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;
59 rtnl_calcit_func calcit;
62 static DEFINE_MUTEX(rtnl_mutex);
66 mutex_lock(&rtnl_mutex);
68 EXPORT_SYMBOL(rtnl_lock);
70 void __rtnl_unlock(void)
72 mutex_unlock(&rtnl_mutex);
75 void rtnl_unlock(void)
77 /* This fellow will unlock it for us. */
80 EXPORT_SYMBOL(rtnl_unlock);
82 int rtnl_trylock(void)
84 return mutex_trylock(&rtnl_mutex);
86 EXPORT_SYMBOL(rtnl_trylock);
88 int rtnl_is_locked(void)
90 return mutex_is_locked(&rtnl_mutex);
92 EXPORT_SYMBOL(rtnl_is_locked);
94 #ifdef CONFIG_PROVE_LOCKING
95 int lockdep_rtnl_is_held(void)
97 return lockdep_is_held(&rtnl_mutex);
99 EXPORT_SYMBOL(lockdep_rtnl_is_held);
100 #endif /* #ifdef CONFIG_PROVE_LOCKING */
102 static struct rtnl_link *rtnl_msg_handlers[RTNL_FAMILY_MAX + 1];
104 static inline int rtm_msgindex(int msgtype)
106 int msgindex = msgtype - RTM_BASE;
109 * msgindex < 0 implies someone tried to register a netlink
110 * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
111 * the message type has not been added to linux/rtnetlink.h
113 BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
118 static rtnl_doit_func rtnl_get_doit(int protocol, int msgindex)
120 struct rtnl_link *tab;
122 if (protocol <= RTNL_FAMILY_MAX)
123 tab = rtnl_msg_handlers[protocol];
127 if (tab == NULL || tab[msgindex].doit == NULL)
128 tab = rtnl_msg_handlers[PF_UNSPEC];
130 return tab ? tab[msgindex].doit : NULL;
133 static rtnl_dumpit_func rtnl_get_dumpit(int protocol, int msgindex)
135 struct rtnl_link *tab;
137 if (protocol <= RTNL_FAMILY_MAX)
138 tab = rtnl_msg_handlers[protocol];
142 if (tab == NULL || tab[msgindex].dumpit == NULL)
143 tab = rtnl_msg_handlers[PF_UNSPEC];
145 return tab ? tab[msgindex].dumpit : NULL;
148 static rtnl_calcit_func rtnl_get_calcit(int protocol, int msgindex)
150 struct rtnl_link *tab;
152 if (protocol <= RTNL_FAMILY_MAX)
153 tab = rtnl_msg_handlers[protocol];
157 if (tab == NULL || tab[msgindex].calcit == NULL)
158 tab = rtnl_msg_handlers[PF_UNSPEC];
160 return tab ? tab[msgindex].calcit : NULL;
164 * __rtnl_register - Register a rtnetlink message type
165 * @protocol: Protocol family or PF_UNSPEC
166 * @msgtype: rtnetlink message type
167 * @doit: Function pointer called for each request message
168 * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
169 * @calcit: Function pointer to calc size of dump message
171 * Registers the specified function pointers (at least one of them has
172 * to be non-NULL) to be called whenever a request message for the
173 * specified protocol family and message type is received.
175 * The special protocol family PF_UNSPEC may be used to define fallback
176 * function pointers for the case when no entry for the specific protocol
179 * Returns 0 on success or a negative error code.
181 int __rtnl_register(int protocol, int msgtype,
182 rtnl_doit_func doit, rtnl_dumpit_func dumpit,
183 rtnl_calcit_func calcit)
185 struct rtnl_link *tab;
188 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
189 msgindex = rtm_msgindex(msgtype);
191 tab = rtnl_msg_handlers[protocol];
193 tab = kcalloc(RTM_NR_MSGTYPES, sizeof(*tab), GFP_KERNEL);
197 rtnl_msg_handlers[protocol] = tab;
201 tab[msgindex].doit = doit;
204 tab[msgindex].dumpit = dumpit;
207 tab[msgindex].calcit = calcit;
211 EXPORT_SYMBOL_GPL(__rtnl_register);
214 * rtnl_register - Register a rtnetlink message type
216 * Identical to __rtnl_register() but panics on failure. This is useful
217 * as failure of this function is very unlikely, it can only happen due
218 * to lack of memory when allocating the chain to store all message
219 * handlers for a protocol. Meant for use in init functions where lack
220 * of memory implies no sense in continuing.
222 void rtnl_register(int protocol, int msgtype,
223 rtnl_doit_func doit, rtnl_dumpit_func dumpit,
224 rtnl_calcit_func calcit)
226 if (__rtnl_register(protocol, msgtype, doit, dumpit, calcit) < 0)
227 panic("Unable to register rtnetlink message handler, "
228 "protocol = %d, message type = %d\n",
231 EXPORT_SYMBOL_GPL(rtnl_register);
234 * rtnl_unregister - Unregister a rtnetlink message type
235 * @protocol: Protocol family or PF_UNSPEC
236 * @msgtype: rtnetlink message type
238 * Returns 0 on success or a negative error code.
240 int rtnl_unregister(int protocol, int msgtype)
244 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
245 msgindex = rtm_msgindex(msgtype);
247 if (rtnl_msg_handlers[protocol] == NULL)
250 rtnl_msg_handlers[protocol][msgindex].doit = NULL;
251 rtnl_msg_handlers[protocol][msgindex].dumpit = NULL;
255 EXPORT_SYMBOL_GPL(rtnl_unregister);
258 * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
259 * @protocol : Protocol family or PF_UNSPEC
261 * Identical to calling rtnl_unregster() for all registered message types
262 * of a certain protocol family.
264 void rtnl_unregister_all(int protocol)
266 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
268 kfree(rtnl_msg_handlers[protocol]);
269 rtnl_msg_handlers[protocol] = NULL;
271 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
273 static LIST_HEAD(link_ops);
275 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
277 const struct rtnl_link_ops *ops;
279 list_for_each_entry(ops, &link_ops, list) {
280 if (!strcmp(ops->kind, kind))
287 * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
288 * @ops: struct rtnl_link_ops * to register
290 * The caller must hold the rtnl_mutex. This function should be used
291 * by drivers that create devices during module initialization. It
292 * must be called before registering the devices.
294 * Returns 0 on success or a negative error code.
296 int __rtnl_link_register(struct rtnl_link_ops *ops)
298 if (rtnl_link_ops_get(ops->kind))
302 ops->dellink = unregister_netdevice_queue;
304 list_add_tail(&ops->list, &link_ops);
307 EXPORT_SYMBOL_GPL(__rtnl_link_register);
310 * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
311 * @ops: struct rtnl_link_ops * to register
313 * Returns 0 on success or a negative error code.
315 int rtnl_link_register(struct rtnl_link_ops *ops)
320 err = __rtnl_link_register(ops);
324 EXPORT_SYMBOL_GPL(rtnl_link_register);
326 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
328 struct net_device *dev;
329 LIST_HEAD(list_kill);
331 for_each_netdev(net, dev) {
332 if (dev->rtnl_link_ops == ops)
333 ops->dellink(dev, &list_kill);
335 unregister_netdevice_many(&list_kill);
339 * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
340 * @ops: struct rtnl_link_ops * to unregister
342 * The caller must hold the rtnl_mutex.
344 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
349 __rtnl_kill_links(net, ops);
351 list_del(&ops->list);
353 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
356 * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
357 * @ops: struct rtnl_link_ops * to unregister
359 void rtnl_link_unregister(struct rtnl_link_ops *ops)
362 __rtnl_link_unregister(ops);
365 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
367 static size_t rtnl_link_get_size(const struct net_device *dev)
369 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
375 size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
376 nla_total_size(strlen(ops->kind) + 1); /* IFLA_INFO_KIND */
379 /* IFLA_INFO_DATA + nested data */
380 size += nla_total_size(sizeof(struct nlattr)) +
383 if (ops->get_xstats_size)
384 /* IFLA_INFO_XSTATS */
385 size += nla_total_size(ops->get_xstats_size(dev));
390 static LIST_HEAD(rtnl_af_ops);
392 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
394 const struct rtnl_af_ops *ops;
396 list_for_each_entry(ops, &rtnl_af_ops, list) {
397 if (ops->family == family)
405 * __rtnl_af_register - Register rtnl_af_ops with rtnetlink.
406 * @ops: struct rtnl_af_ops * to register
408 * The caller must hold the rtnl_mutex.
410 * Returns 0 on success or a negative error code.
412 int __rtnl_af_register(struct rtnl_af_ops *ops)
414 list_add_tail(&ops->list, &rtnl_af_ops);
417 EXPORT_SYMBOL_GPL(__rtnl_af_register);
420 * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
421 * @ops: struct rtnl_af_ops * to register
423 * Returns 0 on success or a negative error code.
425 int rtnl_af_register(struct rtnl_af_ops *ops)
430 err = __rtnl_af_register(ops);
434 EXPORT_SYMBOL_GPL(rtnl_af_register);
437 * __rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
438 * @ops: struct rtnl_af_ops * to unregister
440 * The caller must hold the rtnl_mutex.
442 void __rtnl_af_unregister(struct rtnl_af_ops *ops)
444 list_del(&ops->list);
446 EXPORT_SYMBOL_GPL(__rtnl_af_unregister);
449 * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
450 * @ops: struct rtnl_af_ops * to unregister
452 void rtnl_af_unregister(struct rtnl_af_ops *ops)
455 __rtnl_af_unregister(ops);
458 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
460 static size_t rtnl_link_get_af_size(const struct net_device *dev)
462 struct rtnl_af_ops *af_ops;
466 size = nla_total_size(sizeof(struct nlattr));
468 list_for_each_entry(af_ops, &rtnl_af_ops, list) {
469 if (af_ops->get_link_af_size) {
470 /* AF_* + nested data */
471 size += nla_total_size(sizeof(struct nlattr)) +
472 af_ops->get_link_af_size(dev);
479 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
481 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
482 struct nlattr *linkinfo, *data;
485 linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
486 if (linkinfo == NULL)
489 if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
490 goto err_cancel_link;
491 if (ops->fill_xstats) {
492 err = ops->fill_xstats(skb, dev);
494 goto err_cancel_link;
496 if (ops->fill_info) {
497 data = nla_nest_start(skb, IFLA_INFO_DATA);
499 goto err_cancel_link;
500 err = ops->fill_info(skb, dev);
502 goto err_cancel_data;
503 nla_nest_end(skb, data);
506 nla_nest_end(skb, linkinfo);
510 nla_nest_cancel(skb, data);
512 nla_nest_cancel(skb, linkinfo);
517 static const int rtm_min[RTM_NR_FAMILIES] =
519 [RTM_FAM(RTM_NEWLINK)] = NLMSG_LENGTH(sizeof(struct ifinfomsg)),
520 [RTM_FAM(RTM_NEWADDR)] = NLMSG_LENGTH(sizeof(struct ifaddrmsg)),
521 [RTM_FAM(RTM_NEWROUTE)] = NLMSG_LENGTH(sizeof(struct rtmsg)),
522 [RTM_FAM(RTM_NEWRULE)] = NLMSG_LENGTH(sizeof(struct fib_rule_hdr)),
523 [RTM_FAM(RTM_NEWQDISC)] = NLMSG_LENGTH(sizeof(struct tcmsg)),
524 [RTM_FAM(RTM_NEWTCLASS)] = NLMSG_LENGTH(sizeof(struct tcmsg)),
525 [RTM_FAM(RTM_NEWTFILTER)] = NLMSG_LENGTH(sizeof(struct tcmsg)),
526 [RTM_FAM(RTM_NEWACTION)] = NLMSG_LENGTH(sizeof(struct tcamsg)),
527 [RTM_FAM(RTM_GETMULTICAST)] = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
528 [RTM_FAM(RTM_GETANYCAST)] = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
531 static const int rta_max[RTM_NR_FAMILIES] =
533 [RTM_FAM(RTM_NEWLINK)] = IFLA_MAX,
534 [RTM_FAM(RTM_NEWADDR)] = IFA_MAX,
535 [RTM_FAM(RTM_NEWROUTE)] = RTA_MAX,
536 [RTM_FAM(RTM_NEWRULE)] = FRA_MAX,
537 [RTM_FAM(RTM_NEWQDISC)] = TCA_MAX,
538 [RTM_FAM(RTM_NEWTCLASS)] = TCA_MAX,
539 [RTM_FAM(RTM_NEWTFILTER)] = TCA_MAX,
540 [RTM_FAM(RTM_NEWACTION)] = TCAA_MAX,
543 void __rta_fill(struct sk_buff *skb, int attrtype, int attrlen, const void *data)
546 int size = RTA_LENGTH(attrlen);
548 rta = (struct rtattr *)skb_put(skb, RTA_ALIGN(size));
549 rta->rta_type = attrtype;
551 memcpy(RTA_DATA(rta), data, attrlen);
552 memset(RTA_DATA(rta) + attrlen, 0, RTA_ALIGN(size) - size);
554 EXPORT_SYMBOL(__rta_fill);
556 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned group, int echo)
558 struct sock *rtnl = net->rtnl;
561 NETLINK_CB(skb).dst_group = group;
563 atomic_inc(&skb->users);
564 netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
566 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
570 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
572 struct sock *rtnl = net->rtnl;
574 return nlmsg_unicast(rtnl, skb, pid);
576 EXPORT_SYMBOL(rtnl_unicast);
578 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
579 struct nlmsghdr *nlh, gfp_t flags)
581 struct sock *rtnl = net->rtnl;
585 report = nlmsg_report(nlh);
587 nlmsg_notify(rtnl, skb, pid, group, report, flags);
589 EXPORT_SYMBOL(rtnl_notify);
591 void rtnl_set_sk_err(struct net *net, u32 group, int error)
593 struct sock *rtnl = net->rtnl;
595 netlink_set_err(rtnl, 0, group, error);
597 EXPORT_SYMBOL(rtnl_set_sk_err);
599 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
604 mx = nla_nest_start(skb, RTA_METRICS);
608 for (i = 0; i < RTAX_MAX; i++) {
611 if (nla_put_u32(skb, i+1, metrics[i]))
612 goto nla_put_failure;
617 nla_nest_cancel(skb, mx);
621 return nla_nest_end(skb, mx);
624 nla_nest_cancel(skb, mx);
627 EXPORT_SYMBOL(rtnetlink_put_metrics);
629 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
630 u32 ts, u32 tsage, long expires, u32 error)
632 struct rta_cacheinfo ci = {
633 .rta_lastuse = jiffies_to_clock_t(jiffies - dst->lastuse),
634 .rta_used = dst->__use,
635 .rta_clntref = atomic_read(&(dst->__refcnt)),
643 ci.rta_expires = jiffies_to_clock_t(expires);
645 return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
647 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
649 static void set_operstate(struct net_device *dev, unsigned char transition)
651 unsigned char operstate = dev->operstate;
653 switch (transition) {
655 if ((operstate == IF_OPER_DORMANT ||
656 operstate == IF_OPER_UNKNOWN) &&
658 operstate = IF_OPER_UP;
661 case IF_OPER_DORMANT:
662 if (operstate == IF_OPER_UP ||
663 operstate == IF_OPER_UNKNOWN)
664 operstate = IF_OPER_DORMANT;
668 if (dev->operstate != operstate) {
669 write_lock_bh(&dev_base_lock);
670 dev->operstate = operstate;
671 write_unlock_bh(&dev_base_lock);
672 netdev_state_change(dev);
676 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
677 const struct ifinfomsg *ifm)
679 unsigned int flags = ifm->ifi_flags;
681 /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
683 flags = (flags & ifm->ifi_change) |
684 (dev->flags & ~ifm->ifi_change);
689 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
690 const struct rtnl_link_stats64 *b)
692 a->rx_packets = b->rx_packets;
693 a->tx_packets = b->tx_packets;
694 a->rx_bytes = b->rx_bytes;
695 a->tx_bytes = b->tx_bytes;
696 a->rx_errors = b->rx_errors;
697 a->tx_errors = b->tx_errors;
698 a->rx_dropped = b->rx_dropped;
699 a->tx_dropped = b->tx_dropped;
701 a->multicast = b->multicast;
702 a->collisions = b->collisions;
704 a->rx_length_errors = b->rx_length_errors;
705 a->rx_over_errors = b->rx_over_errors;
706 a->rx_crc_errors = b->rx_crc_errors;
707 a->rx_frame_errors = b->rx_frame_errors;
708 a->rx_fifo_errors = b->rx_fifo_errors;
709 a->rx_missed_errors = b->rx_missed_errors;
711 a->tx_aborted_errors = b->tx_aborted_errors;
712 a->tx_carrier_errors = b->tx_carrier_errors;
713 a->tx_fifo_errors = b->tx_fifo_errors;
714 a->tx_heartbeat_errors = b->tx_heartbeat_errors;
715 a->tx_window_errors = b->tx_window_errors;
717 a->rx_compressed = b->rx_compressed;
718 a->tx_compressed = b->tx_compressed;
721 static void copy_rtnl_link_stats64(void *v, const struct rtnl_link_stats64 *b)
723 memcpy(v, b, sizeof(*b));
727 static inline int rtnl_vfinfo_size(const struct net_device *dev,
730 if (dev->dev.parent && dev_is_pci(dev->dev.parent) &&
731 (ext_filter_mask & RTEXT_FILTER_VF)) {
732 int num_vfs = dev_num_vf(dev->dev.parent);
733 size_t size = nla_total_size(sizeof(struct nlattr));
734 size += nla_total_size(num_vfs * sizeof(struct nlattr));
736 (nla_total_size(sizeof(struct ifla_vf_mac)) +
737 nla_total_size(sizeof(struct ifla_vf_vlan)) +
738 nla_total_size(sizeof(struct ifla_vf_tx_rate)) +
739 nla_total_size(sizeof(struct ifla_vf_spoofchk)));
745 static size_t rtnl_port_size(const struct net_device *dev)
747 size_t port_size = nla_total_size(4) /* PORT_VF */
748 + nla_total_size(PORT_PROFILE_MAX) /* PORT_PROFILE */
749 + nla_total_size(sizeof(struct ifla_port_vsi))
751 + nla_total_size(PORT_UUID_MAX) /* PORT_INSTANCE_UUID */
752 + nla_total_size(PORT_UUID_MAX) /* PORT_HOST_UUID */
753 + nla_total_size(1) /* PROT_VDP_REQUEST */
754 + nla_total_size(2); /* PORT_VDP_RESPONSE */
755 size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
756 size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
758 size_t port_self_size = nla_total_size(sizeof(struct nlattr))
761 if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent)
763 if (dev_num_vf(dev->dev.parent))
764 return port_self_size + vf_ports_size +
765 vf_port_size * dev_num_vf(dev->dev.parent);
767 return port_self_size;
770 static noinline size_t if_nlmsg_size(const struct net_device *dev,
773 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
774 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
775 + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
776 + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
777 + nla_total_size(sizeof(struct rtnl_link_ifmap))
778 + nla_total_size(sizeof(struct rtnl_link_stats))
779 + nla_total_size(sizeof(struct rtnl_link_stats64))
780 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
781 + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
782 + nla_total_size(4) /* IFLA_TXQLEN */
783 + nla_total_size(4) /* IFLA_WEIGHT */
784 + nla_total_size(4) /* IFLA_MTU */
785 + nla_total_size(4) /* IFLA_LINK */
786 + nla_total_size(4) /* IFLA_MASTER */
787 + nla_total_size(1) /* IFLA_OPERSTATE */
788 + nla_total_size(1) /* IFLA_LINKMODE */
789 + nla_total_size(ext_filter_mask
790 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
791 + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
792 + rtnl_port_size(dev) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
793 + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
794 + rtnl_link_get_af_size(dev); /* IFLA_AF_SPEC */
797 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
799 struct nlattr *vf_ports;
800 struct nlattr *vf_port;
804 vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
808 for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
809 vf_port = nla_nest_start(skb, IFLA_VF_PORT);
811 goto nla_put_failure;
812 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
813 goto nla_put_failure;
814 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
815 if (err == -EMSGSIZE)
816 goto nla_put_failure;
818 nla_nest_cancel(skb, vf_port);
821 nla_nest_end(skb, vf_port);
824 nla_nest_end(skb, vf_ports);
829 nla_nest_cancel(skb, vf_ports);
833 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
835 struct nlattr *port_self;
838 port_self = nla_nest_start(skb, IFLA_PORT_SELF);
842 err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
844 nla_nest_cancel(skb, port_self);
845 return (err == -EMSGSIZE) ? err : 0;
848 nla_nest_end(skb, port_self);
853 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev)
857 if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent)
860 err = rtnl_port_self_fill(skb, dev);
864 if (dev_num_vf(dev->dev.parent)) {
865 err = rtnl_vf_ports_fill(skb, dev);
873 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
874 int type, u32 pid, u32 seq, u32 change,
875 unsigned int flags, u32 ext_filter_mask)
877 struct ifinfomsg *ifm;
878 struct nlmsghdr *nlh;
879 struct rtnl_link_stats64 temp;
880 const struct rtnl_link_stats64 *stats;
881 struct nlattr *attr, *af_spec;
882 struct rtnl_af_ops *af_ops;
885 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
889 ifm = nlmsg_data(nlh);
890 ifm->ifi_family = AF_UNSPEC;
892 ifm->ifi_type = dev->type;
893 ifm->ifi_index = dev->ifindex;
894 ifm->ifi_flags = dev_get_flags(dev);
895 ifm->ifi_change = change;
897 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
898 nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
899 nla_put_u8(skb, IFLA_OPERSTATE,
900 netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
901 nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
902 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
903 nla_put_u32(skb, IFLA_GROUP, dev->group) ||
904 (dev->ifindex != dev->iflink &&
905 nla_put_u32(skb, IFLA_LINK, dev->iflink)) ||
907 nla_put_u32(skb, IFLA_MASTER, dev->master->ifindex)) ||
909 nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
911 nla_put_string(skb, IFLA_IFALIAS, dev->ifalias)))
912 goto nla_put_failure;
915 struct rtnl_link_ifmap map = {
916 .mem_start = dev->mem_start,
917 .mem_end = dev->mem_end,
918 .base_addr = dev->base_addr,
921 .port = dev->if_port,
923 if (nla_put(skb, IFLA_MAP, sizeof(map), &map))
924 goto nla_put_failure;
928 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
929 nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
930 goto nla_put_failure;
933 attr = nla_reserve(skb, IFLA_STATS,
934 sizeof(struct rtnl_link_stats));
936 goto nla_put_failure;
938 stats = dev_get_stats(dev, &temp);
939 copy_rtnl_link_stats(nla_data(attr), stats);
941 attr = nla_reserve(skb, IFLA_STATS64,
942 sizeof(struct rtnl_link_stats64));
944 goto nla_put_failure;
945 copy_rtnl_link_stats64(nla_data(attr), stats);
947 if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF) &&
948 nla_put_u32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent)))
949 goto nla_put_failure;
951 if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent
952 && (ext_filter_mask & RTEXT_FILTER_VF)) {
955 struct nlattr *vfinfo, *vf;
956 int num_vfs = dev_num_vf(dev->dev.parent);
958 vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
960 goto nla_put_failure;
961 for (i = 0; i < num_vfs; i++) {
962 struct ifla_vf_info ivi;
963 struct ifla_vf_mac vf_mac;
964 struct ifla_vf_vlan vf_vlan;
965 struct ifla_vf_tx_rate vf_tx_rate;
966 struct ifla_vf_spoofchk vf_spoofchk;
969 * Not all SR-IOV capable drivers support the
970 * spoofcheck query. Preset to -1 so the user
971 * space tool can detect that the driver didn't
975 if (dev->netdev_ops->ndo_get_vf_config(dev, i, &ivi))
980 vf_spoofchk.vf = ivi.vf;
982 memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
983 vf_vlan.vlan = ivi.vlan;
984 vf_vlan.qos = ivi.qos;
985 vf_tx_rate.rate = ivi.tx_rate;
986 vf_spoofchk.setting = ivi.spoofchk;
987 vf = nla_nest_start(skb, IFLA_VF_INFO);
989 nla_nest_cancel(skb, vfinfo);
990 goto nla_put_failure;
992 if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
993 nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
994 nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
996 nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
998 goto nla_put_failure;
999 nla_nest_end(skb, vf);
1001 nla_nest_end(skb, vfinfo);
1004 if (rtnl_port_fill(skb, dev))
1005 goto nla_put_failure;
1007 if (dev->rtnl_link_ops) {
1008 if (rtnl_link_fill(skb, dev) < 0)
1009 goto nla_put_failure;
1012 if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1013 goto nla_put_failure;
1015 list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1016 if (af_ops->fill_link_af) {
1020 if (!(af = nla_nest_start(skb, af_ops->family)))
1021 goto nla_put_failure;
1023 err = af_ops->fill_link_af(skb, dev);
1026 * Caller may return ENODATA to indicate that there
1027 * was no data to be dumped. This is not an error, it
1028 * means we should trim the attribute header and
1031 if (err == -ENODATA)
1032 nla_nest_cancel(skb, af);
1034 goto nla_put_failure;
1036 nla_nest_end(skb, af);
1040 nla_nest_end(skb, af_spec);
1042 return nlmsg_end(skb, nlh);
1045 nlmsg_cancel(skb, nlh);
1049 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1051 struct net *net = sock_net(skb->sk);
1054 struct net_device *dev;
1055 struct hlist_head *head;
1056 struct hlist_node *node;
1057 struct nlattr *tb[IFLA_MAX+1];
1058 u32 ext_filter_mask = 0;
1061 s_idx = cb->args[1];
1064 cb->seq = net->dev_base_seq;
1066 if (nlmsg_parse(cb->nlh, sizeof(struct rtgenmsg), tb, IFLA_MAX,
1067 ifla_policy) >= 0) {
1069 if (tb[IFLA_EXT_MASK])
1070 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1073 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1075 head = &net->dev_index_head[h];
1076 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
1079 if (rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1080 NETLINK_CB(cb->skb).pid,
1081 cb->nlh->nlmsg_seq, 0,
1083 ext_filter_mask) <= 0)
1086 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1099 const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1100 [IFLA_IFNAME] = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1101 [IFLA_ADDRESS] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1102 [IFLA_BROADCAST] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1103 [IFLA_MAP] = { .len = sizeof(struct rtnl_link_ifmap) },
1104 [IFLA_MTU] = { .type = NLA_U32 },
1105 [IFLA_LINK] = { .type = NLA_U32 },
1106 [IFLA_MASTER] = { .type = NLA_U32 },
1107 [IFLA_TXQLEN] = { .type = NLA_U32 },
1108 [IFLA_WEIGHT] = { .type = NLA_U32 },
1109 [IFLA_OPERSTATE] = { .type = NLA_U8 },
1110 [IFLA_LINKMODE] = { .type = NLA_U8 },
1111 [IFLA_LINKINFO] = { .type = NLA_NESTED },
1112 [IFLA_NET_NS_PID] = { .type = NLA_U32 },
1113 [IFLA_NET_NS_FD] = { .type = NLA_U32 },
1114 [IFLA_IFALIAS] = { .type = NLA_STRING, .len = IFALIASZ-1 },
1115 [IFLA_VFINFO_LIST] = {. type = NLA_NESTED },
1116 [IFLA_VF_PORTS] = { .type = NLA_NESTED },
1117 [IFLA_PORT_SELF] = { .type = NLA_NESTED },
1118 [IFLA_AF_SPEC] = { .type = NLA_NESTED },
1119 [IFLA_EXT_MASK] = { .type = NLA_U32 },
1121 EXPORT_SYMBOL(ifla_policy);
1123 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1124 [IFLA_INFO_KIND] = { .type = NLA_STRING },
1125 [IFLA_INFO_DATA] = { .type = NLA_NESTED },
1128 static const struct nla_policy ifla_vfinfo_policy[IFLA_VF_INFO_MAX+1] = {
1129 [IFLA_VF_INFO] = { .type = NLA_NESTED },
1132 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1133 [IFLA_VF_MAC] = { .type = NLA_BINARY,
1134 .len = sizeof(struct ifla_vf_mac) },
1135 [IFLA_VF_VLAN] = { .type = NLA_BINARY,
1136 .len = sizeof(struct ifla_vf_vlan) },
1137 [IFLA_VF_TX_RATE] = { .type = NLA_BINARY,
1138 .len = sizeof(struct ifla_vf_tx_rate) },
1139 [IFLA_VF_SPOOFCHK] = { .type = NLA_BINARY,
1140 .len = sizeof(struct ifla_vf_spoofchk) },
1143 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1144 [IFLA_PORT_VF] = { .type = NLA_U32 },
1145 [IFLA_PORT_PROFILE] = { .type = NLA_STRING,
1146 .len = PORT_PROFILE_MAX },
1147 [IFLA_PORT_VSI_TYPE] = { .type = NLA_BINARY,
1148 .len = sizeof(struct ifla_port_vsi)},
1149 [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1150 .len = PORT_UUID_MAX },
1151 [IFLA_PORT_HOST_UUID] = { .type = NLA_STRING,
1152 .len = PORT_UUID_MAX },
1153 [IFLA_PORT_REQUEST] = { .type = NLA_U8, },
1154 [IFLA_PORT_RESPONSE] = { .type = NLA_U16, },
1157 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1160 /* Examine the link attributes and figure out which
1161 * network namespace we are talking about.
1163 if (tb[IFLA_NET_NS_PID])
1164 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1165 else if (tb[IFLA_NET_NS_FD])
1166 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1168 net = get_net(src_net);
1171 EXPORT_SYMBOL(rtnl_link_get_net);
1173 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1176 if (tb[IFLA_ADDRESS] &&
1177 nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1180 if (tb[IFLA_BROADCAST] &&
1181 nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1185 if (tb[IFLA_AF_SPEC]) {
1189 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1190 const struct rtnl_af_ops *af_ops;
1192 if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1193 return -EAFNOSUPPORT;
1195 if (!af_ops->set_link_af)
1198 if (af_ops->validate_link_af) {
1199 err = af_ops->validate_link_af(dev, af);
1209 static int do_setvfinfo(struct net_device *dev, struct nlattr *attr)
1211 int rem, err = -EINVAL;
1213 const struct net_device_ops *ops = dev->netdev_ops;
1215 nla_for_each_nested(vf, attr, rem) {
1216 switch (nla_type(vf)) {
1218 struct ifla_vf_mac *ivm;
1221 if (ops->ndo_set_vf_mac)
1222 err = ops->ndo_set_vf_mac(dev, ivm->vf,
1226 case IFLA_VF_VLAN: {
1227 struct ifla_vf_vlan *ivv;
1230 if (ops->ndo_set_vf_vlan)
1231 err = ops->ndo_set_vf_vlan(dev, ivv->vf,
1236 case IFLA_VF_TX_RATE: {
1237 struct ifla_vf_tx_rate *ivt;
1240 if (ops->ndo_set_vf_tx_rate)
1241 err = ops->ndo_set_vf_tx_rate(dev, ivt->vf,
1245 case IFLA_VF_SPOOFCHK: {
1246 struct ifla_vf_spoofchk *ivs;
1249 if (ops->ndo_set_vf_spoofchk)
1250 err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1264 static int do_set_master(struct net_device *dev, int ifindex)
1266 struct net_device *master_dev;
1267 const struct net_device_ops *ops;
1271 if (dev->master->ifindex == ifindex)
1273 ops = dev->master->netdev_ops;
1274 if (ops->ndo_del_slave) {
1275 err = ops->ndo_del_slave(dev->master, dev);
1284 master_dev = __dev_get_by_index(dev_net(dev), ifindex);
1287 ops = master_dev->netdev_ops;
1288 if (ops->ndo_add_slave) {
1289 err = ops->ndo_add_slave(master_dev, dev);
1299 static int do_setlink(struct net_device *dev, struct ifinfomsg *ifm,
1300 struct nlattr **tb, char *ifname, int modified)
1302 const struct net_device_ops *ops = dev->netdev_ops;
1303 int send_addr_notify = 0;
1306 if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1307 struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1312 err = dev_change_net_namespace(dev, net, ifname);
1320 struct rtnl_link_ifmap *u_map;
1323 if (!ops->ndo_set_config) {
1328 if (!netif_device_present(dev)) {
1333 u_map = nla_data(tb[IFLA_MAP]);
1334 k_map.mem_start = (unsigned long) u_map->mem_start;
1335 k_map.mem_end = (unsigned long) u_map->mem_end;
1336 k_map.base_addr = (unsigned short) u_map->base_addr;
1337 k_map.irq = (unsigned char) u_map->irq;
1338 k_map.dma = (unsigned char) u_map->dma;
1339 k_map.port = (unsigned char) u_map->port;
1341 err = ops->ndo_set_config(dev, &k_map);
1348 if (tb[IFLA_ADDRESS]) {
1349 struct sockaddr *sa;
1352 if (!ops->ndo_set_mac_address) {
1357 if (!netif_device_present(dev)) {
1362 len = sizeof(sa_family_t) + dev->addr_len;
1363 sa = kmalloc(len, GFP_KERNEL);
1368 sa->sa_family = dev->type;
1369 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1371 err = ops->ndo_set_mac_address(dev, sa);
1375 send_addr_notify = 1;
1380 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1386 if (tb[IFLA_GROUP]) {
1387 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1392 * Interface selected by interface index but interface
1393 * name provided implies that a name change has been
1396 if (ifm->ifi_index > 0 && ifname[0]) {
1397 err = dev_change_name(dev, ifname);
1403 if (tb[IFLA_IFALIAS]) {
1404 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
1405 nla_len(tb[IFLA_IFALIAS]));
1411 if (tb[IFLA_BROADCAST]) {
1412 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
1413 send_addr_notify = 1;
1416 if (ifm->ifi_flags || ifm->ifi_change) {
1417 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1422 if (tb[IFLA_MASTER]) {
1423 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
1429 if (tb[IFLA_TXQLEN])
1430 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1432 if (tb[IFLA_OPERSTATE])
1433 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1435 if (tb[IFLA_LINKMODE]) {
1436 write_lock_bh(&dev_base_lock);
1437 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1438 write_unlock_bh(&dev_base_lock);
1441 if (tb[IFLA_VFINFO_LIST]) {
1442 struct nlattr *attr;
1444 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
1445 if (nla_type(attr) != IFLA_VF_INFO) {
1449 err = do_setvfinfo(dev, attr);
1457 if (tb[IFLA_VF_PORTS]) {
1458 struct nlattr *port[IFLA_PORT_MAX+1];
1459 struct nlattr *attr;
1464 if (!ops->ndo_set_vf_port)
1467 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
1468 if (nla_type(attr) != IFLA_VF_PORT)
1470 err = nla_parse_nested(port, IFLA_PORT_MAX,
1471 attr, ifla_port_policy);
1474 if (!port[IFLA_PORT_VF]) {
1478 vf = nla_get_u32(port[IFLA_PORT_VF]);
1479 err = ops->ndo_set_vf_port(dev, vf, port);
1487 if (tb[IFLA_PORT_SELF]) {
1488 struct nlattr *port[IFLA_PORT_MAX+1];
1490 err = nla_parse_nested(port, IFLA_PORT_MAX,
1491 tb[IFLA_PORT_SELF], ifla_port_policy);
1496 if (ops->ndo_set_vf_port)
1497 err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
1503 if (tb[IFLA_AF_SPEC]) {
1507 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1508 const struct rtnl_af_ops *af_ops;
1510 if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1513 err = af_ops->set_link_af(dev, af);
1523 if (err < 0 && modified && net_ratelimit())
1524 printk(KERN_WARNING "A link change request failed with "
1525 "some changes committed already. Interface %s may "
1526 "have been left with an inconsistent configuration, "
1527 "please check.\n", dev->name);
1529 if (send_addr_notify)
1530 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
1535 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1537 struct net *net = sock_net(skb->sk);
1538 struct ifinfomsg *ifm;
1539 struct net_device *dev;
1541 struct nlattr *tb[IFLA_MAX+1];
1542 char ifname[IFNAMSIZ];
1544 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1548 if (tb[IFLA_IFNAME])
1549 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1554 ifm = nlmsg_data(nlh);
1555 if (ifm->ifi_index > 0)
1556 dev = __dev_get_by_index(net, ifm->ifi_index);
1557 else if (tb[IFLA_IFNAME])
1558 dev = __dev_get_by_name(net, ifname);
1567 err = validate_linkmsg(dev, tb);
1571 err = do_setlink(dev, ifm, tb, ifname, 0);
1576 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1578 struct net *net = sock_net(skb->sk);
1579 const struct rtnl_link_ops *ops;
1580 struct net_device *dev;
1581 struct ifinfomsg *ifm;
1582 char ifname[IFNAMSIZ];
1583 struct nlattr *tb[IFLA_MAX+1];
1585 LIST_HEAD(list_kill);
1587 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1591 if (tb[IFLA_IFNAME])
1592 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1594 ifm = nlmsg_data(nlh);
1595 if (ifm->ifi_index > 0)
1596 dev = __dev_get_by_index(net, ifm->ifi_index);
1597 else if (tb[IFLA_IFNAME])
1598 dev = __dev_get_by_name(net, ifname);
1605 ops = dev->rtnl_link_ops;
1609 ops->dellink(dev, &list_kill);
1610 unregister_netdevice_many(&list_kill);
1611 list_del(&list_kill);
1615 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
1617 unsigned int old_flags;
1620 old_flags = dev->flags;
1621 if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
1622 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1627 dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
1628 rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U);
1630 __dev_notify_flags(dev, old_flags);
1633 EXPORT_SYMBOL(rtnl_configure_link);
1635 struct net_device *rtnl_create_link(struct net *src_net, struct net *net,
1636 char *ifname, const struct rtnl_link_ops *ops, struct nlattr *tb[])
1639 struct net_device *dev;
1640 unsigned int num_queues = 1;
1641 unsigned int real_num_queues = 1;
1643 if (ops->get_tx_queues) {
1644 err = ops->get_tx_queues(src_net, tb, &num_queues,
1650 dev = alloc_netdev_mq(ops->priv_size, ifname, ops->setup, num_queues);
1654 dev_net_set(dev, net);
1655 dev->rtnl_link_ops = ops;
1656 dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
1659 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
1660 if (tb[IFLA_ADDRESS])
1661 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
1662 nla_len(tb[IFLA_ADDRESS]));
1663 if (tb[IFLA_BROADCAST])
1664 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
1665 nla_len(tb[IFLA_BROADCAST]));
1666 if (tb[IFLA_TXQLEN])
1667 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1668 if (tb[IFLA_OPERSTATE])
1669 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1670 if (tb[IFLA_LINKMODE])
1671 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1673 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1678 return ERR_PTR(err);
1680 EXPORT_SYMBOL(rtnl_create_link);
1682 static int rtnl_group_changelink(struct net *net, int group,
1683 struct ifinfomsg *ifm,
1686 struct net_device *dev;
1689 for_each_netdev(net, dev) {
1690 if (dev->group == group) {
1691 err = do_setlink(dev, ifm, tb, NULL, 0);
1700 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1702 struct net *net = sock_net(skb->sk);
1703 const struct rtnl_link_ops *ops;
1704 struct net_device *dev;
1705 struct ifinfomsg *ifm;
1706 char kind[MODULE_NAME_LEN];
1707 char ifname[IFNAMSIZ];
1708 struct nlattr *tb[IFLA_MAX+1];
1709 struct nlattr *linkinfo[IFLA_INFO_MAX+1];
1712 #ifdef CONFIG_MODULES
1715 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1719 if (tb[IFLA_IFNAME])
1720 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1724 ifm = nlmsg_data(nlh);
1725 if (ifm->ifi_index > 0)
1726 dev = __dev_get_by_index(net, ifm->ifi_index);
1729 dev = __dev_get_by_name(net, ifname);
1734 err = validate_linkmsg(dev, tb);
1738 if (tb[IFLA_LINKINFO]) {
1739 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
1740 tb[IFLA_LINKINFO], ifla_info_policy);
1744 memset(linkinfo, 0, sizeof(linkinfo));
1746 if (linkinfo[IFLA_INFO_KIND]) {
1747 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
1748 ops = rtnl_link_ops_get(kind);
1755 struct nlattr *attr[ops ? ops->maxtype + 1 : 0], **data = NULL;
1756 struct net *dest_net;
1759 if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
1760 err = nla_parse_nested(attr, ops->maxtype,
1761 linkinfo[IFLA_INFO_DATA],
1767 if (ops->validate) {
1768 err = ops->validate(tb, data);
1777 if (nlh->nlmsg_flags & NLM_F_EXCL)
1779 if (nlh->nlmsg_flags & NLM_F_REPLACE)
1782 if (linkinfo[IFLA_INFO_DATA]) {
1783 if (!ops || ops != dev->rtnl_link_ops ||
1787 err = ops->changelink(dev, tb, data);
1793 return do_setlink(dev, ifm, tb, ifname, modified);
1796 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
1797 if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
1798 return rtnl_group_changelink(net,
1799 nla_get_u32(tb[IFLA_GROUP]),
1806 if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
1810 #ifdef CONFIG_MODULES
1813 request_module("rtnl-link-%s", kind);
1815 ops = rtnl_link_ops_get(kind);
1824 snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
1826 dest_net = rtnl_link_get_net(net, tb);
1827 if (IS_ERR(dest_net))
1828 return PTR_ERR(dest_net);
1830 dev = rtnl_create_link(net, dest_net, ifname, ops, tb);
1834 else if (ops->newlink)
1835 err = ops->newlink(net, dev, tb, data);
1837 err = register_netdevice(dev);
1839 if (err < 0 && !IS_ERR(dev))
1844 err = rtnl_configure_link(dev, ifm);
1846 unregister_netdevice(dev);
1853 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
1855 struct net *net = sock_net(skb->sk);
1856 struct ifinfomsg *ifm;
1857 char ifname[IFNAMSIZ];
1858 struct nlattr *tb[IFLA_MAX+1];
1859 struct net_device *dev = NULL;
1860 struct sk_buff *nskb;
1862 u32 ext_filter_mask = 0;
1864 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1868 if (tb[IFLA_IFNAME])
1869 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1871 if (tb[IFLA_EXT_MASK])
1872 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1874 ifm = nlmsg_data(nlh);
1875 if (ifm->ifi_index > 0)
1876 dev = __dev_get_by_index(net, ifm->ifi_index);
1877 else if (tb[IFLA_IFNAME])
1878 dev = __dev_get_by_name(net, ifname);
1885 nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
1889 err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).pid,
1890 nlh->nlmsg_seq, 0, 0, ext_filter_mask);
1892 /* -EMSGSIZE implies BUG in if_nlmsg_size */
1893 WARN_ON(err == -EMSGSIZE);
1896 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).pid);
1901 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
1903 struct net *net = sock_net(skb->sk);
1904 struct net_device *dev;
1905 struct nlattr *tb[IFLA_MAX+1];
1906 u32 ext_filter_mask = 0;
1907 u16 min_ifinfo_dump_size = 0;
1909 if (nlmsg_parse(nlh, sizeof(struct rtgenmsg), tb, IFLA_MAX,
1910 ifla_policy) >= 0) {
1911 if (tb[IFLA_EXT_MASK])
1912 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1915 if (!ext_filter_mask)
1916 return NLMSG_GOODSIZE;
1918 * traverse the list of net devices and compute the minimum
1919 * buffer size based upon the filter mask.
1921 list_for_each_entry(dev, &net->dev_base_head, dev_list) {
1922 min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
1927 return min_ifinfo_dump_size;
1930 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
1933 int s_idx = cb->family;
1937 for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
1938 int type = cb->nlh->nlmsg_type-RTM_BASE;
1939 if (idx < s_idx || idx == PF_PACKET)
1941 if (rtnl_msg_handlers[idx] == NULL ||
1942 rtnl_msg_handlers[idx][type].dumpit == NULL)
1945 memset(&cb->args[0], 0, sizeof(cb->args));
1946 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
1954 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned change)
1956 struct net *net = dev_net(dev);
1957 struct sk_buff *skb;
1959 size_t if_info_size;
1961 skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), GFP_KERNEL);
1965 err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0);
1967 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
1968 WARN_ON(err == -EMSGSIZE);
1972 rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_KERNEL);
1976 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
1979 /* Protected by RTNL sempahore. */
1980 static struct rtattr **rta_buf;
1981 static int rtattr_max;
1983 /* Process one rtnetlink message. */
1985 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
1987 struct net *net = sock_net(skb->sk);
1988 rtnl_doit_func doit;
1995 type = nlh->nlmsg_type;
2001 /* All the messages must have at least 1 byte length */
2002 if (nlh->nlmsg_len < NLMSG_LENGTH(sizeof(struct rtgenmsg)))
2005 family = ((struct rtgenmsg *)NLMSG_DATA(nlh))->rtgen_family;
2009 if (kind != 2 && !capable(CAP_NET_ADMIN))
2012 if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
2014 rtnl_dumpit_func dumpit;
2015 rtnl_calcit_func calcit;
2016 u16 min_dump_alloc = 0;
2018 dumpit = rtnl_get_dumpit(family, type);
2021 calcit = rtnl_get_calcit(family, type);
2023 min_dump_alloc = calcit(skb, nlh);
2028 struct netlink_dump_control c = {
2030 .min_dump_alloc = min_dump_alloc,
2032 err = netlink_dump_start(rtnl, skb, nlh, &c);
2038 memset(rta_buf, 0, (rtattr_max * sizeof(struct rtattr *)));
2040 min_len = rtm_min[sz_idx];
2041 if (nlh->nlmsg_len < min_len)
2044 if (nlh->nlmsg_len > min_len) {
2045 int attrlen = nlh->nlmsg_len - NLMSG_ALIGN(min_len);
2046 struct rtattr *attr = (void *)nlh + NLMSG_ALIGN(min_len);
2048 while (RTA_OK(attr, attrlen)) {
2049 unsigned flavor = attr->rta_type;
2051 if (flavor > rta_max[sz_idx])
2053 rta_buf[flavor-1] = attr;
2055 attr = RTA_NEXT(attr, attrlen);
2059 doit = rtnl_get_doit(family, type);
2063 return doit(skb, nlh, (void *)&rta_buf[0]);
2066 static void rtnetlink_rcv(struct sk_buff *skb)
2069 netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
2073 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
2075 struct net_device *dev = ptr;
2081 case NETDEV_POST_INIT:
2082 case NETDEV_REGISTER:
2084 case NETDEV_PRE_TYPE_CHANGE:
2085 case NETDEV_GOING_DOWN:
2086 case NETDEV_UNREGISTER:
2087 case NETDEV_UNREGISTER_BATCH:
2088 case NETDEV_RELEASE:
2092 rtmsg_ifinfo(RTM_NEWLINK, dev, 0);
2098 static struct notifier_block rtnetlink_dev_notifier = {
2099 .notifier_call = rtnetlink_event,
2103 static int __net_init rtnetlink_net_init(struct net *net)
2106 sk = netlink_kernel_create(net, NETLINK_ROUTE, RTNLGRP_MAX,
2107 rtnetlink_rcv, &rtnl_mutex, THIS_MODULE);
2114 static void __net_exit rtnetlink_net_exit(struct net *net)
2116 netlink_kernel_release(net->rtnl);
2120 static struct pernet_operations rtnetlink_net_ops = {
2121 .init = rtnetlink_net_init,
2122 .exit = rtnetlink_net_exit,
2125 void __init rtnetlink_init(void)
2130 for (i = 0; i < ARRAY_SIZE(rta_max); i++)
2131 if (rta_max[i] > rtattr_max)
2132 rtattr_max = rta_max[i];
2133 rta_buf = kmalloc(rtattr_max * sizeof(struct rtattr *), GFP_KERNEL);
2135 panic("rtnetlink_init: cannot allocate rta_buf\n");
2137 if (register_pernet_subsys(&rtnetlink_net_ops))
2138 panic("rtnetlink_init: cannot initialize rtnetlink\n");
2140 netlink_set_nonroot(NETLINK_ROUTE, NL_NONROOT_RECV);
2141 register_netdevice_notifier(&rtnetlink_dev_notifier);
2143 rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
2144 rtnl_dump_ifinfo, rtnl_calcit);
2145 rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
2146 rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
2147 rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
2149 rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
2150 rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);