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>
39 #include <asm/uaccess.h>
40 #include <asm/system.h>
41 #include <asm/string.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>
58 rtnl_dumpit_func dumpit;
61 static DEFINE_MUTEX(rtnl_mutex);
62 static struct sock *rtnl;
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);
77 if (rtnl && rtnl->sk_receive_queue.qlen)
78 rtnl->sk_data_ready(rtnl, 0);
82 int rtnl_trylock(void)
84 return mutex_trylock(&rtnl_mutex);
87 int rtattr_parse(struct rtattr *tb[], int maxattr, struct rtattr *rta, int len)
89 memset(tb, 0, sizeof(struct rtattr*)*maxattr);
91 while (RTA_OK(rta, len)) {
92 unsigned flavor = rta->rta_type;
93 if (flavor && flavor <= maxattr)
95 rta = RTA_NEXT(rta, len);
100 int __rtattr_parse_nested_compat(struct rtattr *tb[], int maxattr,
101 struct rtattr *rta, int len)
103 if (RTA_PAYLOAD(rta) < len)
105 if (RTA_PAYLOAD(rta) >= RTA_ALIGN(len) + sizeof(struct rtattr)) {
106 rta = RTA_DATA(rta) + RTA_ALIGN(len);
107 return rtattr_parse_nested(tb, maxattr, rta);
109 memset(tb, 0, sizeof(struct rtattr *) * maxattr);
113 static struct rtnl_link *rtnl_msg_handlers[NPROTO];
115 static inline int rtm_msgindex(int msgtype)
117 int msgindex = msgtype - RTM_BASE;
120 * msgindex < 0 implies someone tried to register a netlink
121 * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
122 * the message type has not been added to linux/rtnetlink.h
124 BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
129 static rtnl_doit_func rtnl_get_doit(int protocol, int msgindex)
131 struct rtnl_link *tab;
133 tab = rtnl_msg_handlers[protocol];
134 if (tab == NULL || tab[msgindex].doit == NULL)
135 tab = rtnl_msg_handlers[PF_UNSPEC];
137 return tab ? tab[msgindex].doit : NULL;
140 static rtnl_dumpit_func rtnl_get_dumpit(int protocol, int msgindex)
142 struct rtnl_link *tab;
144 tab = rtnl_msg_handlers[protocol];
145 if (tab == NULL || tab[msgindex].dumpit == NULL)
146 tab = rtnl_msg_handlers[PF_UNSPEC];
148 return tab ? tab[msgindex].dumpit : NULL;
152 * __rtnl_register - Register a rtnetlink message type
153 * @protocol: Protocol family or PF_UNSPEC
154 * @msgtype: rtnetlink message type
155 * @doit: Function pointer called for each request message
156 * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
158 * Registers the specified function pointers (at least one of them has
159 * to be non-NULL) to be called whenever a request message for the
160 * specified protocol family and message type is received.
162 * The special protocol family PF_UNSPEC may be used to define fallback
163 * function pointers for the case when no entry for the specific protocol
166 * Returns 0 on success or a negative error code.
168 int __rtnl_register(int protocol, int msgtype,
169 rtnl_doit_func doit, rtnl_dumpit_func dumpit)
171 struct rtnl_link *tab;
174 BUG_ON(protocol < 0 || protocol >= NPROTO);
175 msgindex = rtm_msgindex(msgtype);
177 tab = rtnl_msg_handlers[protocol];
179 tab = kcalloc(RTM_NR_MSGTYPES, sizeof(*tab), GFP_KERNEL);
183 rtnl_msg_handlers[protocol] = tab;
187 tab[msgindex].doit = doit;
190 tab[msgindex].dumpit = dumpit;
195 EXPORT_SYMBOL_GPL(__rtnl_register);
198 * rtnl_register - Register a rtnetlink message type
200 * Identical to __rtnl_register() but panics on failure. This is useful
201 * as failure of this function is very unlikely, it can only happen due
202 * to lack of memory when allocating the chain to store all message
203 * handlers for a protocol. Meant for use in init functions where lack
204 * of memory implies no sense in continueing.
206 void rtnl_register(int protocol, int msgtype,
207 rtnl_doit_func doit, rtnl_dumpit_func dumpit)
209 if (__rtnl_register(protocol, msgtype, doit, dumpit) < 0)
210 panic("Unable to register rtnetlink message handler, "
211 "protocol = %d, message type = %d\n",
215 EXPORT_SYMBOL_GPL(rtnl_register);
218 * rtnl_unregister - Unregister a rtnetlink message type
219 * @protocol: Protocol family or PF_UNSPEC
220 * @msgtype: rtnetlink message type
222 * Returns 0 on success or a negative error code.
224 int rtnl_unregister(int protocol, int msgtype)
228 BUG_ON(protocol < 0 || protocol >= NPROTO);
229 msgindex = rtm_msgindex(msgtype);
231 if (rtnl_msg_handlers[protocol] == NULL)
234 rtnl_msg_handlers[protocol][msgindex].doit = NULL;
235 rtnl_msg_handlers[protocol][msgindex].dumpit = NULL;
240 EXPORT_SYMBOL_GPL(rtnl_unregister);
243 * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
244 * @protocol : Protocol family or PF_UNSPEC
246 * Identical to calling rtnl_unregster() for all registered message types
247 * of a certain protocol family.
249 void rtnl_unregister_all(int protocol)
251 BUG_ON(protocol < 0 || protocol >= NPROTO);
253 kfree(rtnl_msg_handlers[protocol]);
254 rtnl_msg_handlers[protocol] = NULL;
257 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
259 static LIST_HEAD(link_ops);
262 * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
263 * @ops: struct rtnl_link_ops * to register
265 * The caller must hold the rtnl_mutex. This function should be used
266 * by drivers that create devices during module initialization. It
267 * must be called before registering the devices.
269 * Returns 0 on success or a negative error code.
271 int __rtnl_link_register(struct rtnl_link_ops *ops)
274 ops->dellink = unregister_netdevice;
276 list_add_tail(&ops->list, &link_ops);
280 EXPORT_SYMBOL_GPL(__rtnl_link_register);
283 * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
284 * @ops: struct rtnl_link_ops * to register
286 * Returns 0 on success or a negative error code.
288 int rtnl_link_register(struct rtnl_link_ops *ops)
293 err = __rtnl_link_register(ops);
298 EXPORT_SYMBOL_GPL(rtnl_link_register);
301 * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
302 * @ops: struct rtnl_link_ops * to unregister
304 * The caller must hold the rtnl_mutex.
306 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
308 struct net_device *dev, *n;
312 for_each_netdev_safe(net, dev, n) {
313 if (dev->rtnl_link_ops == ops)
317 list_del(&ops->list);
320 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
323 * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
324 * @ops: struct rtnl_link_ops * to unregister
326 void rtnl_link_unregister(struct rtnl_link_ops *ops)
329 __rtnl_link_unregister(ops);
333 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
335 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
337 const struct rtnl_link_ops *ops;
339 list_for_each_entry(ops, &link_ops, list) {
340 if (!strcmp(ops->kind, kind))
346 static size_t rtnl_link_get_size(const struct net_device *dev)
348 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
354 size = nlmsg_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
355 nlmsg_total_size(strlen(ops->kind) + 1); /* IFLA_INFO_KIND */
358 /* IFLA_INFO_DATA + nested data */
359 size += nlmsg_total_size(sizeof(struct nlattr)) +
362 if (ops->get_xstats_size)
363 size += ops->get_xstats_size(dev); /* IFLA_INFO_XSTATS */
368 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
370 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
371 struct nlattr *linkinfo, *data;
374 linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
375 if (linkinfo == NULL)
378 if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
379 goto err_cancel_link;
380 if (ops->fill_xstats) {
381 err = ops->fill_xstats(skb, dev);
383 goto err_cancel_link;
385 if (ops->fill_info) {
386 data = nla_nest_start(skb, IFLA_INFO_DATA);
388 goto err_cancel_link;
389 err = ops->fill_info(skb, dev);
391 goto err_cancel_data;
392 nla_nest_end(skb, data);
395 nla_nest_end(skb, linkinfo);
399 nla_nest_cancel(skb, data);
401 nla_nest_cancel(skb, linkinfo);
406 static const int rtm_min[RTM_NR_FAMILIES] =
408 [RTM_FAM(RTM_NEWLINK)] = NLMSG_LENGTH(sizeof(struct ifinfomsg)),
409 [RTM_FAM(RTM_NEWADDR)] = NLMSG_LENGTH(sizeof(struct ifaddrmsg)),
410 [RTM_FAM(RTM_NEWROUTE)] = NLMSG_LENGTH(sizeof(struct rtmsg)),
411 [RTM_FAM(RTM_NEWRULE)] = NLMSG_LENGTH(sizeof(struct fib_rule_hdr)),
412 [RTM_FAM(RTM_NEWQDISC)] = NLMSG_LENGTH(sizeof(struct tcmsg)),
413 [RTM_FAM(RTM_NEWTCLASS)] = NLMSG_LENGTH(sizeof(struct tcmsg)),
414 [RTM_FAM(RTM_NEWTFILTER)] = NLMSG_LENGTH(sizeof(struct tcmsg)),
415 [RTM_FAM(RTM_NEWACTION)] = NLMSG_LENGTH(sizeof(struct tcamsg)),
416 [RTM_FAM(RTM_GETMULTICAST)] = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
417 [RTM_FAM(RTM_GETANYCAST)] = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
420 static const int rta_max[RTM_NR_FAMILIES] =
422 [RTM_FAM(RTM_NEWLINK)] = IFLA_MAX,
423 [RTM_FAM(RTM_NEWADDR)] = IFA_MAX,
424 [RTM_FAM(RTM_NEWROUTE)] = RTA_MAX,
425 [RTM_FAM(RTM_NEWRULE)] = FRA_MAX,
426 [RTM_FAM(RTM_NEWQDISC)] = TCA_MAX,
427 [RTM_FAM(RTM_NEWTCLASS)] = TCA_MAX,
428 [RTM_FAM(RTM_NEWTFILTER)] = TCA_MAX,
429 [RTM_FAM(RTM_NEWACTION)] = TCAA_MAX,
432 void __rta_fill(struct sk_buff *skb, int attrtype, int attrlen, const void *data)
435 int size = RTA_LENGTH(attrlen);
437 rta = (struct rtattr*)skb_put(skb, RTA_ALIGN(size));
438 rta->rta_type = attrtype;
440 memcpy(RTA_DATA(rta), data, attrlen);
441 memset(RTA_DATA(rta) + attrlen, 0, RTA_ALIGN(size) - size);
444 size_t rtattr_strlcpy(char *dest, const struct rtattr *rta, size_t size)
446 size_t ret = RTA_PAYLOAD(rta);
447 char *src = RTA_DATA(rta);
449 if (ret > 0 && src[ret - 1] == '\0')
452 size_t len = (ret >= size) ? size - 1 : ret;
453 memset(dest, 0, size);
454 memcpy(dest, src, len);
459 int rtnetlink_send(struct sk_buff *skb, u32 pid, unsigned group, int echo)
463 NETLINK_CB(skb).dst_group = group;
465 atomic_inc(&skb->users);
466 netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
468 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
472 int rtnl_unicast(struct sk_buff *skb, u32 pid)
474 return nlmsg_unicast(rtnl, skb, pid);
477 int rtnl_notify(struct sk_buff *skb, u32 pid, u32 group,
478 struct nlmsghdr *nlh, gfp_t flags)
483 report = nlmsg_report(nlh);
485 return nlmsg_notify(rtnl, skb, pid, group, report, flags);
488 void rtnl_set_sk_err(u32 group, int error)
490 netlink_set_err(rtnl, 0, group, error);
493 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
498 mx = nla_nest_start(skb, RTA_METRICS);
502 for (i = 0; i < RTAX_MAX; i++) {
505 NLA_PUT_U32(skb, i+1, metrics[i]);
510 nla_nest_cancel(skb, mx);
514 return nla_nest_end(skb, mx);
517 return nla_nest_cancel(skb, mx);
520 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
521 u32 ts, u32 tsage, long expires, u32 error)
523 struct rta_cacheinfo ci = {
524 .rta_lastuse = jiffies_to_clock_t(jiffies - dst->lastuse),
525 .rta_used = dst->__use,
526 .rta_clntref = atomic_read(&(dst->__refcnt)),
534 ci.rta_expires = jiffies_to_clock_t(expires);
536 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;
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 void copy_rtnl_link_stats(struct rtnl_link_stats *a,
569 struct net_device_stats *b)
571 a->rx_packets = b->rx_packets;
572 a->tx_packets = b->tx_packets;
573 a->rx_bytes = b->rx_bytes;
574 a->tx_bytes = b->tx_bytes;
575 a->rx_errors = b->rx_errors;
576 a->tx_errors = b->tx_errors;
577 a->rx_dropped = b->rx_dropped;
578 a->tx_dropped = b->tx_dropped;
580 a->multicast = b->multicast;
581 a->collisions = b->collisions;
583 a->rx_length_errors = b->rx_length_errors;
584 a->rx_over_errors = b->rx_over_errors;
585 a->rx_crc_errors = b->rx_crc_errors;
586 a->rx_frame_errors = b->rx_frame_errors;
587 a->rx_fifo_errors = b->rx_fifo_errors;
588 a->rx_missed_errors = b->rx_missed_errors;
590 a->tx_aborted_errors = b->tx_aborted_errors;
591 a->tx_carrier_errors = b->tx_carrier_errors;
592 a->tx_fifo_errors = b->tx_fifo_errors;
593 a->tx_heartbeat_errors = b->tx_heartbeat_errors;
594 a->tx_window_errors = b->tx_window_errors;
596 a->rx_compressed = b->rx_compressed;
597 a->tx_compressed = b->tx_compressed;
600 static inline size_t if_nlmsg_size(const struct net_device *dev)
602 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
603 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
604 + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
605 + nla_total_size(sizeof(struct rtnl_link_ifmap))
606 + nla_total_size(sizeof(struct rtnl_link_stats))
607 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
608 + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
609 + nla_total_size(4) /* IFLA_TXQLEN */
610 + nla_total_size(4) /* IFLA_WEIGHT */
611 + nla_total_size(4) /* IFLA_MTU */
612 + nla_total_size(4) /* IFLA_LINK */
613 + nla_total_size(4) /* IFLA_MASTER */
614 + nla_total_size(1) /* IFLA_OPERSTATE */
615 + nla_total_size(1) /* IFLA_LINKMODE */
616 + rtnl_link_get_size(dev); /* IFLA_LINKINFO */
619 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
620 int type, u32 pid, u32 seq, u32 change,
623 struct ifinfomsg *ifm;
624 struct nlmsghdr *nlh;
626 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
630 ifm = nlmsg_data(nlh);
631 ifm->ifi_family = AF_UNSPEC;
633 ifm->ifi_type = dev->type;
634 ifm->ifi_index = dev->ifindex;
635 ifm->ifi_flags = dev_get_flags(dev);
636 ifm->ifi_change = change;
638 NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
639 NLA_PUT_U32(skb, IFLA_TXQLEN, dev->tx_queue_len);
640 NLA_PUT_U8(skb, IFLA_OPERSTATE,
641 netif_running(dev) ? dev->operstate : IF_OPER_DOWN);
642 NLA_PUT_U8(skb, IFLA_LINKMODE, dev->link_mode);
643 NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
645 if (dev->ifindex != dev->iflink)
646 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
649 NLA_PUT_U32(skb, IFLA_MASTER, dev->master->ifindex);
651 if (dev->qdisc_sleeping)
652 NLA_PUT_STRING(skb, IFLA_QDISC, dev->qdisc_sleeping->ops->id);
655 struct rtnl_link_ifmap map = {
656 .mem_start = dev->mem_start,
657 .mem_end = dev->mem_end,
658 .base_addr = dev->base_addr,
661 .port = dev->if_port,
663 NLA_PUT(skb, IFLA_MAP, sizeof(map), &map);
667 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
668 NLA_PUT(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast);
671 if (dev->get_stats) {
672 struct net_device_stats *stats = dev->get_stats(dev);
676 attr = nla_reserve(skb, IFLA_STATS,
677 sizeof(struct rtnl_link_stats));
679 goto nla_put_failure;
681 copy_rtnl_link_stats(nla_data(attr), stats);
685 if (dev->rtnl_link_ops) {
686 if (rtnl_link_fill(skb, dev) < 0)
687 goto nla_put_failure;
690 return nlmsg_end(skb, nlh);
693 nlmsg_cancel(skb, nlh);
697 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
699 struct net *net = skb->sk->sk_net;
701 int s_idx = cb->args[0];
702 struct net_device *dev;
705 for_each_netdev(net, dev) {
708 if (rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
709 NETLINK_CB(cb->skb).pid,
710 cb->nlh->nlmsg_seq, 0, NLM_F_MULTI) <= 0)
720 const struct nla_policy ifla_policy[IFLA_MAX+1] = {
721 [IFLA_IFNAME] = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
722 [IFLA_ADDRESS] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
723 [IFLA_BROADCAST] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
724 [IFLA_MAP] = { .len = sizeof(struct rtnl_link_ifmap) },
725 [IFLA_MTU] = { .type = NLA_U32 },
726 [IFLA_TXQLEN] = { .type = NLA_U32 },
727 [IFLA_WEIGHT] = { .type = NLA_U32 },
728 [IFLA_OPERSTATE] = { .type = NLA_U8 },
729 [IFLA_LINKMODE] = { .type = NLA_U8 },
732 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
733 [IFLA_INFO_KIND] = { .type = NLA_STRING },
734 [IFLA_INFO_DATA] = { .type = NLA_NESTED },
737 static int do_setlink(struct net_device *dev, struct ifinfomsg *ifm,
738 struct nlattr **tb, char *ifname, int modified)
740 int send_addr_notify = 0;
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, 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 sk_buff *skb;
1189 skb = nlmsg_new(if_nlmsg_size(dev), GFP_KERNEL);
1193 err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0);
1195 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
1196 WARN_ON(err == -EMSGSIZE);
1200 err = rtnl_notify(skb, 0, RTNLGRP_LINK, NULL, GFP_KERNEL);
1203 rtnl_set_sk_err(RTNLGRP_LINK, err);
1206 /* Protected by RTNL sempahore. */
1207 static struct rtattr **rta_buf;
1208 static int rtattr_max;
1210 /* Process one rtnetlink message. */
1212 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
1214 rtnl_doit_func doit;
1221 type = nlh->nlmsg_type;
1227 /* All the messages must have at least 1 byte length */
1228 if (nlh->nlmsg_len < NLMSG_LENGTH(sizeof(struct rtgenmsg)))
1231 family = ((struct rtgenmsg*)NLMSG_DATA(nlh))->rtgen_family;
1232 if (family >= NPROTO)
1233 return -EAFNOSUPPORT;
1238 if (kind != 2 && security_netlink_recv(skb, CAP_NET_ADMIN))
1241 if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
1242 rtnl_dumpit_func dumpit;
1244 dumpit = rtnl_get_dumpit(family, type);
1249 err = netlink_dump_start(rtnl, skb, nlh, dumpit, NULL);
1254 memset(rta_buf, 0, (rtattr_max * sizeof(struct rtattr *)));
1256 min_len = rtm_min[sz_idx];
1257 if (nlh->nlmsg_len < min_len)
1260 if (nlh->nlmsg_len > min_len) {
1261 int attrlen = nlh->nlmsg_len - NLMSG_ALIGN(min_len);
1262 struct rtattr *attr = (void*)nlh + NLMSG_ALIGN(min_len);
1264 while (RTA_OK(attr, attrlen)) {
1265 unsigned flavor = attr->rta_type;
1267 if (flavor > rta_max[sz_idx])
1269 rta_buf[flavor-1] = attr;
1271 attr = RTA_NEXT(attr, attrlen);
1275 doit = rtnl_get_doit(family, type);
1279 return doit(skb, nlh, (void *)&rta_buf[0]);
1282 static void rtnetlink_rcv(struct sock *sk, int len)
1284 unsigned int qlen = 0;
1287 mutex_lock(&rtnl_mutex);
1288 netlink_run_queue(sk, &qlen, &rtnetlink_rcv_msg);
1289 mutex_unlock(&rtnl_mutex);
1295 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
1297 struct net_device *dev = ptr;
1299 if (dev->nd_net != &init_net)
1303 case NETDEV_UNREGISTER:
1304 rtmsg_ifinfo(RTM_DELLINK, dev, ~0U);
1306 case NETDEV_REGISTER:
1307 rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U);
1311 rtmsg_ifinfo(RTM_NEWLINK, dev, IFF_UP|IFF_RUNNING);
1314 case NETDEV_GOING_DOWN:
1317 rtmsg_ifinfo(RTM_NEWLINK, dev, 0);
1323 static struct notifier_block rtnetlink_dev_notifier = {
1324 .notifier_call = rtnetlink_event,
1327 void __init rtnetlink_init(void)
1332 for (i = 0; i < ARRAY_SIZE(rta_max); i++)
1333 if (rta_max[i] > rtattr_max)
1334 rtattr_max = rta_max[i];
1335 rta_buf = kmalloc(rtattr_max * sizeof(struct rtattr *), GFP_KERNEL);
1337 panic("rtnetlink_init: cannot allocate rta_buf\n");
1339 rtnl = netlink_kernel_create(&init_net, NETLINK_ROUTE, RTNLGRP_MAX,
1340 rtnetlink_rcv, &rtnl_mutex, THIS_MODULE);
1342 panic("rtnetlink_init: cannot initialize rtnetlink\n");
1343 netlink_set_nonroot(NETLINK_ROUTE, NL_NONROOT_RECV);
1344 register_netdevice_notifier(&rtnetlink_dev_notifier);
1346 rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink, rtnl_dump_ifinfo);
1347 rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL);
1348 rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL);
1349 rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL);
1351 rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all);
1352 rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all);
1355 EXPORT_SYMBOL(__rta_fill);
1356 EXPORT_SYMBOL(rtattr_strlcpy);
1357 EXPORT_SYMBOL(rtattr_parse);
1358 EXPORT_SYMBOL(__rtattr_parse_nested_compat);
1359 EXPORT_SYMBOL(rtnetlink_put_metrics);
1360 EXPORT_SYMBOL(rtnl_lock);
1361 EXPORT_SYMBOL(rtnl_trylock);
1362 EXPORT_SYMBOL(rtnl_unlock);
1363 EXPORT_SYMBOL(rtnl_unicast);
1364 EXPORT_SYMBOL(rtnl_notify);
1365 EXPORT_SYMBOL(rtnl_set_sk_err);
1366 EXPORT_SYMBOL(rtnl_create_link);
1367 EXPORT_SYMBOL(ifla_policy);