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 static struct rtnl_link *rtnl_msg_handlers[NPROTO];
95 static inline int rtm_msgindex(int msgtype)
97 int msgindex = msgtype - RTM_BASE;
100 * msgindex < 0 implies someone tried to register a netlink
101 * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
102 * the message type has not been added to linux/rtnetlink.h
104 BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
109 static rtnl_doit_func rtnl_get_doit(int protocol, int msgindex)
111 struct rtnl_link *tab;
113 tab = rtnl_msg_handlers[protocol];
114 if (tab == NULL || tab[msgindex].doit == NULL)
115 tab = rtnl_msg_handlers[PF_UNSPEC];
117 return tab ? tab[msgindex].doit : NULL;
120 static rtnl_dumpit_func rtnl_get_dumpit(int protocol, int msgindex)
122 struct rtnl_link *tab;
124 tab = rtnl_msg_handlers[protocol];
125 if (tab == NULL || tab[msgindex].dumpit == NULL)
126 tab = rtnl_msg_handlers[PF_UNSPEC];
128 return tab ? tab[msgindex].dumpit : NULL;
132 * __rtnl_register - Register a rtnetlink message type
133 * @protocol: Protocol family or PF_UNSPEC
134 * @msgtype: rtnetlink message type
135 * @doit: Function pointer called for each request message
136 * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
138 * Registers the specified function pointers (at least one of them has
139 * to be non-NULL) to be called whenever a request message for the
140 * specified protocol family and message type is received.
142 * The special protocol family PF_UNSPEC may be used to define fallback
143 * function pointers for the case when no entry for the specific protocol
146 * Returns 0 on success or a negative error code.
148 int __rtnl_register(int protocol, int msgtype,
149 rtnl_doit_func doit, rtnl_dumpit_func dumpit)
151 struct rtnl_link *tab;
154 BUG_ON(protocol < 0 || protocol >= NPROTO);
155 msgindex = rtm_msgindex(msgtype);
157 tab = rtnl_msg_handlers[protocol];
159 tab = kcalloc(RTM_NR_MSGTYPES, sizeof(*tab), GFP_KERNEL);
163 rtnl_msg_handlers[protocol] = tab;
167 tab[msgindex].doit = doit;
170 tab[msgindex].dumpit = dumpit;
174 EXPORT_SYMBOL_GPL(__rtnl_register);
177 * rtnl_register - Register a rtnetlink message type
179 * Identical to __rtnl_register() but panics on failure. This is useful
180 * as failure of this function is very unlikely, it can only happen due
181 * to lack of memory when allocating the chain to store all message
182 * handlers for a protocol. Meant for use in init functions where lack
183 * of memory implies no sense in continueing.
185 void rtnl_register(int protocol, int msgtype,
186 rtnl_doit_func doit, rtnl_dumpit_func dumpit)
188 if (__rtnl_register(protocol, msgtype, doit, dumpit) < 0)
189 panic("Unable to register rtnetlink message handler, "
190 "protocol = %d, message type = %d\n",
193 EXPORT_SYMBOL_GPL(rtnl_register);
196 * rtnl_unregister - Unregister a rtnetlink message type
197 * @protocol: Protocol family or PF_UNSPEC
198 * @msgtype: rtnetlink message type
200 * Returns 0 on success or a negative error code.
202 int rtnl_unregister(int protocol, int msgtype)
206 BUG_ON(protocol < 0 || protocol >= NPROTO);
207 msgindex = rtm_msgindex(msgtype);
209 if (rtnl_msg_handlers[protocol] == NULL)
212 rtnl_msg_handlers[protocol][msgindex].doit = NULL;
213 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;
233 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
235 static LIST_HEAD(link_ops);
238 * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
239 * @ops: struct rtnl_link_ops * to register
241 * The caller must hold the rtnl_mutex. This function should be used
242 * by drivers that create devices during module initialization. It
243 * must be called before registering the devices.
245 * Returns 0 on success or a negative error code.
247 int __rtnl_link_register(struct rtnl_link_ops *ops)
250 ops->dellink = unregister_netdevice_queue;
252 list_add_tail(&ops->list, &link_ops);
255 EXPORT_SYMBOL_GPL(__rtnl_link_register);
258 * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
259 * @ops: struct rtnl_link_ops * to register
261 * Returns 0 on success or a negative error code.
263 int rtnl_link_register(struct rtnl_link_ops *ops)
268 err = __rtnl_link_register(ops);
272 EXPORT_SYMBOL_GPL(rtnl_link_register);
274 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
276 struct net_device *dev;
277 LIST_HEAD(list_kill);
279 for_each_netdev(net, dev) {
280 if (dev->rtnl_link_ops == ops)
281 ops->dellink(dev, &list_kill);
283 unregister_netdevice_many(&list_kill);
286 void rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
289 __rtnl_kill_links(net, ops);
292 EXPORT_SYMBOL_GPL(rtnl_kill_links);
295 * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
296 * @ops: struct rtnl_link_ops * to unregister
298 * The caller must hold the rtnl_mutex.
300 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
305 __rtnl_kill_links(net, ops);
307 list_del(&ops->list);
309 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
312 * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
313 * @ops: struct rtnl_link_ops * to unregister
315 void rtnl_link_unregister(struct rtnl_link_ops *ops)
318 __rtnl_link_unregister(ops);
321 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
323 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
325 const struct rtnl_link_ops *ops;
327 list_for_each_entry(ops, &link_ops, list) {
328 if (!strcmp(ops->kind, kind))
334 static size_t rtnl_link_get_size(const struct net_device *dev)
336 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
342 size = nlmsg_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
343 nlmsg_total_size(strlen(ops->kind) + 1); /* IFLA_INFO_KIND */
346 /* IFLA_INFO_DATA + nested data */
347 size += nlmsg_total_size(sizeof(struct nlattr)) +
350 if (ops->get_xstats_size)
351 size += ops->get_xstats_size(dev); /* IFLA_INFO_XSTATS */
356 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
358 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
359 struct nlattr *linkinfo, *data;
362 linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
363 if (linkinfo == NULL)
366 if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
367 goto err_cancel_link;
368 if (ops->fill_xstats) {
369 err = ops->fill_xstats(skb, dev);
371 goto err_cancel_link;
373 if (ops->fill_info) {
374 data = nla_nest_start(skb, IFLA_INFO_DATA);
376 goto err_cancel_link;
377 err = ops->fill_info(skb, dev);
379 goto err_cancel_data;
380 nla_nest_end(skb, data);
383 nla_nest_end(skb, linkinfo);
387 nla_nest_cancel(skb, data);
389 nla_nest_cancel(skb, linkinfo);
394 static const int rtm_min[RTM_NR_FAMILIES] =
396 [RTM_FAM(RTM_NEWLINK)] = NLMSG_LENGTH(sizeof(struct ifinfomsg)),
397 [RTM_FAM(RTM_NEWADDR)] = NLMSG_LENGTH(sizeof(struct ifaddrmsg)),
398 [RTM_FAM(RTM_NEWROUTE)] = NLMSG_LENGTH(sizeof(struct rtmsg)),
399 [RTM_FAM(RTM_NEWRULE)] = NLMSG_LENGTH(sizeof(struct fib_rule_hdr)),
400 [RTM_FAM(RTM_NEWQDISC)] = NLMSG_LENGTH(sizeof(struct tcmsg)),
401 [RTM_FAM(RTM_NEWTCLASS)] = NLMSG_LENGTH(sizeof(struct tcmsg)),
402 [RTM_FAM(RTM_NEWTFILTER)] = NLMSG_LENGTH(sizeof(struct tcmsg)),
403 [RTM_FAM(RTM_NEWACTION)] = NLMSG_LENGTH(sizeof(struct tcamsg)),
404 [RTM_FAM(RTM_GETMULTICAST)] = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
405 [RTM_FAM(RTM_GETANYCAST)] = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
408 static const int rta_max[RTM_NR_FAMILIES] =
410 [RTM_FAM(RTM_NEWLINK)] = IFLA_MAX,
411 [RTM_FAM(RTM_NEWADDR)] = IFA_MAX,
412 [RTM_FAM(RTM_NEWROUTE)] = RTA_MAX,
413 [RTM_FAM(RTM_NEWRULE)] = FRA_MAX,
414 [RTM_FAM(RTM_NEWQDISC)] = TCA_MAX,
415 [RTM_FAM(RTM_NEWTCLASS)] = TCA_MAX,
416 [RTM_FAM(RTM_NEWTFILTER)] = TCA_MAX,
417 [RTM_FAM(RTM_NEWACTION)] = TCAA_MAX,
420 void __rta_fill(struct sk_buff *skb, int attrtype, int attrlen, const void *data)
423 int size = RTA_LENGTH(attrlen);
425 rta = (struct rtattr *)skb_put(skb, RTA_ALIGN(size));
426 rta->rta_type = attrtype;
428 memcpy(RTA_DATA(rta), data, attrlen);
429 memset(RTA_DATA(rta) + attrlen, 0, RTA_ALIGN(size) - size);
431 EXPORT_SYMBOL(__rta_fill);
433 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned group, int echo)
435 struct sock *rtnl = net->rtnl;
438 NETLINK_CB(skb).dst_group = group;
440 atomic_inc(&skb->users);
441 netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
443 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
447 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
449 struct sock *rtnl = net->rtnl;
451 return nlmsg_unicast(rtnl, skb, pid);
453 EXPORT_SYMBOL(rtnl_unicast);
455 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
456 struct nlmsghdr *nlh, gfp_t flags)
458 struct sock *rtnl = net->rtnl;
462 report = nlmsg_report(nlh);
464 nlmsg_notify(rtnl, skb, pid, group, report, flags);
466 EXPORT_SYMBOL(rtnl_notify);
468 void rtnl_set_sk_err(struct net *net, u32 group, int error)
470 struct sock *rtnl = net->rtnl;
472 netlink_set_err(rtnl, 0, group, error);
474 EXPORT_SYMBOL(rtnl_set_sk_err);
476 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
481 mx = nla_nest_start(skb, RTA_METRICS);
485 for (i = 0; i < RTAX_MAX; i++) {
488 NLA_PUT_U32(skb, i+1, metrics[i]);
493 nla_nest_cancel(skb, mx);
497 return nla_nest_end(skb, mx);
500 nla_nest_cancel(skb, mx);
503 EXPORT_SYMBOL(rtnetlink_put_metrics);
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);
523 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
525 static void set_operstate(struct net_device *dev, unsigned char transition)
527 unsigned char operstate = dev->operstate;
529 switch (transition) {
531 if ((operstate == IF_OPER_DORMANT ||
532 operstate == IF_OPER_UNKNOWN) &&
534 operstate = IF_OPER_UP;
537 case IF_OPER_DORMANT:
538 if (operstate == IF_OPER_UP ||
539 operstate == IF_OPER_UNKNOWN)
540 operstate = IF_OPER_DORMANT;
544 if (dev->operstate != operstate) {
545 write_lock_bh(&dev_base_lock);
546 dev->operstate = operstate;
547 write_unlock_bh(&dev_base_lock);
548 netdev_state_change(dev);
552 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
553 const struct ifinfomsg *ifm)
555 unsigned int flags = ifm->ifi_flags;
557 /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
559 flags = (flags & ifm->ifi_change) |
560 (dev->flags & ~ifm->ifi_change);
565 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
566 const struct net_device_stats *b)
568 a->rx_packets = b->rx_packets;
569 a->tx_packets = b->tx_packets;
570 a->rx_bytes = b->rx_bytes;
571 a->tx_bytes = b->tx_bytes;
572 a->rx_errors = b->rx_errors;
573 a->tx_errors = b->tx_errors;
574 a->rx_dropped = b->rx_dropped;
575 a->tx_dropped = b->tx_dropped;
577 a->multicast = b->multicast;
578 a->collisions = b->collisions;
580 a->rx_length_errors = b->rx_length_errors;
581 a->rx_over_errors = b->rx_over_errors;
582 a->rx_crc_errors = b->rx_crc_errors;
583 a->rx_frame_errors = b->rx_frame_errors;
584 a->rx_fifo_errors = b->rx_fifo_errors;
585 a->rx_missed_errors = b->rx_missed_errors;
587 a->tx_aborted_errors = b->tx_aborted_errors;
588 a->tx_carrier_errors = b->tx_carrier_errors;
589 a->tx_fifo_errors = b->tx_fifo_errors;
590 a->tx_heartbeat_errors = b->tx_heartbeat_errors;
591 a->tx_window_errors = b->tx_window_errors;
593 a->rx_compressed = b->rx_compressed;
594 a->tx_compressed = b->tx_compressed;
597 static inline int rtnl_vfinfo_size(const struct net_device *dev)
599 if (dev->dev.parent && dev_is_pci(dev->dev.parent))
600 return dev_num_vf(dev->dev.parent) *
601 sizeof(struct ifla_vf_info);
606 static inline size_t if_nlmsg_size(const struct net_device *dev)
608 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
609 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
610 + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
611 + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
612 + nla_total_size(sizeof(struct rtnl_link_ifmap))
613 + nla_total_size(sizeof(struct rtnl_link_stats))
614 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
615 + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
616 + nla_total_size(4) /* IFLA_TXQLEN */
617 + nla_total_size(4) /* IFLA_WEIGHT */
618 + nla_total_size(4) /* IFLA_MTU */
619 + nla_total_size(4) /* IFLA_LINK */
620 + nla_total_size(4) /* IFLA_MASTER */
621 + nla_total_size(1) /* IFLA_OPERSTATE */
622 + nla_total_size(1) /* IFLA_LINKMODE */
623 + nla_total_size(4) /* IFLA_NUM_VF */
624 + nla_total_size(rtnl_vfinfo_size(dev)) /* IFLA_VFINFO */
625 + rtnl_link_get_size(dev); /* IFLA_LINKINFO */
628 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
629 int type, u32 pid, u32 seq, u32 change,
632 struct ifinfomsg *ifm;
633 struct nlmsghdr *nlh;
634 const struct net_device_stats *stats;
637 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
641 ifm = nlmsg_data(nlh);
642 ifm->ifi_family = AF_UNSPEC;
644 ifm->ifi_type = dev->type;
645 ifm->ifi_index = dev->ifindex;
646 ifm->ifi_flags = dev_get_flags(dev);
647 ifm->ifi_change = change;
649 NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
650 NLA_PUT_U32(skb, IFLA_TXQLEN, dev->tx_queue_len);
651 NLA_PUT_U8(skb, IFLA_OPERSTATE,
652 netif_running(dev) ? dev->operstate : IF_OPER_DOWN);
653 NLA_PUT_U8(skb, IFLA_LINKMODE, dev->link_mode);
654 NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
656 if (dev->ifindex != dev->iflink)
657 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
660 NLA_PUT_U32(skb, IFLA_MASTER, dev->master->ifindex);
663 NLA_PUT_STRING(skb, IFLA_QDISC, dev->qdisc->ops->id);
666 NLA_PUT_STRING(skb, IFLA_IFALIAS, dev->ifalias);
669 struct rtnl_link_ifmap map = {
670 .mem_start = dev->mem_start,
671 .mem_end = dev->mem_end,
672 .base_addr = dev->base_addr,
675 .port = dev->if_port,
677 NLA_PUT(skb, IFLA_MAP, sizeof(map), &map);
681 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
682 NLA_PUT(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast);
685 attr = nla_reserve(skb, IFLA_STATS,
686 sizeof(struct rtnl_link_stats));
688 goto nla_put_failure;
690 stats = dev_get_stats(dev);
691 copy_rtnl_link_stats(nla_data(attr), stats);
693 if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent) {
695 struct ifla_vf_info ivi;
697 NLA_PUT_U32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent));
698 for (i = 0; i < dev_num_vf(dev->dev.parent); i++) {
699 if (dev->netdev_ops->ndo_get_vf_config(dev, i, &ivi))
701 NLA_PUT(skb, IFLA_VFINFO, sizeof(ivi), &ivi);
704 if (dev->rtnl_link_ops) {
705 if (rtnl_link_fill(skb, dev) < 0)
706 goto nla_put_failure;
709 return nlmsg_end(skb, nlh);
712 nlmsg_cancel(skb, nlh);
716 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
718 struct net *net = sock_net(skb->sk);
721 struct net_device *dev;
722 struct hlist_head *head;
723 struct hlist_node *node;
728 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
730 head = &net->dev_index_head[h];
731 hlist_for_each_entry(dev, node, head, index_hlist) {
734 if (rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
735 NETLINK_CB(cb->skb).pid,
736 cb->nlh->nlmsg_seq, 0,
750 const struct nla_policy ifla_policy[IFLA_MAX+1] = {
751 [IFLA_IFNAME] = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
752 [IFLA_ADDRESS] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
753 [IFLA_BROADCAST] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
754 [IFLA_MAP] = { .len = sizeof(struct rtnl_link_ifmap) },
755 [IFLA_MTU] = { .type = NLA_U32 },
756 [IFLA_LINK] = { .type = NLA_U32 },
757 [IFLA_TXQLEN] = { .type = NLA_U32 },
758 [IFLA_WEIGHT] = { .type = NLA_U32 },
759 [IFLA_OPERSTATE] = { .type = NLA_U8 },
760 [IFLA_LINKMODE] = { .type = NLA_U8 },
761 [IFLA_LINKINFO] = { .type = NLA_NESTED },
762 [IFLA_NET_NS_PID] = { .type = NLA_U32 },
763 [IFLA_IFALIAS] = { .type = NLA_STRING, .len = IFALIASZ-1 },
764 [IFLA_VF_MAC] = { .type = NLA_BINARY,
765 .len = sizeof(struct ifla_vf_mac) },
766 [IFLA_VF_VLAN] = { .type = NLA_BINARY,
767 .len = sizeof(struct ifla_vf_vlan) },
768 [IFLA_VF_TX_RATE] = { .type = NLA_BINARY,
769 .len = sizeof(struct ifla_vf_tx_rate) },
771 EXPORT_SYMBOL(ifla_policy);
773 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
774 [IFLA_INFO_KIND] = { .type = NLA_STRING },
775 [IFLA_INFO_DATA] = { .type = NLA_NESTED },
778 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
781 /* Examine the link attributes and figure out which
782 * network namespace we are talking about.
784 if (tb[IFLA_NET_NS_PID])
785 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
787 net = get_net(src_net);
790 EXPORT_SYMBOL(rtnl_link_get_net);
792 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
795 if (tb[IFLA_ADDRESS] &&
796 nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
799 if (tb[IFLA_BROADCAST] &&
800 nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
807 static int do_setlink(struct net_device *dev, struct ifinfomsg *ifm,
808 struct nlattr **tb, char *ifname, int modified)
810 const struct net_device_ops *ops = dev->netdev_ops;
811 int send_addr_notify = 0;
814 if (tb[IFLA_NET_NS_PID]) {
815 struct net *net = rtnl_link_get_net(dev_net(dev), tb);
820 err = dev_change_net_namespace(dev, net, ifname);
828 struct rtnl_link_ifmap *u_map;
831 if (!ops->ndo_set_config) {
836 if (!netif_device_present(dev)) {
841 u_map = nla_data(tb[IFLA_MAP]);
842 k_map.mem_start = (unsigned long) u_map->mem_start;
843 k_map.mem_end = (unsigned long) u_map->mem_end;
844 k_map.base_addr = (unsigned short) u_map->base_addr;
845 k_map.irq = (unsigned char) u_map->irq;
846 k_map.dma = (unsigned char) u_map->dma;
847 k_map.port = (unsigned char) u_map->port;
849 err = ops->ndo_set_config(dev, &k_map);
856 if (tb[IFLA_ADDRESS]) {
860 if (!ops->ndo_set_mac_address) {
865 if (!netif_device_present(dev)) {
870 len = sizeof(sa_family_t) + dev->addr_len;
871 sa = kmalloc(len, GFP_KERNEL);
876 sa->sa_family = dev->type;
877 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
879 err = ops->ndo_set_mac_address(dev, sa);
883 send_addr_notify = 1;
888 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
895 * Interface selected by interface index but interface
896 * name provided implies that a name change has been
899 if (ifm->ifi_index > 0 && ifname[0]) {
900 err = dev_change_name(dev, ifname);
906 if (tb[IFLA_IFALIAS]) {
907 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
908 nla_len(tb[IFLA_IFALIAS]));
914 if (tb[IFLA_BROADCAST]) {
915 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
916 send_addr_notify = 1;
919 if (ifm->ifi_flags || ifm->ifi_change) {
920 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
926 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
928 if (tb[IFLA_OPERSTATE])
929 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
931 if (tb[IFLA_LINKMODE]) {
932 write_lock_bh(&dev_base_lock);
933 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
934 write_unlock_bh(&dev_base_lock);
937 if (tb[IFLA_VF_MAC]) {
938 struct ifla_vf_mac *ivm;
939 ivm = nla_data(tb[IFLA_VF_MAC]);
941 if (ops->ndo_set_vf_mac)
942 err = ops->ndo_set_vf_mac(dev, ivm->vf, ivm->mac);
948 if (tb[IFLA_VF_VLAN]) {
949 struct ifla_vf_vlan *ivv;
950 ivv = nla_data(tb[IFLA_VF_VLAN]);
952 if (ops->ndo_set_vf_vlan)
953 err = ops->ndo_set_vf_vlan(dev, ivv->vf,
962 if (tb[IFLA_VF_TX_RATE]) {
963 struct ifla_vf_tx_rate *ivt;
964 ivt = nla_data(tb[IFLA_VF_TX_RATE]);
966 if (ops->ndo_set_vf_tx_rate)
967 err = ops->ndo_set_vf_tx_rate(dev, ivt->vf, ivt->rate);
975 if (err < 0 && modified && net_ratelimit())
976 printk(KERN_WARNING "A link change request failed with "
977 "some changes comitted already. Interface %s may "
978 "have been left with an inconsistent configuration, "
979 "please check.\n", dev->name);
981 if (send_addr_notify)
982 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
986 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
988 struct net *net = sock_net(skb->sk);
989 struct ifinfomsg *ifm;
990 struct net_device *dev;
992 struct nlattr *tb[IFLA_MAX+1];
993 char ifname[IFNAMSIZ];
995 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1000 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1005 ifm = nlmsg_data(nlh);
1006 if (ifm->ifi_index > 0)
1007 dev = __dev_get_by_index(net, ifm->ifi_index);
1008 else if (tb[IFLA_IFNAME])
1009 dev = __dev_get_by_name(net, ifname);
1018 err = validate_linkmsg(dev, tb);
1022 err = do_setlink(dev, ifm, tb, ifname, 0);
1027 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1029 struct net *net = sock_net(skb->sk);
1030 const struct rtnl_link_ops *ops;
1031 struct net_device *dev;
1032 struct ifinfomsg *ifm;
1033 char ifname[IFNAMSIZ];
1034 struct nlattr *tb[IFLA_MAX+1];
1037 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1041 if (tb[IFLA_IFNAME])
1042 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1044 ifm = nlmsg_data(nlh);
1045 if (ifm->ifi_index > 0)
1046 dev = __dev_get_by_index(net, ifm->ifi_index);
1047 else if (tb[IFLA_IFNAME])
1048 dev = __dev_get_by_name(net, ifname);
1055 ops = dev->rtnl_link_ops;
1059 ops->dellink(dev, NULL);
1063 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
1065 unsigned int old_flags;
1068 old_flags = dev->flags;
1069 if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
1070 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1075 dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
1076 rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U);
1078 __dev_notify_flags(dev, old_flags);
1081 EXPORT_SYMBOL(rtnl_configure_link);
1083 struct net_device *rtnl_create_link(struct net *src_net, struct net *net,
1084 char *ifname, const struct rtnl_link_ops *ops, struct nlattr *tb[])
1087 struct net_device *dev;
1088 unsigned int num_queues = 1;
1089 unsigned int real_num_queues = 1;
1091 if (ops->get_tx_queues) {
1092 err = ops->get_tx_queues(src_net, tb, &num_queues,
1098 dev = alloc_netdev_mq(ops->priv_size, ifname, ops->setup, num_queues);
1102 dev_net_set(dev, net);
1103 dev->rtnl_link_ops = ops;
1104 dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
1105 dev->real_num_tx_queues = real_num_queues;
1107 if (strchr(dev->name, '%')) {
1108 err = dev_alloc_name(dev, dev->name);
1114 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
1115 if (tb[IFLA_ADDRESS])
1116 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
1117 nla_len(tb[IFLA_ADDRESS]));
1118 if (tb[IFLA_BROADCAST])
1119 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
1120 nla_len(tb[IFLA_BROADCAST]));
1121 if (tb[IFLA_TXQLEN])
1122 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1123 if (tb[IFLA_OPERSTATE])
1124 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1125 if (tb[IFLA_LINKMODE])
1126 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1133 return ERR_PTR(err);
1135 EXPORT_SYMBOL(rtnl_create_link);
1137 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1139 struct net *net = sock_net(skb->sk);
1140 const struct rtnl_link_ops *ops;
1141 struct net_device *dev;
1142 struct ifinfomsg *ifm;
1143 char kind[MODULE_NAME_LEN];
1144 char ifname[IFNAMSIZ];
1145 struct nlattr *tb[IFLA_MAX+1];
1146 struct nlattr *linkinfo[IFLA_INFO_MAX+1];
1149 #ifdef CONFIG_MODULES
1152 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1156 if (tb[IFLA_IFNAME])
1157 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1161 ifm = nlmsg_data(nlh);
1162 if (ifm->ifi_index > 0)
1163 dev = __dev_get_by_index(net, ifm->ifi_index);
1165 dev = __dev_get_by_name(net, ifname);
1169 err = validate_linkmsg(dev, tb);
1173 if (tb[IFLA_LINKINFO]) {
1174 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
1175 tb[IFLA_LINKINFO], ifla_info_policy);
1179 memset(linkinfo, 0, sizeof(linkinfo));
1181 if (linkinfo[IFLA_INFO_KIND]) {
1182 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
1183 ops = rtnl_link_ops_get(kind);
1190 struct nlattr *attr[ops ? ops->maxtype + 1 : 0], **data = NULL;
1191 struct net *dest_net;
1194 if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
1195 err = nla_parse_nested(attr, ops->maxtype,
1196 linkinfo[IFLA_INFO_DATA],
1202 if (ops->validate) {
1203 err = ops->validate(tb, data);
1212 if (nlh->nlmsg_flags & NLM_F_EXCL)
1214 if (nlh->nlmsg_flags & NLM_F_REPLACE)
1217 if (linkinfo[IFLA_INFO_DATA]) {
1218 if (!ops || ops != dev->rtnl_link_ops ||
1222 err = ops->changelink(dev, tb, data);
1228 return do_setlink(dev, ifm, tb, ifname, modified);
1231 if (!(nlh->nlmsg_flags & NLM_F_CREATE))
1236 if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
1240 #ifdef CONFIG_MODULES
1243 request_module("rtnl-link-%s", kind);
1245 ops = rtnl_link_ops_get(kind);
1254 snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
1256 dest_net = rtnl_link_get_net(net, tb);
1257 dev = rtnl_create_link(net, dest_net, ifname, ops, tb);
1261 else if (ops->newlink)
1262 err = ops->newlink(net, dev, tb, data);
1264 err = register_netdevice(dev);
1265 if (err < 0 && !IS_ERR(dev)) {
1270 err = rtnl_configure_link(dev, ifm);
1272 unregister_netdevice(dev);
1279 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
1281 struct net *net = sock_net(skb->sk);
1282 struct ifinfomsg *ifm;
1283 char ifname[IFNAMSIZ];
1284 struct nlattr *tb[IFLA_MAX+1];
1285 struct net_device *dev = NULL;
1286 struct sk_buff *nskb;
1289 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1293 if (tb[IFLA_IFNAME])
1294 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1296 ifm = nlmsg_data(nlh);
1297 if (ifm->ifi_index > 0)
1298 dev = __dev_get_by_index(net, ifm->ifi_index);
1299 else if (tb[IFLA_IFNAME])
1300 dev = __dev_get_by_name(net, ifname);
1307 nskb = nlmsg_new(if_nlmsg_size(dev), GFP_KERNEL);
1311 err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).pid,
1312 nlh->nlmsg_seq, 0, 0);
1314 /* -EMSGSIZE implies BUG in if_nlmsg_size */
1315 WARN_ON(err == -EMSGSIZE);
1318 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).pid);
1323 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
1326 int s_idx = cb->family;
1330 for (idx = 1; idx < NPROTO; idx++) {
1331 int type = cb->nlh->nlmsg_type-RTM_BASE;
1332 if (idx < s_idx || idx == PF_PACKET)
1334 if (rtnl_msg_handlers[idx] == NULL ||
1335 rtnl_msg_handlers[idx][type].dumpit == NULL)
1338 memset(&cb->args[0], 0, sizeof(cb->args));
1339 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
1347 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned change)
1349 struct net *net = dev_net(dev);
1350 struct sk_buff *skb;
1353 skb = nlmsg_new(if_nlmsg_size(dev), GFP_KERNEL);
1357 err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0);
1359 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
1360 WARN_ON(err == -EMSGSIZE);
1364 rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_KERNEL);
1368 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
1371 /* Protected by RTNL sempahore. */
1372 static struct rtattr **rta_buf;
1373 static int rtattr_max;
1375 /* Process one rtnetlink message. */
1377 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
1379 struct net *net = sock_net(skb->sk);
1380 rtnl_doit_func doit;
1387 type = nlh->nlmsg_type;
1393 /* All the messages must have at least 1 byte length */
1394 if (nlh->nlmsg_len < NLMSG_LENGTH(sizeof(struct rtgenmsg)))
1397 family = ((struct rtgenmsg *)NLMSG_DATA(nlh))->rtgen_family;
1398 if (family >= NPROTO)
1399 return -EAFNOSUPPORT;
1404 if (kind != 2 && security_netlink_recv(skb, CAP_NET_ADMIN))
1407 if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
1409 rtnl_dumpit_func dumpit;
1411 dumpit = rtnl_get_dumpit(family, type);
1417 err = netlink_dump_start(rtnl, skb, nlh, dumpit, NULL);
1422 memset(rta_buf, 0, (rtattr_max * sizeof(struct rtattr *)));
1424 min_len = rtm_min[sz_idx];
1425 if (nlh->nlmsg_len < min_len)
1428 if (nlh->nlmsg_len > min_len) {
1429 int attrlen = nlh->nlmsg_len - NLMSG_ALIGN(min_len);
1430 struct rtattr *attr = (void *)nlh + NLMSG_ALIGN(min_len);
1432 while (RTA_OK(attr, attrlen)) {
1433 unsigned flavor = attr->rta_type;
1435 if (flavor > rta_max[sz_idx])
1437 rta_buf[flavor-1] = attr;
1439 attr = RTA_NEXT(attr, attrlen);
1443 doit = rtnl_get_doit(family, type);
1447 return doit(skb, nlh, (void *)&rta_buf[0]);
1450 static void rtnetlink_rcv(struct sk_buff *skb)
1453 netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
1457 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
1459 struct net_device *dev = ptr;
1465 case NETDEV_POST_INIT:
1466 case NETDEV_REGISTER:
1468 case NETDEV_GOING_DOWN:
1469 case NETDEV_UNREGISTER:
1470 case NETDEV_UNREGISTER_BATCH:
1473 rtmsg_ifinfo(RTM_NEWLINK, dev, 0);
1479 static struct notifier_block rtnetlink_dev_notifier = {
1480 .notifier_call = rtnetlink_event,
1484 static int __net_init rtnetlink_net_init(struct net *net)
1487 sk = netlink_kernel_create(net, NETLINK_ROUTE, RTNLGRP_MAX,
1488 rtnetlink_rcv, &rtnl_mutex, THIS_MODULE);
1495 static void __net_exit rtnetlink_net_exit(struct net *net)
1497 netlink_kernel_release(net->rtnl);
1501 static struct pernet_operations rtnetlink_net_ops = {
1502 .init = rtnetlink_net_init,
1503 .exit = rtnetlink_net_exit,
1506 void __init rtnetlink_init(void)
1511 for (i = 0; i < ARRAY_SIZE(rta_max); i++)
1512 if (rta_max[i] > rtattr_max)
1513 rtattr_max = rta_max[i];
1514 rta_buf = kmalloc(rtattr_max * sizeof(struct rtattr *), GFP_KERNEL);
1516 panic("rtnetlink_init: cannot allocate rta_buf\n");
1518 if (register_pernet_subsys(&rtnetlink_net_ops))
1519 panic("rtnetlink_init: cannot initialize rtnetlink\n");
1521 netlink_set_nonroot(NETLINK_ROUTE, NL_NONROOT_RECV);
1522 register_netdevice_notifier(&rtnetlink_dev_notifier);
1524 rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink, rtnl_dump_ifinfo);
1525 rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL);
1526 rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL);
1527 rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL);
1529 rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all);
1530 rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all);