTATE])
841 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
843 if (tb[IFLA_LINKMODE]) {
844 write_lock_bh(&dev_base_lock);
845 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
846 write_unlock_bh(&dev_base_lock);
852 if (err < 0 && modified && net_ratelimit())
853 printk(KERN_WARNING "A link change request failed with "
854 "some changes comitted already. Interface %s may "
855 "have been left with an inconsistent configuration, "
856 "please check.\n", dev->name);
858 if (send_addr_notify)
859 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
863 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
865 struct net *net = skb->sk->sk_net;
866 struct ifinfomsg *ifm;
867 struct net_device *dev;
869 struct nlattr *tb[IFLA_MAX+1];
870 char ifname[IFNAMSIZ];
872 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
877 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
882 ifm = nlmsg_data(nlh);
883 if (ifm->ifi_index > 0)
884 dev = dev_get_by_index(net, ifm->ifi_index);
885 else if (tb[IFLA_IFNAME])
886 dev = dev_get_by_name(net, ifname);
895 if (tb[IFLA_ADDRESS] &&
896 nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
899 if (tb[IFLA_BROADCAST] &&
900 nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
903 err = do_setlink(dev, ifm, tb, ifname, 0);
910 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
912 struct net *net = skb->sk->sk_net;
913 const struct rtnl_link_ops *ops;
914 struct net_device *dev;
915 struct ifinfomsg *ifm;
916 char ifname[IFNAMSIZ];
917 struct nlattr *tb[IFLA_MAX+1];
920 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
925 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
927 ifm = nlmsg_data(nlh);
928 if (ifm->ifi_index > 0)
929 dev = __dev_get_by_index(net, ifm->ifi_index);
930 else if (tb[IFLA_IFNAME])
931 dev = __dev_get_by_name(net, ifname);
938 ops = dev->rtnl_link_ops;
946 struct net_device *rtnl_create_link(struct net *net, char *ifname,
947 const struct rtnl_link_ops *ops, struct nlattr *tb[])
950 struct net_device *dev;
953 dev = alloc_netdev(ops->priv_size, ifname, ops->setup);
957 if (strchr(dev->name, '%')) {
958 err = dev_alloc_name(dev, dev->name);
964 dev->rtnl_link_ops = ops;
967 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
968 if (tb[IFLA_ADDRESS])
969 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
970 nla_len(tb[IFLA_ADDRESS]));
971 if (tb[IFLA_BROADCAST])
972 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
973 nla_len(tb[IFLA_BROADCAST]));
975 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
976 if (tb[IFLA_OPERSTATE])
977 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
978 if (tb[IFLA_LINKMODE])
979 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
989 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
991 struct net *net = skb->sk->sk_net;
992 const struct rtnl_link_ops *ops;
993 struct net_device *dev;
994 struct ifinfomsg *ifm;
995 char kind[MODULE_NAME_LEN];
996 char ifname[IFNAMSIZ];
997 struct nlattr *tb[IFLA_MAX+1];
998 struct nlattr *linkinfo[IFLA_INFO_MAX+1];
1004 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1008 if (tb[IFLA_IFNAME])
1009 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1013 ifm = nlmsg_data(nlh);
1014 if (ifm->ifi_index > 0)
1015 dev = __dev_get_by_index(net, ifm->ifi_index);
1017 dev = __dev_get_by_name(net, ifname);
1021 if (tb[IFLA_LINKINFO]) {
1022 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
1023 tb[IFLA_LINKINFO], ifla_info_policy);
1027 memset(linkinfo, 0, sizeof(linkinfo));
1029 if (linkinfo[IFLA_INFO_KIND]) {
1030 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
1031 ops = rtnl_link_ops_get(kind);
1038 struct nlattr *attr[ops ? ops->maxtype + 1 : 0], **data = NULL;
1041 if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
1042 err = nla_parse_nested(attr, ops->maxtype,
1043 linkinfo[IFLA_INFO_DATA],
1049 if (ops->validate) {
1050 err = ops->validate(tb, data);
1059 if (nlh->nlmsg_flags & NLM_F_EXCL)
1061 if (nlh->nlmsg_flags & NLM_F_REPLACE)
1064 if (linkinfo[IFLA_INFO_DATA]) {
1065 if (!ops || ops != dev->rtnl_link_ops ||
1069 err = ops->changelink(dev, tb, data);
1075 return do_setlink(dev, ifm, tb, ifname, modified);
1078 if (!(nlh->nlmsg_flags & NLM_F_CREATE))
1081 if (ifm->ifi_index || ifm->ifi_flags || ifm->ifi_change)
1083 if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
1090 request_module("rtnl-link-%s", kind);
1092 ops = rtnl_link_ops_get(kind);
1101 snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
1103 dev = rtnl_create_link(net, ifname, ops, tb);
1107 else if (ops->newlink)
1108 err = ops->newlink(dev, tb, data);
1110 err = register_netdevice(dev);
1112 if (err < 0 && !IS_ERR(dev))
1118 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
1120 struct net *net = skb->sk->sk_net;
1121 struct ifinfomsg *ifm;
1122 struct nlattr *tb[IFLA_MAX+1];
1123 struct net_device *dev = NULL;
1124 struct sk_buff *nskb;
1127 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1131 ifm = nlmsg_data(nlh);
1132 if (ifm->ifi_index > 0) {
1133 dev = dev_get_by_index(net, ifm->ifi_index);
1139 nskb = nlmsg_new(if_nlmsg_size(dev), GFP_KERNEL);
1145 err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).pid,
1146 nlh->nlmsg_seq, 0, 0);
1148 /* -EMSGSIZE implies BUG in if_nlmsg_size */
1149 WARN_ON(err == -EMSGSIZE);
1153 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).pid);
1160 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
1163 int s_idx = cb->family;
1167 for (idx=1; idx<NPROTO; idx++) {
1168 int type = cb->nlh->nlmsg_type-RTM_BASE;
1169 if (idx < s_idx || idx == PF_PACKET)
1171 if (rtnl_msg_handlers[idx] == NULL ||
1172 rtnl_msg_handlers[idx][type].dumpit == NULL)
1175 memset(&cb->args[0], 0, sizeof(cb->args));
1176 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
1184 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned change)
1186 struct net *net = dev->nd_net;
1187 struct sk_buff *skb;
1190 skb = nlmsg_new(if_nlmsg_size(dev), GFP_KERNEL);
1194 err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0);
1196 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
1197 WARN_ON(err == -EMSGSIZE);
1201 err = rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_KERNEL);
1204 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
1207 /* Protected by RTNL sempahore. */
1208 static struct rtattr **rta_buf;
1209 static int rtattr_max;
1211 /* Process one rtnetlink message. */
1213 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
1215 struct net *net = skb->sk->sk_net;
1216 rtnl_doit_func doit;
1223 type = nlh->nlmsg_type;
1229 /* All the messages must have at least 1 byte length */
1230 if (nlh->nlmsg_len < NLMSG_LENGTH(sizeof(struct rtgenmsg)))
1233 family = ((struct rtgenmsg*)NLMSG_DATA(nlh))->rtgen_family;
1234 if (family >= NPROTO)
1235 return -EAFNOSUPPORT;
1240 if (kind != 2 && security_netlink_recv(skb, CAP_NET_ADMIN))
1243 if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
1245 rtnl_dumpit_func dumpit;
1247 dumpit = rtnl_get_dumpit(family, type);
1253 err = netlink_dump_start(rtnl, skb, nlh, dumpit, NULL);
1258 memset(rta_buf, 0, (rtattr_max * sizeof(struct rtattr *)));
1260 min_len = rtm_min[sz_idx];
1261 if (nlh->nlmsg_len < min_len)
1264 if (nlh->nlmsg_len > min_len) {
1265 int attrlen = nlh->nlmsg_len - NLMSG_ALIGN(min_len);
1266 struct rtattr *attr = (void*)nlh + NLMSG_ALIGN(min_len);
1268 while (RTA_OK(attr, attrlen)) {
1269 unsigned flavor = attr->rta_type;
1271 if (flavor > rta_max[sz_idx])
1273 rta_buf[flavor-1] = attr;
1275 attr = RTA_NEXT(attr, attrlen);
1279 doit = rtnl_get_doit(family, type);
1283 return doit(skb, nlh, (void *)&rta_buf[0]);
1286 static void rtnetlink_rcv(struct sk_buff *skb)
1289 netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
1293 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
1295 struct net_device *dev = ptr;
1298 case NETDEV_UNREGISTER:
1299 rtmsg_ifinfo(RTM_DELLINK, dev, ~0U);
1301 case NETDEV_REGISTER:
1302 rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U);
1306 rtmsg_ifinfo(RTM_NEWLINK, dev, IFF_UP|IFF_RUNNING);
1309 case NETDEV_GOING_DOWN:
1312 rtmsg_ifinfo(RTM_NEWLINK, dev, 0);
1318 static struct notifier_block rtnetlink_dev_notifier = {
1319 .notifier_call = rtnetlink_event,
1323 static int rtnetlink_net_init(struct net *net)
1326 sk = netlink_kernel_create(net, NETLINK_ROUTE, RTNLGRP_MAX,
1327 rtnetlink_rcv, &rtnl_mutex, THIS_MODULE);
1334 static void rtnetlink_net_exit(struct net *net)
1336 netlink_kernel_release(net->rtnl);
1340 static struct pernet_operations rtnetlink_net_ops = {
1341 .init = rtnetlink_net_init,
1342 .exit = rtnetlink_net_exit,
1345 void __init rtnetlink_init(void)
1350 for (i = 0; i < ARRAY_SIZE(rta_max); i++)
1351 if (rta_max[i] > rtattr_max)
1352 rtattr_max = rta_max[i];
1353 rta_buf = kmalloc(rtattr_max * sizeof(struct rtattr *), GFP_KERNEL);
1355 panic("rtnetlink_init: cannot allocate rta_buf\n");
1357 if (register_pernet_subsys(&rtnetlink_net_ops))
1358 panic("rtnetlink_init: cannot initialize rtnetlink\n");
1360 netlink_set_nonroot(NETLINK_ROUTE, NL_NONROOT_RECV);
1361 register_netdevice_notifier(&rtnetlink_dev_notifier);
1363 rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink, rtnl_dump_ifinfo);
1364 rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL);
1365 rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL);
1366 rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL);
1368 rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all);
1369 rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all);
1372 EXPORT_SYMBOL(__rta_fill);
1373 EXPORT_SYMBOL(rtnetlink_put_metrics);
1374 EXPORT_SYMBOL(rtnl_lock);
1375 EXPORT_SYMBOL(rtnl_trylock);
1376 EXPORT_SYMBOL(rtnl_unlock);
1377 EXPORT_SYMBOL(rtnl_unicast);
1378 EXPORT_SYMBOL(rtnl_notify);
1379 EXPORT_SYMBOL(rtnl_set_sk_err);
1380 EXPORT_SYMBOL(rtnl_create_link);
1381 EXPORT_SYMBOL(ifla_policy);