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);
63 static u16 min_ifinfo_dump_size;
67 mutex_lock(&rtnl_mutex);
69 EXPORT_SYMBOL(rtnl_lock);
71 void __rtnl_unlock(void)
73 mutex_unlock(&rtnl_mutex);
76 void rtnl_unlock(void)
78 /* This fellow will unlock it for us. */
81 EXPORT_SYMBOL(rtnl_unlock);
83 int rtnl_trylock(void)
85 return mutex_trylock(&rtnl_mutex);
87 EXPORT_SYMBOL(rtnl_trylock);
89 int rtnl_is_locked(void)
91 return mutex_is_locked(&rtnl_mutex);
93 EXPORT_SYMBOL(rtnl_is_locked);
95 #ifdef CONFIG_PROVE_LOCKING
96 int lockdep_rtnl_is_held(void)
98 return lockdep_is_held(&rtnl_mutex);
100 EXPORT_SYMBOL(lockdep_rtnl_is_held);
101 #endif /* #ifdef CONFIG_PROVE_LOCKING */
103 static struct rtnl_link *rtnl_msg_handlers[RTNL_FAMILY_MAX + 1];
105 static inline int rtm_msgindex(int msgtype)
107 int msgindex = msgtype - RTM_BASE;
110 * msgindex < 0 implies someone tried to register a netlink
111 * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
112 * the message type has not been added to linux/rtnetlink.h
114 BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
119 static rtnl_doit_func rtnl_get_doit(int protocol, int msgindex)
121 struct rtnl_link *tab;
123 if (protocol <= RTNL_FAMILY_MAX)
124 tab = rtnl_msg_handlers[protocol];
128 if (tab == NULL || tab[msgindex].doit == NULL)
129 tab = rtnl_msg_handlers[PF_UNSPEC];
131 return tab ? tab[msgindex].doit : NULL;
134 static rtnl_dumpit_func rtnl_get_dumpit(int protocol, int msgindex)
136 struct rtnl_link *tab;
138 if (protocol <= RTNL_FAMILY_MAX)
139 tab = rtnl_msg_handlers[protocol];
143 if (tab == NULL || tab[msgindex].dumpit == NULL)
144 tab = rtnl_msg_handlers[PF_UNSPEC];
146 return tab ? tab[msgindex].dumpit : NULL;
149 static rtnl_calcit_func rtnl_get_calcit(int protocol, int msgindex)
151 struct rtnl_link *tab;
153 if (protocol <= RTNL_FAMILY_MAX)
154 tab = rtnl_msg_handlers[protocol];
158 if (tab == NULL || tab[msgindex].calcit == NULL)
159 tab = rtnl_msg_handlers[PF_UNSPEC];
161 return tab ? tab[msgindex].calcit : NULL;
165 * __rtnl_register - Register a rtnetlink message type
166 * @protocol: Protocol family or PF_UNSPEC
167 * @msgtype: rtnetlink message type
168 * @doit: Function pointer called for each request message
169 * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
170 * @calcit: Function pointer to calc size of dump message
172 * Registers the specified function pointers (at least one of them has
173 * to be non-NULL) to be called whenever a request message for the
174 * specified protocol family and message type is received.
176 * The special protocol family PF_UNSPEC may be used to define fallback
177 * function pointers for the case when no entry for the specific protocol
180 * Returns 0 on success or a negative error code.
182 int __rtnl_register(int protocol, int msgtype,
183 rtnl_doit_func doit, rtnl_dumpit_func dumpit,
184 rtnl_calcit_func calcit)
186 struct rtnl_link *tab;
189 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
190 msgindex = rtm_msgindex(msgtype);
192 tab = rtnl_msg_handlers[protocol];
194 tab = kcalloc(RTM_NR_MSGTYPES, sizeof(*tab), GFP_KERNEL);
198 rtnl_msg_handlers[protocol] = tab;
202 tab[msgindex].doit = doit;
205 tab[msgindex].dumpit = dumpit;
208 tab[msgindex].calcit = calcit;
212 EXPORT_SYMBOL_GPL(__rtnl_register);
215 * rtnl_register - Register a rtnetlink message type
217 * Identical to __rtnl_register() but panics on failure. This is useful
218 * as failure of this function is very unlikely, it can only happen due
219 * to lack of memory when allocating the chain to store all message
220 * handlers for a protocol. Meant for use in init functions where lack
221 * of memory implies no sense in continuing.
223 void rtnl_register(int protocol, int msgtype,
224 rtnl_doit_func doit, rtnl_dumpit_func dumpit,
225 rtnl_calcit_func calcit)
227 if (__rtnl_register(protocol, msgtype, doit, dumpit, calcit) < 0)
228 panic("Unable to register rtnetlink message handler, "
229 "protocol = %d, message type = %d\n",
232 EXPORT_SYMBOL_GPL(rtnl_register);
235 * rtnl_unregister - Unregister a rtnetlink message type
236 * @protocol: Protocol family or PF_UNSPEC
237 * @msgtype: rtnetlink message type
239 * Returns 0 on success or a negative error code.
241 int rtnl_unregister(int protocol, int msgtype)
245 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
246 msgindex = rtm_msgindex(msgtype);
248 if (rtnl_msg_handlers[protocol] == NULL)
251 rtnl_msg_handlers[protocol][msgindex].doit = NULL;
252 rtnl_msg_handlers[protocol][msgindex].dumpit = NULL;
256 EXPORT_SYMBOL_GPL(rtnl_unregister);
259 * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
260 * @protocol : Protocol family or PF_UNSPEC
262 * Identical to calling rtnl_unregster() for all registered message types
263 * of a certain protocol family.
265 void rtnl_unregister_all(int protocol)
267 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
269 kfree(rtnl_msg_handlers[protocol]);
270 rtnl_msg_handlers[protocol] = NULL;
272 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
274 static LIST_HEAD(link_ops);
277 * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
278 * @ops: struct rtnl_link_ops * to register
280 * The caller must hold the rtnl_mutex. This function should be used
281 * by drivers that create devices during module initialization. It
282 * must be called before registering the devices.
284 * Returns 0 on success or a negative error code.
286 int __rtnl_link_register(struct rtnl_link_ops *ops)
289 ops->dellink = unregister_netdevice_queue;
291 list_add_tail(&ops->list, &link_ops);
294 EXPORT_SYMBOL_GPL(__rtnl_link_register);
297 * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
298 * @ops: struct rtnl_link_ops * to register
300 * Returns 0 on success or a negative error code.
302 int rtnl_link_register(struct rtnl_link_ops *ops)
307 err = __rtnl_link_register(ops);
311 EXPORT_SYMBOL_GPL(rtnl_link_register);
313 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
315 struct net_device *dev;
316 LIST_HEAD(list_kill);
318 for_each_netdev(net, dev) {
319 if (dev->rtnl_link_ops == ops)
320 ops->dellink(dev, &list_kill);
322 unregister_netdevice_many(&list_kill);
326 * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
327 * @ops: struct rtnl_link_ops * to unregister
329 * The caller must hold the rtnl_mutex.
331 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
336 __rtnl_kill_links(net, ops);
338 list_del(&ops->list);
340 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
343 * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
344 * @ops: struct rtnl_link_ops * to unregister
346 void rtnl_link_unregister(struct rtnl_link_ops *ops)
349 __rtnl_link_unregister(ops);
352 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
354 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
356 const struct rtnl_link_ops *ops;
358 list_for_each_entry(ops, &link_ops, list) {
359 if (!strcmp(ops->kind, kind))
365 static size_t rtnl_link_get_size(const struct net_device *dev)
367 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
373 size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
374 nla_total_size(strlen(ops->kind) + 1); /* IFLA_INFO_KIND */
377 /* IFLA_INFO_DATA + nested data */
378 size += nla_total_size(sizeof(struct nlattr)) +
381 if (ops->get_xstats_size)
382 /* IFLA_INFO_XSTATS */
383 size += nla_total_size(ops->get_xstats_size(dev));
388 static LIST_HEAD(rtnl_af_ops);
390 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
392 const struct rtnl_af_ops *ops;
394 list_for_each_entry(ops, &rtnl_af_ops, list) {
395 if (ops->family == family)
403 * __rtnl_af_register - Register rtnl_af_ops with rtnetlink.
404 * @ops: struct rtnl_af_ops * to register
406 * The caller must hold the rtnl_mutex.
408 * Returns 0 on success or a negative error code.
410 int __rtnl_af_register(struct rtnl_af_ops *ops)
412 list_add_tail(&ops->list, &rtnl_af_ops);
415 EXPORT_SYMBOL_GPL(__rtnl_af_register);
418 * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
419 * @ops: struct rtnl_af_ops * to register
421 * Returns 0 on success or a negative error code.
423 int rtnl_af_register(struct rtnl_af_ops *ops)
428 err = __rtnl_af_register(ops);
432 EXPORT_SYMBOL_GPL(rtnl_af_register);
435 * __rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
436 * @ops: struct rtnl_af_ops * to unregister
438 * The caller must hold the rtnl_mutex.
440 void __rtnl_af_unregister(struct rtnl_af_ops *ops)
442 list_del(&ops->list);
444 EXPORT_SYMBOL_GPL(__rtnl_af_unregister);
447 * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
448 * @ops: struct rtnl_af_ops * to unregister
450 void rtnl_af_unregister(struct rtnl_af_ops *ops)
453 __rtnl_af_unregister(ops);
456 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
458 static size_t rtnl_link_get_af_size(const struct net_device *dev)
460 struct rtnl_af_ops *af_ops;
464 size = nla_total_size(sizeof(struct nlattr));
466 list_for_each_entry(af_ops, &rtnl_af_ops, list) {
467 if (af_ops->get_link_af_size) {
468 /* AF_* + nested data */
469 size += nla_total_size(sizeof(struct nlattr)) +
470 af_ops->get_link_af_size(dev);
477 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
479 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
480 struct nlattr *linkinfo, *data;
483 linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
484 if (linkinfo == NULL)
487 if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
488 goto err_cancel_link;
489 if (ops->fill_xstats) {
490 err = ops->fill_xstats(skb, dev);
492 goto err_cancel_link;
494 if (ops->fill_info) {
495 data = nla_nest_start(skb, IFLA_INFO_DATA);
497 goto err_cancel_link;
498 err = ops->fill_info(skb, dev);
500 goto err_cancel_data;
501 nla_nest_end(skb, data);
504 nla_nest_end(skb, linkinfo);
508 nla_nest_cancel(skb, data);
510 nla_nest_cancel(skb, linkinfo);
515 static const int rtm_min[RTM_NR_FAMILIES] =
517 [RTM_FAM(RTM_NEWLINK)] = NLMSG_LENGTH(sizeof(struct ifinfomsg)),
518 [RTM_FAM(RTM_NEWADDR)] = NLMSG_LENGTH(sizeof(struct ifaddrmsg)),
519 [RTM_FAM(RTM_NEWROUTE)] = NLMSG_LENGTH(sizeof(struct rtmsg)),
520 [RTM_FAM(RTM_NEWRULE)] = NLMSG_LENGTH(sizeof(struct fib_rule_hdr)),
521 [RTM_FAM(RTM_NEWQDISC)] = NLMSG_LENGTH(sizeof(struct tcmsg)),
522 [RTM_FAM(RTM_NEWTCLASS)] = NLMSG_LENGTH(sizeof(struct tcmsg)),
523 [RTM_FAM(RTM_NEWTFILTER)] = NLMSG_LENGTH(sizeof(struct tcmsg)),
524 [RTM_FAM(RTM_NEWACTION)] = NLMSG_LENGTH(sizeof(struct tcamsg)),
525 [RTM_FAM(RTM_GETMULTICAST)] = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
526 [RTM_FAM(RTM_GETANYCAST)] = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
529 static const int rta_max[RTM_NR_FAMILIES] =
531 [RTM_FAM(RTM_NEWLINK)] = IFLA_MAX,
532 [RTM_FAM(RTM_NEWADDR)] = IFA_MAX,
533 [RTM_FAM(RTM_NEWROUTE)] = RTA_MAX,
534 [RTM_FAM(RTM_NEWRULE)] = FRA_MAX,
535 [RTM_FAM(RTM_NEWQDISC)] = TCA_MAX,
536 [RTM_FAM(RTM_NEWTCLASS)] = TCA_MAX,
537 [RTM_FAM(RTM_NEWTFILTER)] = TCA_MAX,
538 [RTM_FAM(RTM_NEWACTION)] = TCAA_MAX,
541 void __rta_fill(struct sk_buff *skb, int attrtype, int attrlen, const void *data)
544 int size = RTA_LENGTH(attrlen);
546 rta = (struct rtattr *)skb_put(skb, RTA_ALIGN(size));
547 rta->rta_type = attrtype;
549 memcpy(RTA_DATA(rta), data, attrlen);
550 memset(RTA_DATA(rta) + attrlen, 0, RTA_ALIGN(size) - size);
552 EXPORT_SYMBOL(__rta_fill);
554 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned group, int echo)
556 struct sock *rtnl = net->rtnl;
559 NETLINK_CB(skb).dst_group = group;
561 atomic_inc(&skb->users);
562 netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
564 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
568 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
570 struct sock *rtnl = net->rtnl;
572 return nlmsg_unicast(rtnl, skb, pid);
574 EXPORT_SYMBOL(rtnl_unicast);
576 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
577 struct nlmsghdr *nlh, gfp_t flags)
579 struct sock *rtnl = net->rtnl;
583 report = nlmsg_report(nlh);
585 nlmsg_notify(rtnl, skb, pid, group, report, flags);
587 EXPORT_SYMBOL(rtnl_notify);
589 void rtnl_set_sk_err(struct net *net, u32 group, int error)
591 struct sock *rtnl = net->rtnl;
593 netlink_set_err(rtnl, 0, group, error);
595 EXPORT_SYMBOL(rtnl_set_sk_err);
597 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
602 mx = nla_nest_start(skb, RTA_METRICS);
606 for (i = 0; i < RTAX_MAX; i++) {
609 NLA_PUT_U32(skb, i+1, metrics[i]);
614 nla_nest_cancel(skb, mx);
618 return nla_nest_end(skb, mx);
621 nla_nest_cancel(skb, mx);
624 EXPORT_SYMBOL(rtnetlink_put_metrics);
626 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
627 u32 ts, u32 tsage, long expires, u32 error)
629 struct rta_cacheinfo ci = {
630 .rta_lastuse = jiffies_to_clock_t(jiffies - dst->lastuse),
631 .rta_used = dst->__use,
632 .rta_clntref = atomic_read(&(dst->__refcnt)),
640 ci.rta_expires = jiffies_to_clock_t(expires);
642 return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
644 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
646 static void set_operstate(struct net_device *dev, unsigned char transition)
648 unsigned char operstate = dev->operstate;
650 switch (transition) {
652 if ((operstate == IF_OPER_DORMANT ||
653 operstate == IF_OPER_UNKNOWN) &&
655 operstate = IF_OPER_UP;
658 case IF_OPER_DORMANT:
659 if (operstate == IF_OPER_UP ||
660 operstate == IF_OPER_UNKNOWN)
661 operstate = IF_OPER_DORMANT;
665 if (dev->operstate != operstate) {
666 write_lock_bh(&dev_base_lock);
667 dev->operstate = operstate;
668 write_unlock_bh(&dev_base_lock);
669 netdev_state_change(dev);
673 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
674 const struct ifinfomsg *ifm)
676 unsigned int flags = ifm->ifi_flags;
678 /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
680 flags = (flags & ifm->ifi_change) |
681 (dev->flags & ~ifm->ifi_change);
686 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
687 const struct rtnl_link_stats64 *b)
689 a->rx_packets = b->rx_packets;
690 a->tx_packets = b->tx_packets;
691 a->rx_bytes = b->rx_bytes;
692 a->tx_bytes = b->tx_bytes;
693 a->rx_errors = b->rx_errors;
694 a->tx_errors = b->tx_errors;
695 a->rx_dropped = b->rx_dropped;
696 a->tx_dropped = b->tx_dropped;
698 a->multicast = b->multicast;
699 a->collisions = b->collisions;
701 a->rx_length_errors = b->rx_length_errors;
702 a->rx_over_errors = b->rx_over_errors;
703 a->rx_crc_errors = b->rx_crc_errors;
704 a->rx_frame_errors = b->rx_frame_errors;
705 a->rx_fifo_errors = b->rx_fifo_errors;
706 a->rx_missed_errors = b->rx_missed_errors;
708 a->tx_aborted_errors = b->tx_aborted_errors;
709 a->tx_carrier_errors = b->tx_carrier_errors;
710 a->tx_fifo_errors = b->tx_fifo_errors;
711 a->tx_heartbeat_errors = b->tx_heartbeat_errors;
712 a->tx_window_errors = b->tx_window_errors;
714 a->rx_compressed = b->rx_compressed;
715 a->tx_compressed = b->tx_compressed;
718 static void copy_rtnl_link_stats64(void *v, const struct rtnl_link_stats64 *b)
720 memcpy(v, b, sizeof(*b));
724 static inline int rtnl_vfinfo_size(const struct net_device *dev)
726 if (dev->dev.parent && dev_is_pci(dev->dev.parent)) {
728 int num_vfs = dev_num_vf(dev->dev.parent);
729 size_t size = nla_total_size(sizeof(struct nlattr));
730 size += nla_total_size(num_vfs * sizeof(struct nlattr));
732 (nla_total_size(sizeof(struct ifla_vf_mac)) +
733 nla_total_size(sizeof(struct ifla_vf_vlan)) +
734 nla_total_size(sizeof(struct ifla_vf_tx_rate)) +
735 nla_total_size(sizeof(struct ifla_vf_spoofchk)));
741 static size_t rtnl_port_size(const struct net_device *dev)
743 size_t port_size = nla_total_size(4) /* PORT_VF */
744 + nla_total_size(PORT_PROFILE_MAX) /* PORT_PROFILE */
745 + nla_total_size(sizeof(struct ifla_port_vsi))
747 + nla_total_size(PORT_UUID_MAX) /* PORT_INSTANCE_UUID */
748 + nla_total_size(PORT_UUID_MAX) /* PORT_HOST_UUID */
749 + nla_total_size(1) /* PROT_VDP_REQUEST */
750 + nla_total_size(2); /* PORT_VDP_RESPONSE */
751 size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
752 size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
754 size_t port_self_size = nla_total_size(sizeof(struct nlattr))
757 if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent)
759 if (dev_num_vf(dev->dev.parent))
760 return port_self_size + vf_ports_size +
761 vf_port_size * dev_num_vf(dev->dev.parent);
763 return port_self_size;
766 static noinline size_t if_nlmsg_size(const struct net_device *dev)
768 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
769 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
770 + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
771 + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
772 + nla_total_size(sizeof(struct rtnl_link_ifmap))
773 + nla_total_size(sizeof(struct rtnl_link_stats))
774 + nla_total_size(sizeof(struct rtnl_link_stats64))
775 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
776 + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
777 + nla_total_size(4) /* IFLA_TXQLEN */
778 + nla_total_size(4) /* IFLA_WEIGHT */
779 + nla_total_size(4) /* IFLA_MTU */
780 + nla_total_size(4) /* IFLA_LINK */
781 + nla_total_size(4) /* IFLA_MASTER */
782 + nla_total_size(1) /* IFLA_OPERSTATE */
783 + nla_total_size(1) /* IFLA_LINKMODE */
784 + nla_total_size(4) /* IFLA_NUM_VF */
785 + rtnl_vfinfo_size(dev) /* IFLA_VFINFO_LIST */
786 + rtnl_port_size(dev) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
787 + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
788 + rtnl_link_get_af_size(dev); /* IFLA_AF_SPEC */
791 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
793 struct nlattr *vf_ports;
794 struct nlattr *vf_port;
798 vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
802 for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
803 vf_port = nla_nest_start(skb, IFLA_VF_PORT);
805 goto nla_put_failure;
806 NLA_PUT_U32(skb, IFLA_PORT_VF, vf);
807 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
808 if (err == -EMSGSIZE)
809 goto nla_put_failure;
811 nla_nest_cancel(skb, vf_port);
814 nla_nest_end(skb, vf_port);
817 nla_nest_end(skb, vf_ports);
822 nla_nest_cancel(skb, vf_ports);
826 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
828 struct nlattr *port_self;
831 port_self = nla_nest_start(skb, IFLA_PORT_SELF);
835 err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
837 nla_nest_cancel(skb, port_self);
838 return (err == -EMSGSIZE) ? err : 0;
841 nla_nest_end(skb, port_self);
846 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev)
850 if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent)
853 err = rtnl_port_self_fill(skb, dev);
857 if (dev_num_vf(dev->dev.parent)) {
858 err = rtnl_vf_ports_fill(skb, dev);
866 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
867 int type, u32 pid, u32 seq, u32 change,
870 struct ifinfomsg *ifm;
871 struct nlmsghdr *nlh;
872 struct rtnl_link_stats64 temp;
873 const struct rtnl_link_stats64 *stats;
874 struct nlattr *attr, *af_spec;
875 struct rtnl_af_ops *af_ops;
878 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
882 ifm = nlmsg_data(nlh);
883 ifm->ifi_family = AF_UNSPEC;
885 ifm->ifi_type = dev->type;
886 ifm->ifi_index = dev->ifindex;
887 ifm->ifi_flags = dev_get_flags(dev);
888 ifm->ifi_change = change;
890 NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
891 NLA_PUT_U32(skb, IFLA_TXQLEN, dev->tx_queue_len);
892 NLA_PUT_U8(skb, IFLA_OPERSTATE,
893 netif_running(dev) ? dev->operstate : IF_OPER_DOWN);
894 NLA_PUT_U8(skb, IFLA_LINKMODE, dev->link_mode);
895 NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
896 NLA_PUT_U32(skb, IFLA_GROUP, dev->group);
898 if (dev->ifindex != dev->iflink)
899 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
902 NLA_PUT_U32(skb, IFLA_MASTER, dev->master->ifindex);
905 NLA_PUT_STRING(skb, IFLA_QDISC, dev->qdisc->ops->id);
908 NLA_PUT_STRING(skb, IFLA_IFALIAS, dev->ifalias);
911 struct rtnl_link_ifmap map = {
912 .mem_start = dev->mem_start,
913 .mem_end = dev->mem_end,
914 .base_addr = dev->base_addr,
917 .port = dev->if_port,
919 NLA_PUT(skb, IFLA_MAP, sizeof(map), &map);
923 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
924 NLA_PUT(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast);
927 attr = nla_reserve(skb, IFLA_STATS,
928 sizeof(struct rtnl_link_stats));
930 goto nla_put_failure;
932 stats = dev_get_stats(dev, &temp);
933 copy_rtnl_link_stats(nla_data(attr), stats);
935 attr = nla_reserve(skb, IFLA_STATS64,
936 sizeof(struct rtnl_link_stats64));
938 goto nla_put_failure;
939 copy_rtnl_link_stats64(nla_data(attr), stats);
942 NLA_PUT_U32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent));
944 if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent) {
947 struct nlattr *vfinfo, *vf;
948 int num_vfs = dev_num_vf(dev->dev.parent);
950 vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
952 goto nla_put_failure;
953 for (i = 0; i < num_vfs; i++) {
954 struct ifla_vf_info ivi;
955 struct ifla_vf_mac vf_mac;
956 struct ifla_vf_vlan vf_vlan;
957 struct ifla_vf_tx_rate vf_tx_rate;
958 struct ifla_vf_spoofchk vf_spoofchk;
961 * Not all SR-IOV capable drivers support the
962 * spoofcheck query. Preset to -1 so the user
963 * space tool can detect that the driver didn't
967 if (dev->netdev_ops->ndo_get_vf_config(dev, i, &ivi))
972 vf_spoofchk.vf = ivi.vf;
974 memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
975 vf_vlan.vlan = ivi.vlan;
976 vf_vlan.qos = ivi.qos;
977 vf_tx_rate.rate = ivi.tx_rate;
978 vf_spoofchk.setting = ivi.spoofchk;
979 vf = nla_nest_start(skb, IFLA_VF_INFO);
981 nla_nest_cancel(skb, vfinfo);
982 goto nla_put_failure;
984 NLA_PUT(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac);
985 NLA_PUT(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan);
986 NLA_PUT(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
988 NLA_PUT(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
990 nla_nest_end(skb, vf);
992 nla_nest_end(skb, vfinfo);
995 if (rtnl_port_fill(skb, dev))
996 goto nla_put_failure;
998 if (dev->rtnl_link_ops) {
999 if (rtnl_link_fill(skb, dev) < 0)
1000 goto nla_put_failure;
1003 if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1004 goto nla_put_failure;
1006 list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1007 if (af_ops->fill_link_af) {
1011 if (!(af = nla_nest_start(skb, af_ops->family)))
1012 goto nla_put_failure;
1014 err = af_ops->fill_link_af(skb, dev);
1017 * Caller may return ENODATA to indicate that there
1018 * was no data to be dumped. This is not an error, it
1019 * means we should trim the attribute header and
1022 if (err == -ENODATA)
1023 nla_nest_cancel(skb, af);
1025 goto nla_put_failure;
1027 nla_nest_end(skb, af);
1031 nla_nest_end(skb, af_spec);
1033 return nlmsg_end(skb, nlh);
1036 nlmsg_cancel(skb, nlh);
1040 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1042 struct net *net = sock_net(skb->sk);
1045 struct net_device *dev;
1046 struct hlist_head *head;
1047 struct hlist_node *node;
1050 s_idx = cb->args[1];
1053 cb->seq = net->dev_base_seq;
1055 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1057 head = &net->dev_index_head[h];
1058 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
1061 if (rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1062 NETLINK_CB(cb->skb).pid,
1063 cb->nlh->nlmsg_seq, 0,
1067 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1080 const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1081 [IFLA_IFNAME] = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1082 [IFLA_ADDRESS] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1083 [IFLA_BROADCAST] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1084 [IFLA_MAP] = { .len = sizeof(struct rtnl_link_ifmap) },
1085 [IFLA_MTU] = { .type = NLA_U32 },
1086 [IFLA_LINK] = { .type = NLA_U32 },
1087 [IFLA_MASTER] = { .type = NLA_U32 },
1088 [IFLA_TXQLEN] = { .type = NLA_U32 },
1089 [IFLA_WEIGHT] = { .type = NLA_U32 },
1090 [IFLA_OPERSTATE] = { .type = NLA_U8 },
1091 [IFLA_LINKMODE] = { .type = NLA_U8 },
1092 [IFLA_LINKINFO] = { .type = NLA_NESTED },
1093 [IFLA_NET_NS_PID] = { .type = NLA_U32 },
1094 [IFLA_NET_NS_FD] = { .type = NLA_U32 },
1095 [IFLA_IFALIAS] = { .type = NLA_STRING, .len = IFALIASZ-1 },
1096 [IFLA_VFINFO_LIST] = {. type = NLA_NESTED },
1097 [IFLA_VF_PORTS] = { .type = NLA_NESTED },
1098 [IFLA_PORT_SELF] = { .type = NLA_NESTED },
1099 [IFLA_AF_SPEC] = { .type = NLA_NESTED },
1101 EXPORT_SYMBOL(ifla_policy);
1103 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1104 [IFLA_INFO_KIND] = { .type = NLA_STRING },
1105 [IFLA_INFO_DATA] = { .type = NLA_NESTED },
1108 static const struct nla_policy ifla_vfinfo_policy[IFLA_VF_INFO_MAX+1] = {
1109 [IFLA_VF_INFO] = { .type = NLA_NESTED },
1112 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1113 [IFLA_VF_MAC] = { .type = NLA_BINARY,
1114 .len = sizeof(struct ifla_vf_mac) },
1115 [IFLA_VF_VLAN] = { .type = NLA_BINARY,
1116 .len = sizeof(struct ifla_vf_vlan) },
1117 [IFLA_VF_TX_RATE] = { .type = NLA_BINARY,
1118 .len = sizeof(struct ifla_vf_tx_rate) },
1121 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1122 [IFLA_PORT_VF] = { .type = NLA_U32 },
1123 [IFLA_PORT_PROFILE] = { .type = NLA_STRING,
1124 .len = PORT_PROFILE_MAX },
1125 [IFLA_PORT_VSI_TYPE] = { .type = NLA_BINARY,
1126 .len = sizeof(struct ifla_port_vsi)},
1127 [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1128 .len = PORT_UUID_MAX },
1129 [IFLA_PORT_HOST_UUID] = { .type = NLA_STRING,
1130 .len = PORT_UUID_MAX },
1131 [IFLA_PORT_REQUEST] = { .type = NLA_U8, },
1132 [IFLA_PORT_RESPONSE] = { .type = NLA_U16, },
1135 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1138 /* Examine the link attributes and figure out which
1139 * network namespace we are talking about.
1141 if (tb[IFLA_NET_NS_PID])
1142 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1143 else if (tb[IFLA_NET_NS_FD])
1144 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1146 net = get_net(src_net);
1149 EXPORT_SYMBOL(rtnl_link_get_net);
1151 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1154 if (tb[IFLA_ADDRESS] &&
1155 nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1158 if (tb[IFLA_BROADCAST] &&
1159 nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1163 if (tb[IFLA_AF_SPEC]) {
1167 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1168 const struct rtnl_af_ops *af_ops;
1170 if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1171 return -EAFNOSUPPORT;
1173 if (!af_ops->set_link_af)
1176 if (af_ops->validate_link_af) {
1177 err = af_ops->validate_link_af(dev, af);
1187 static int do_setvfinfo(struct net_device *dev, struct nlattr *attr)
1189 int rem, err = -EINVAL;
1191 const struct net_device_ops *ops = dev->netdev_ops;
1193 nla_for_each_nested(vf, attr, rem) {
1194 switch (nla_type(vf)) {
1196 struct ifla_vf_mac *ivm;
1199 if (ops->ndo_set_vf_mac)
1200 err = ops->ndo_set_vf_mac(dev, ivm->vf,
1204 case IFLA_VF_VLAN: {
1205 struct ifla_vf_vlan *ivv;
1208 if (ops->ndo_set_vf_vlan)
1209 err = ops->ndo_set_vf_vlan(dev, ivv->vf,
1214 case IFLA_VF_TX_RATE: {
1215 struct ifla_vf_tx_rate *ivt;
1218 if (ops->ndo_set_vf_tx_rate)
1219 err = ops->ndo_set_vf_tx_rate(dev, ivt->vf,
1223 case IFLA_VF_SPOOFCHK: {
1224 struct ifla_vf_spoofchk *ivs;
1227 if (ops->ndo_set_vf_spoofchk)
1228 err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1242 static int do_set_master(struct net_device *dev, int ifindex)
1244 struct net_device *master_dev;
1245 const struct net_device_ops *ops;
1249 if (dev->master->ifindex == ifindex)
1251 ops = dev->master->netdev_ops;
1252 if (ops->ndo_del_slave) {
1253 err = ops->ndo_del_slave(dev->master, dev);
1262 master_dev = __dev_get_by_index(dev_net(dev), ifindex);
1265 ops = master_dev->netdev_ops;
1266 if (ops->ndo_add_slave) {
1267 err = ops->ndo_add_slave(master_dev, dev);
1277 static int do_setlink(struct net_device *dev, struct ifinfomsg *ifm,
1278 struct nlattr **tb, char *ifname, int modified)
1280 const struct net_device_ops *ops = dev->netdev_ops;
1281 int send_addr_notify = 0;
1284 if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1285 struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1290 err = dev_change_net_namespace(dev, net, ifname);
1298 struct rtnl_link_ifmap *u_map;
1301 if (!ops->ndo_set_config) {
1306 if (!netif_device_present(dev)) {
1311 u_map = nla_data(tb[IFLA_MAP]);
1312 k_map.mem_start = (unsigned long) u_map->mem_start;
1313 k_map.mem_end = (unsigned long) u_map->mem_end;
1314 k_map.base_addr = (unsigned short) u_map->base_addr;
1315 k_map.irq = (unsigned char) u_map->irq;
1316 k_map.dma = (unsigned char) u_map->dma;
1317 k_map.port = (unsigned char) u_map->port;
1319 err = ops->ndo_set_config(dev, &k_map);
1326 if (tb[IFLA_ADDRESS]) {
1327 struct sockaddr *sa;
1330 if (!ops->ndo_set_mac_address) {
1335 if (!netif_device_present(dev)) {
1340 len = sizeof(sa_family_t) + dev->addr_len;
1341 sa = kmalloc(len, GFP_KERNEL);
1346 sa->sa_family = dev->type;
1347 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1349 err = ops->ndo_set_mac_address(dev, sa);
1353 send_addr_notify = 1;
1358 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1364 if (tb[IFLA_GROUP]) {
1365 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1370 * Interface selected by interface index but interface
1371 * name provided implies that a name change has been
1374 if (ifm->ifi_index > 0 && ifname[0]) {
1375 err = dev_change_name(dev, ifname);
1381 if (tb[IFLA_IFALIAS]) {
1382 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
1383 nla_len(tb[IFLA_IFALIAS]));
1389 if (tb[IFLA_BROADCAST]) {
1390 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
1391 send_addr_notify = 1;
1394 if (ifm->ifi_flags || ifm->ifi_change) {
1395 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1400 if (tb[IFLA_MASTER]) {
1401 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
1407 if (tb[IFLA_TXQLEN])
1408 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1410 if (tb[IFLA_OPERSTATE])
1411 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1413 if (tb[IFLA_LINKMODE]) {
1414 write_lock_bh(&dev_base_lock);
1415 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1416 write_unlock_bh(&dev_base_lock);
1419 if (tb[IFLA_VFINFO_LIST]) {
1420 struct nlattr *attr;
1422 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
1423 if (nla_type(attr) != IFLA_VF_INFO) {
1427 err = do_setvfinfo(dev, attr);
1435 if (tb[IFLA_VF_PORTS]) {
1436 struct nlattr *port[IFLA_PORT_MAX+1];
1437 struct nlattr *attr;
1442 if (!ops->ndo_set_vf_port)
1445 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
1446 if (nla_type(attr) != IFLA_VF_PORT)
1448 err = nla_parse_nested(port, IFLA_PORT_MAX,
1449 attr, ifla_port_policy);
1452 if (!port[IFLA_PORT_VF]) {
1456 vf = nla_get_u32(port[IFLA_PORT_VF]);
1457 err = ops->ndo_set_vf_port(dev, vf, port);
1465 if (tb[IFLA_PORT_SELF]) {
1466 struct nlattr *port[IFLA_PORT_MAX+1];
1468 err = nla_parse_nested(port, IFLA_PORT_MAX,
1469 tb[IFLA_PORT_SELF], ifla_port_policy);
1474 if (ops->ndo_set_vf_port)
1475 err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
1481 if (tb[IFLA_AF_SPEC]) {
1485 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1486 const struct rtnl_af_ops *af_ops;
1488 if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1491 err = af_ops->set_link_af(dev, af);
1501 if (err < 0 && modified && net_ratelimit())
1502 printk(KERN_WARNING "A link change request failed with "
1503 "some changes committed already. Interface %s may "
1504 "have been left with an inconsistent configuration, "
1505 "please check.\n", dev->name);
1507 if (send_addr_notify)
1508 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
1512 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1514 struct net *net = sock_net(skb->sk);
1515 struct ifinfomsg *ifm;
1516 struct net_device *dev;
1518 struct nlattr *tb[IFLA_MAX+1];
1519 char ifname[IFNAMSIZ];
1521 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1525 if (tb[IFLA_IFNAME])
1526 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1531 ifm = nlmsg_data(nlh);
1532 if (ifm->ifi_index > 0)
1533 dev = __dev_get_by_index(net, ifm->ifi_index);
1534 else if (tb[IFLA_IFNAME])
1535 dev = __dev_get_by_name(net, ifname);
1544 err = validate_linkmsg(dev, tb);
1548 err = do_setlink(dev, ifm, tb, ifname, 0);
1553 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1555 struct net *net = sock_net(skb->sk);
1556 const struct rtnl_link_ops *ops;
1557 struct net_device *dev;
1558 struct ifinfomsg *ifm;
1559 char ifname[IFNAMSIZ];
1560 struct nlattr *tb[IFLA_MAX+1];
1562 LIST_HEAD(list_kill);
1564 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1568 if (tb[IFLA_IFNAME])
1569 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1571 ifm = nlmsg_data(nlh);
1572 if (ifm->ifi_index > 0)
1573 dev = __dev_get_by_index(net, ifm->ifi_index);
1574 else if (tb[IFLA_IFNAME])
1575 dev = __dev_get_by_name(net, ifname);
1582 ops = dev->rtnl_link_ops;
1586 ops->dellink(dev, &list_kill);
1587 unregister_netdevice_many(&list_kill);
1588 list_del(&list_kill);
1592 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
1594 unsigned int old_flags;
1597 old_flags = dev->flags;
1598 if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
1599 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1604 dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
1605 rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U);
1607 __dev_notify_flags(dev, old_flags);
1610 EXPORT_SYMBOL(rtnl_configure_link);
1612 struct net_device *rtnl_create_link(struct net *src_net, struct net *net,
1613 char *ifname, const struct rtnl_link_ops *ops, struct nlattr *tb[])
1616 struct net_device *dev;
1617 unsigned int num_queues = 1;
1618 unsigned int real_num_queues = 1;
1620 if (ops->get_tx_queues) {
1621 err = ops->get_tx_queues(src_net, tb, &num_queues,
1627 dev = alloc_netdev_mq(ops->priv_size, ifname, ops->setup, num_queues);
1631 dev_net_set(dev, net);
1632 dev->rtnl_link_ops = ops;
1633 dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
1636 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
1637 if (tb[IFLA_ADDRESS])
1638 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
1639 nla_len(tb[IFLA_ADDRESS]));
1640 if (tb[IFLA_BROADCAST])
1641 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
1642 nla_len(tb[IFLA_BROADCAST]));
1643 if (tb[IFLA_TXQLEN])
1644 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1645 if (tb[IFLA_OPERSTATE])
1646 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1647 if (tb[IFLA_LINKMODE])
1648 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1650 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1655 return ERR_PTR(err);
1657 EXPORT_SYMBOL(rtnl_create_link);
1659 static int rtnl_group_changelink(struct net *net, int group,
1660 struct ifinfomsg *ifm,
1663 struct net_device *dev;
1666 for_each_netdev(net, dev) {
1667 if (dev->group == group) {
1668 err = do_setlink(dev, ifm, tb, NULL, 0);
1677 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1679 struct net *net = sock_net(skb->sk);
1680 const struct rtnl_link_ops *ops;
1681 struct net_device *dev;
1682 struct ifinfomsg *ifm;
1683 char kind[MODULE_NAME_LEN];
1684 char ifname[IFNAMSIZ];
1685 struct nlattr *tb[IFLA_MAX+1];
1686 struct nlattr *linkinfo[IFLA_INFO_MAX+1];
1689 #ifdef CONFIG_MODULES
1692 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1696 if (tb[IFLA_IFNAME])
1697 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1701 ifm = nlmsg_data(nlh);
1702 if (ifm->ifi_index > 0)
1703 dev = __dev_get_by_index(net, ifm->ifi_index);
1706 dev = __dev_get_by_name(net, ifname);
1711 err = validate_linkmsg(dev, tb);
1715 if (tb[IFLA_LINKINFO]) {
1716 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
1717 tb[IFLA_LINKINFO], ifla_info_policy);
1721 memset(linkinfo, 0, sizeof(linkinfo));
1723 if (linkinfo[IFLA_INFO_KIND]) {
1724 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
1725 ops = rtnl_link_ops_get(kind);
1732 struct nlattr *attr[ops ? ops->maxtype + 1 : 0], **data = NULL;
1733 struct net *dest_net;
1736 if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
1737 err = nla_parse_nested(attr, ops->maxtype,
1738 linkinfo[IFLA_INFO_DATA],
1744 if (ops->validate) {
1745 err = ops->validate(tb, data);
1754 if (nlh->nlmsg_flags & NLM_F_EXCL)
1756 if (nlh->nlmsg_flags & NLM_F_REPLACE)
1759 if (linkinfo[IFLA_INFO_DATA]) {
1760 if (!ops || ops != dev->rtnl_link_ops ||
1764 err = ops->changelink(dev, tb, data);
1770 return do_setlink(dev, ifm, tb, ifname, modified);
1773 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
1774 if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
1775 return rtnl_group_changelink(net,
1776 nla_get_u32(tb[IFLA_GROUP]),
1783 if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
1787 #ifdef CONFIG_MODULES
1790 request_module("rtnl-link-%s", kind);
1792 ops = rtnl_link_ops_get(kind);
1801 snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
1803 dest_net = rtnl_link_get_net(net, tb);
1804 if (IS_ERR(dest_net))
1805 return PTR_ERR(dest_net);
1807 dev = rtnl_create_link(net, dest_net, ifname, ops, tb);
1811 else if (ops->newlink)
1812 err = ops->newlink(net, dev, tb, data);
1814 err = register_netdevice(dev);
1816 if (err < 0 && !IS_ERR(dev))
1821 err = rtnl_configure_link(dev, ifm);
1823 unregister_netdevice(dev);
1830 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
1832 struct net *net = sock_net(skb->sk);
1833 struct ifinfomsg *ifm;
1834 char ifname[IFNAMSIZ];
1835 struct nlattr *tb[IFLA_MAX+1];
1836 struct net_device *dev = NULL;
1837 struct sk_buff *nskb;
1840 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1844 if (tb[IFLA_IFNAME])
1845 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1847 ifm = nlmsg_data(nlh);
1848 if (ifm->ifi_index > 0)
1849 dev = __dev_get_by_index(net, ifm->ifi_index);
1850 else if (tb[IFLA_IFNAME])
1851 dev = __dev_get_by_name(net, ifname);
1858 nskb = nlmsg_new(if_nlmsg_size(dev), GFP_KERNEL);
1862 err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).pid,
1863 nlh->nlmsg_seq, 0, 0);
1865 /* -EMSGSIZE implies BUG in if_nlmsg_size */
1866 WARN_ON(err == -EMSGSIZE);
1869 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).pid);
1874 static u16 rtnl_calcit(struct sk_buff *skb)
1876 return min_ifinfo_dump_size;
1879 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
1882 int s_idx = cb->family;
1886 for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
1887 int type = cb->nlh->nlmsg_type-RTM_BASE;
1888 if (idx < s_idx || idx == PF_PACKET)
1890 if (rtnl_msg_handlers[idx] == NULL ||
1891 rtnl_msg_handlers[idx][type].dumpit == NULL)
1894 memset(&cb->args[0], 0, sizeof(cb->args));
1895 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
1903 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned change)
1905 struct net *net = dev_net(dev);
1906 struct sk_buff *skb;
1908 size_t if_info_size;
1910 skb = nlmsg_new((if_info_size = if_nlmsg_size(dev)), GFP_KERNEL);
1914 min_ifinfo_dump_size = max_t(u16, if_info_size, min_ifinfo_dump_size);
1916 err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0);
1918 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
1919 WARN_ON(err == -EMSGSIZE);
1923 rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_KERNEL);
1927 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
1930 /* Protected by RTNL sempahore. */
1931 static struct rtattr **rta_buf;
1932 static int rtattr_max;
1934 /* Process one rtnetlink message. */
1936 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
1938 struct net *net = sock_net(skb->sk);
1939 rtnl_doit_func doit;
1946 type = nlh->nlmsg_type;
1952 /* All the messages must have at least 1 byte length */
1953 if (nlh->nlmsg_len < NLMSG_LENGTH(sizeof(struct rtgenmsg)))
1956 family = ((struct rtgenmsg *)NLMSG_DATA(nlh))->rtgen_family;
1960 if (kind != 2 && security_netlink_recv(skb, CAP_NET_ADMIN))
1963 if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
1965 rtnl_dumpit_func dumpit;
1966 rtnl_calcit_func calcit;
1967 u16 min_dump_alloc = 0;
1969 dumpit = rtnl_get_dumpit(family, type);
1972 calcit = rtnl_get_calcit(family, type);
1974 min_dump_alloc = calcit(skb);
1978 err = netlink_dump_start(rtnl, skb, nlh, dumpit,
1979 NULL, min_dump_alloc);
1984 memset(rta_buf, 0, (rtattr_max * sizeof(struct rtattr *)));
1986 min_len = rtm_min[sz_idx];
1987 if (nlh->nlmsg_len < min_len)
1990 if (nlh->nlmsg_len > min_len) {
1991 int attrlen = nlh->nlmsg_len - NLMSG_ALIGN(min_len);
1992 struct rtattr *attr = (void *)nlh + NLMSG_ALIGN(min_len);
1994 while (RTA_OK(attr, attrlen)) {
1995 unsigned flavor = attr->rta_type;
1997 if (flavor > rta_max[sz_idx])
1999 rta_buf[flavor-1] = attr;
2001 attr = RTA_NEXT(attr, attrlen);
2005 doit = rtnl_get_doit(family, type);
2009 return doit(skb, nlh, (void *)&rta_buf[0]);
2012 static void rtnetlink_rcv(struct sk_buff *skb)
2015 netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
2019 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
2021 struct net_device *dev = ptr;
2027 case NETDEV_POST_INIT:
2028 case NETDEV_REGISTER:
2030 case NETDEV_PRE_TYPE_CHANGE:
2031 case NETDEV_GOING_DOWN:
2032 case NETDEV_UNREGISTER:
2033 case NETDEV_UNREGISTER_BATCH:
2034 case NETDEV_RELEASE:
2038 rtmsg_ifinfo(RTM_NEWLINK, dev, 0);
2044 static struct notifier_block rtnetlink_dev_notifier = {
2045 .notifier_call = rtnetlink_event,
2049 static int __net_init rtnetlink_net_init(struct net *net)
2052 sk = netlink_kernel_create(net, NETLINK_ROUTE, RTNLGRP_MAX,
2053 rtnetlink_rcv, &rtnl_mutex, THIS_MODULE);
2060 static void __net_exit rtnetlink_net_exit(struct net *net)
2062 netlink_kernel_release(net->rtnl);
2066 static struct pernet_operations rtnetlink_net_ops = {
2067 .init = rtnetlink_net_init,
2068 .exit = rtnetlink_net_exit,
2071 void __init rtnetlink_init(void)
2076 for (i = 0; i < ARRAY_SIZE(rta_max); i++)
2077 if (rta_max[i] > rtattr_max)
2078 rtattr_max = rta_max[i];
2079 rta_buf = kmalloc(rtattr_max * sizeof(struct rtattr *), GFP_KERNEL);
2081 panic("rtnetlink_init: cannot allocate rta_buf\n");
2083 if (register_pernet_subsys(&rtnetlink_net_ops))
2084 panic("rtnetlink_init: cannot initialize rtnetlink\n");
2086 netlink_set_nonroot(NETLINK_ROUTE, NL_NONROOT_RECV);
2087 register_netdevice_notifier(&rtnetlink_dev_notifier);
2089 rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
2090 rtnl_dump_ifinfo, rtnl_calcit);
2091 rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
2092 rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
2093 rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
2095 rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
2096 rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);