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 NLA_PUT_U32(skb, i+1, metrics[i]);
616 nla_nest_cancel(skb, mx);
620 return nla_nest_end(skb, mx);
623 nla_nest_cancel(skb, mx);
626 EXPORT_SYMBOL(rtnetlink_put_metrics);
628 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
629 u32 ts, u32 tsage, long expires, u32 error)
631 struct rta_cacheinfo ci = {
632 .rta_lastuse = jiffies_to_clock_t(jiffies - dst->lastuse),
633 .rta_used = dst->__use,
634 .rta_clntref = atomic_read(&(dst->__refcnt)),
642 ci.rta_expires = jiffies_to_clock_t(expires);
644 return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
646 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
648 static void set_operstate(struct net_device *dev, unsigned char transition)
650 unsigned char operstate = dev->operstate;
652 switch (transition) {
654 if ((operstate == IF_OPER_DORMANT ||
655 operstate == IF_OPER_UNKNOWN) &&
657 operstate = IF_OPER_UP;
660 case IF_OPER_DORMANT:
661 if (operstate == IF_OPER_UP ||
662 operstate == IF_OPER_UNKNOWN)
663 operstate = IF_OPER_DORMANT;
667 if (dev->operstate != operstate) {
668 write_lock_bh(&dev_base_lock);
669 dev->operstate = operstate;
670 write_unlock_bh(&dev_base_lock);
671 netdev_state_change(dev);
675 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
676 const struct ifinfomsg *ifm)
678 unsigned int flags = ifm->ifi_flags;
680 /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
682 flags = (flags & ifm->ifi_change) |
683 (dev->flags & ~ifm->ifi_change);
688 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
689 const struct rtnl_link_stats64 *b)
691 a->rx_packets = b->rx_packets;
692 a->tx_packets = b->tx_packets;
693 a->rx_bytes = b->rx_bytes;
694 a->tx_bytes = b->tx_bytes;
695 a->rx_errors = b->rx_errors;
696 a->tx_errors = b->tx_errors;
697 a->rx_dropped = b->rx_dropped;
698 a->tx_dropped = b->tx_dropped;
700 a->multicast = b->multicast;
701 a->collisions = b->collisions;
703 a->rx_length_errors = b->rx_length_errors;
704 a->rx_over_errors = b->rx_over_errors;
705 a->rx_crc_errors = b->rx_crc_errors;
706 a->rx_frame_errors = b->rx_frame_errors;
707 a->rx_fifo_errors = b->rx_fifo_errors;
708 a->rx_missed_errors = b->rx_missed_errors;
710 a->tx_aborted_errors = b->tx_aborted_errors;
711 a->tx_carrier_errors = b->tx_carrier_errors;
712 a->tx_fifo_errors = b->tx_fifo_errors;
713 a->tx_heartbeat_errors = b->tx_heartbeat_errors;
714 a->tx_window_errors = b->tx_window_errors;
716 a->rx_compressed = b->rx_compressed;
717 a->tx_compressed = b->tx_compressed;
720 static void copy_rtnl_link_stats64(void *v, const struct rtnl_link_stats64 *b)
722 memcpy(v, b, sizeof(*b));
726 static inline int rtnl_vfinfo_size(const struct net_device *dev,
729 if (dev->dev.parent && dev_is_pci(dev->dev.parent) &&
730 (ext_filter_mask & RTEXT_FILTER_VF)) {
731 int num_vfs = dev_num_vf(dev->dev.parent);
732 size_t size = nla_total_size(sizeof(struct nlattr));
733 size += nla_total_size(num_vfs * sizeof(struct nlattr));
735 (nla_total_size(sizeof(struct ifla_vf_mac)) +
736 nla_total_size(sizeof(struct ifla_vf_vlan)) +
737 nla_total_size(sizeof(struct ifla_vf_tx_rate)) +
738 nla_total_size(sizeof(struct ifla_vf_spoofchk)));
744 static size_t rtnl_port_size(const struct net_device *dev)
746 size_t port_size = nla_total_size(4) /* PORT_VF */
747 + nla_total_size(PORT_PROFILE_MAX) /* PORT_PROFILE */
748 + nla_total_size(sizeof(struct ifla_port_vsi))
750 + nla_total_size(PORT_UUID_MAX) /* PORT_INSTANCE_UUID */
751 + nla_total_size(PORT_UUID_MAX) /* PORT_HOST_UUID */
752 + nla_total_size(1) /* PROT_VDP_REQUEST */
753 + nla_total_size(2); /* PORT_VDP_RESPONSE */
754 size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
755 size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
757 size_t port_self_size = nla_total_size(sizeof(struct nlattr))
760 if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent)
762 if (dev_num_vf(dev->dev.parent))
763 return port_self_size + vf_ports_size +
764 vf_port_size * dev_num_vf(dev->dev.parent);
766 return port_self_size;
769 static noinline size_t if_nlmsg_size(const struct net_device *dev,
772 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
773 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
774 + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
775 + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
776 + nla_total_size(sizeof(struct rtnl_link_ifmap))
777 + nla_total_size(sizeof(struct rtnl_link_stats))
778 + nla_total_size(sizeof(struct rtnl_link_stats64))
779 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
780 + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
781 + nla_total_size(4) /* IFLA_TXQLEN */
782 + nla_total_size(4) /* IFLA_WEIGHT */
783 + nla_total_size(4) /* IFLA_MTU */
784 + nla_total_size(4) /* IFLA_LINK */
785 + nla_total_size(4) /* IFLA_MASTER */
786 + nla_total_size(1) /* IFLA_OPERSTATE */
787 + nla_total_size(1) /* IFLA_LINKMODE */
788 + nla_total_size(ext_filter_mask
789 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
790 + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
791 + rtnl_port_size(dev) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
792 + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
793 + rtnl_link_get_af_size(dev); /* IFLA_AF_SPEC */
796 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
798 struct nlattr *vf_ports;
799 struct nlattr *vf_port;
803 vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
807 for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
808 vf_port = nla_nest_start(skb, IFLA_VF_PORT);
810 goto nla_put_failure;
811 NLA_PUT_U32(skb, IFLA_PORT_VF, vf);
812 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
813 if (err == -EMSGSIZE)
814 goto nla_put_failure;
816 nla_nest_cancel(skb, vf_port);
819 nla_nest_end(skb, vf_port);
822 nla_nest_end(skb, vf_ports);
827 nla_nest_cancel(skb, vf_ports);
831 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
833 struct nlattr *port_self;
836 port_self = nla_nest_start(skb, IFLA_PORT_SELF);
840 err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
842 nla_nest_cancel(skb, port_self);
843 return (err == -EMSGSIZE) ? err : 0;
846 nla_nest_end(skb, port_self);
851 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev)
855 if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent)
858 err = rtnl_port_self_fill(skb, dev);
862 if (dev_num_vf(dev->dev.parent)) {
863 err = rtnl_vf_ports_fill(skb, dev);
871 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
872 int type, u32 pid, u32 seq, u32 change,
873 unsigned int flags, u32 ext_filter_mask)
875 struct ifinfomsg *ifm;
876 struct nlmsghdr *nlh;
877 struct rtnl_link_stats64 temp;
878 const struct rtnl_link_stats64 *stats;
879 struct nlattr *attr, *af_spec;
880 struct rtnl_af_ops *af_ops;
883 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
887 ifm = nlmsg_data(nlh);
888 ifm->ifi_family = AF_UNSPEC;
890 ifm->ifi_type = dev->type;
891 ifm->ifi_index = dev->ifindex;
892 ifm->ifi_flags = dev_get_flags(dev);
893 ifm->ifi_change = change;
895 NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
896 NLA_PUT_U32(skb, IFLA_TXQLEN, dev->tx_queue_len);
897 NLA_PUT_U8(skb, IFLA_OPERSTATE,
898 netif_running(dev) ? dev->operstate : IF_OPER_DOWN);
899 NLA_PUT_U8(skb, IFLA_LINKMODE, dev->link_mode);
900 NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
901 NLA_PUT_U32(skb, IFLA_GROUP, dev->group);
903 if (dev->ifindex != dev->iflink)
904 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
907 NLA_PUT_U32(skb, IFLA_MASTER, dev->master->ifindex);
910 NLA_PUT_STRING(skb, IFLA_QDISC, dev->qdisc->ops->id);
913 NLA_PUT_STRING(skb, IFLA_IFALIAS, dev->ifalias);
916 struct rtnl_link_ifmap map = {
917 .mem_start = dev->mem_start,
918 .mem_end = dev->mem_end,
919 .base_addr = dev->base_addr,
922 .port = dev->if_port,
924 NLA_PUT(skb, IFLA_MAP, sizeof(map), &map);
928 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
929 NLA_PUT(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast);
932 attr = nla_reserve(skb, IFLA_STATS,
933 sizeof(struct rtnl_link_stats));
935 goto nla_put_failure;
937 stats = dev_get_stats(dev, &temp);
938 copy_rtnl_link_stats(nla_data(attr), stats);
940 attr = nla_reserve(skb, IFLA_STATS64,
941 sizeof(struct rtnl_link_stats64));
943 goto nla_put_failure;
944 copy_rtnl_link_stats64(nla_data(attr), stats);
946 if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF))
947 NLA_PUT_U32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent));
949 if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent
950 && (ext_filter_mask & RTEXT_FILTER_VF)) {
953 struct nlattr *vfinfo, *vf;
954 int num_vfs = dev_num_vf(dev->dev.parent);
956 vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
958 goto nla_put_failure;
959 for (i = 0; i < num_vfs; i++) {
960 struct ifla_vf_info ivi;
961 struct ifla_vf_mac vf_mac;
962 struct ifla_vf_vlan vf_vlan;
963 struct ifla_vf_tx_rate vf_tx_rate;
964 struct ifla_vf_spoofchk vf_spoofchk;
967 * Not all SR-IOV capable drivers support the
968 * spoofcheck query. Preset to -1 so the user
969 * space tool can detect that the driver didn't
973 if (dev->netdev_ops->ndo_get_vf_config(dev, i, &ivi))
978 vf_spoofchk.vf = ivi.vf;
980 memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
981 vf_vlan.vlan = ivi.vlan;
982 vf_vlan.qos = ivi.qos;
983 vf_tx_rate.rate = ivi.tx_rate;
984 vf_spoofchk.setting = ivi.spoofchk;
985 vf = nla_nest_start(skb, IFLA_VF_INFO);
987 nla_nest_cancel(skb, vfinfo);
988 goto nla_put_failure;
990 NLA_PUT(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac);
991 NLA_PUT(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan);
992 NLA_PUT(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
994 NLA_PUT(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
996 nla_nest_end(skb, vf);
998 nla_nest_end(skb, vfinfo);
1001 if (rtnl_port_fill(skb, dev))
1002 goto nla_put_failure;
1004 if (dev->rtnl_link_ops) {
1005 if (rtnl_link_fill(skb, dev) < 0)
1006 goto nla_put_failure;
1009 if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1010 goto nla_put_failure;
1012 list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1013 if (af_ops->fill_link_af) {
1017 if (!(af = nla_nest_start(skb, af_ops->family)))
1018 goto nla_put_failure;
1020 err = af_ops->fill_link_af(skb, dev);
1023 * Caller may return ENODATA to indicate that there
1024 * was no data to be dumped. This is not an error, it
1025 * means we should trim the attribute header and
1028 if (err == -ENODATA)
1029 nla_nest_cancel(skb, af);
1031 goto nla_put_failure;
1033 nla_nest_end(skb, af);
1037 nla_nest_end(skb, af_spec);
1039 return nlmsg_end(skb, nlh);
1042 nlmsg_cancel(skb, nlh);
1046 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1048 struct net *net = sock_net(skb->sk);
1051 struct net_device *dev;
1052 struct hlist_head *head;
1053 struct hlist_node *node;
1054 struct nlattr *tb[IFLA_MAX+1];
1055 u32 ext_filter_mask = 0;
1058 s_idx = cb->args[1];
1061 cb->seq = net->dev_base_seq;
1063 if (nlmsg_parse(cb->nlh, sizeof(struct rtgenmsg), tb, IFLA_MAX,
1064 ifla_policy) >= 0) {
1066 if (tb[IFLA_EXT_MASK])
1067 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1070 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1072 head = &net->dev_index_head[h];
1073 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
1076 if (rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1077 NETLINK_CB(cb->skb).pid,
1078 cb->nlh->nlmsg_seq, 0,
1080 ext_filter_mask) <= 0)
1083 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1096 const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1097 [IFLA_IFNAME] = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1098 [IFLA_ADDRESS] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1099 [IFLA_BROADCAST] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1100 [IFLA_MAP] = { .len = sizeof(struct rtnl_link_ifmap) },
1101 [IFLA_MTU] = { .type = NLA_U32 },
1102 [IFLA_LINK] = { .type = NLA_U32 },
1103 [IFLA_MASTER] = { .type = NLA_U32 },
1104 [IFLA_TXQLEN] = { .type = NLA_U32 },
1105 [IFLA_WEIGHT] = { .type = NLA_U32 },
1106 [IFLA_OPERSTATE] = { .type = NLA_U8 },
1107 [IFLA_LINKMODE] = { .type = NLA_U8 },
1108 [IFLA_LINKINFO] = { .type = NLA_NESTED },
1109 [IFLA_NET_NS_PID] = { .type = NLA_U32 },
1110 [IFLA_NET_NS_FD] = { .type = NLA_U32 },
1111 [IFLA_IFALIAS] = { .type = NLA_STRING, .len = IFALIASZ-1 },
1112 [IFLA_VFINFO_LIST] = {. type = NLA_NESTED },
1113 [IFLA_VF_PORTS] = { .type = NLA_NESTED },
1114 [IFLA_PORT_SELF] = { .type = NLA_NESTED },
1115 [IFLA_AF_SPEC] = { .type = NLA_NESTED },
1116 [IFLA_EXT_MASK] = { .type = NLA_U32 },
1118 EXPORT_SYMBOL(ifla_policy);
1120 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1121 [IFLA_INFO_KIND] = { .type = NLA_STRING },
1122 [IFLA_INFO_DATA] = { .type = NLA_NESTED },
1125 static const struct nla_policy ifla_vfinfo_policy[IFLA_VF_INFO_MAX+1] = {
1126 [IFLA_VF_INFO] = { .type = NLA_NESTED },
1129 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1130 [IFLA_VF_MAC] = { .type = NLA_BINARY,
1131 .len = sizeof(struct ifla_vf_mac) },
1132 [IFLA_VF_VLAN] = { .type = NLA_BINARY,
1133 .len = sizeof(struct ifla_vf_vlan) },
1134 [IFLA_VF_TX_RATE] = { .type = NLA_BINARY,
1135 .len = sizeof(struct ifla_vf_tx_rate) },
1136 [IFLA_VF_SPOOFCHK] = { .type = NLA_BINARY,
1137 .len = sizeof(struct ifla_vf_spoofchk) },
1140 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1141 [IFLA_PORT_VF] = { .type = NLA_U32 },
1142 [IFLA_PORT_PROFILE] = { .type = NLA_STRING,
1143 .len = PORT_PROFILE_MAX },
1144 [IFLA_PORT_VSI_TYPE] = { .type = NLA_BINARY,
1145 .len = sizeof(struct ifla_port_vsi)},
1146 [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1147 .len = PORT_UUID_MAX },
1148 [IFLA_PORT_HOST_UUID] = { .type = NLA_STRING,
1149 .len = PORT_UUID_MAX },
1150 [IFLA_PORT_REQUEST] = { .type = NLA_U8, },
1151 [IFLA_PORT_RESPONSE] = { .type = NLA_U16, },
1154 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1157 /* Examine the link attributes and figure out which
1158 * network namespace we are talking about.
1160 if (tb[IFLA_NET_NS_PID])
1161 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1162 else if (tb[IFLA_NET_NS_FD])
1163 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1165 net = get_net(src_net);
1168 EXPORT_SYMBOL(rtnl_link_get_net);
1170 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1173 if (tb[IFLA_ADDRESS] &&
1174 nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1177 if (tb[IFLA_BROADCAST] &&
1178 nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1182 if (tb[IFLA_AF_SPEC]) {
1186 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1187 const struct rtnl_af_ops *af_ops;
1189 if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1190 return -EAFNOSUPPORT;
1192 if (!af_ops->set_link_af)
1195 if (af_ops->validate_link_af) {
1196 err = af_ops->validate_link_af(dev, af);
1206 static int do_setvfinfo(struct net_device *dev, struct nlattr *attr)
1208 int rem, err = -EINVAL;
1210 const struct net_device_ops *ops = dev->netdev_ops;
1212 nla_for_each_nested(vf, attr, rem) {
1213 switch (nla_type(vf)) {
1215 struct ifla_vf_mac *ivm;
1218 if (ops->ndo_set_vf_mac)
1219 err = ops->ndo_set_vf_mac(dev, ivm->vf,
1223 case IFLA_VF_VLAN: {
1224 struct ifla_vf_vlan *ivv;
1227 if (ops->ndo_set_vf_vlan)
1228 err = ops->ndo_set_vf_vlan(dev, ivv->vf,
1233 case IFLA_VF_TX_RATE: {
1234 struct ifla_vf_tx_rate *ivt;
1237 if (ops->ndo_set_vf_tx_rate)
1238 err = ops->ndo_set_vf_tx_rate(dev, ivt->vf,
1242 case IFLA_VF_SPOOFCHK: {
1243 struct ifla_vf_spoofchk *ivs;
1246 if (ops->ndo_set_vf_spoofchk)
1247 err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1261 static int do_set_master(struct net_device *dev, int ifindex)
1263 struct net_device *master_dev;
1264 const struct net_device_ops *ops;
1268 if (dev->master->ifindex == ifindex)
1270 ops = dev->master->netdev_ops;
1271 if (ops->ndo_del_slave) {
1272 err = ops->ndo_del_slave(dev->master, dev);
1281 master_dev = __dev_get_by_index(dev_net(dev), ifindex);
1284 ops = master_dev->netdev_ops;
1285 if (ops->ndo_add_slave) {
1286 err = ops->ndo_add_slave(master_dev, dev);
1296 static int do_setlink(struct net_device *dev, struct ifinfomsg *ifm,
1297 struct nlattr **tb, char *ifname, int modified)
1299 const struct net_device_ops *ops = dev->netdev_ops;
1300 int send_addr_notify = 0;
1303 if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1304 struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1309 err = dev_change_net_namespace(dev, net, ifname);
1317 struct rtnl_link_ifmap *u_map;
1320 if (!ops->ndo_set_config) {
1325 if (!netif_device_present(dev)) {
1330 u_map = nla_data(tb[IFLA_MAP]);
1331 k_map.mem_start = (unsigned long) u_map->mem_start;
1332 k_map.mem_end = (unsigned long) u_map->mem_end;
1333 k_map.base_addr = (unsigned short) u_map->base_addr;
1334 k_map.irq = (unsigned char) u_map->irq;
1335 k_map.dma = (unsigned char) u_map->dma;
1336 k_map.port = (unsigned char) u_map->port;
1338 err = ops->ndo_set_config(dev, &k_map);
1345 if (tb[IFLA_ADDRESS]) {
1346 struct sockaddr *sa;
1349 if (!ops->ndo_set_mac_address) {
1354 if (!netif_device_present(dev)) {
1359 len = sizeof(sa_family_t) + dev->addr_len;
1360 sa = kmalloc(len, GFP_KERNEL);
1365 sa->sa_family = dev->type;
1366 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1368 err = ops->ndo_set_mac_address(dev, sa);
1372 send_addr_notify = 1;
1377 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1383 if (tb[IFLA_GROUP]) {
1384 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1389 * Interface selected by interface index but interface
1390 * name provided implies that a name change has been
1393 if (ifm->ifi_index > 0 && ifname[0]) {
1394 err = dev_change_name(dev, ifname);
1400 if (tb[IFLA_IFALIAS]) {
1401 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
1402 nla_len(tb[IFLA_IFALIAS]));
1408 if (tb[IFLA_BROADCAST]) {
1409 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
1410 send_addr_notify = 1;
1413 if (ifm->ifi_flags || ifm->ifi_change) {
1414 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1419 if (tb[IFLA_MASTER]) {
1420 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
1426 if (tb[IFLA_TXQLEN])
1427 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1429 if (tb[IFLA_OPERSTATE])
1430 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1432 if (tb[IFLA_LINKMODE]) {
1433 write_lock_bh(&dev_base_lock);
1434 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1435 write_unlock_bh(&dev_base_lock);
1438 if (tb[IFLA_VFINFO_LIST]) {
1439 struct nlattr *attr;
1441 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
1442 if (nla_type(attr) != IFLA_VF_INFO) {
1446 err = do_setvfinfo(dev, attr);
1454 if (tb[IFLA_VF_PORTS]) {
1455 struct nlattr *port[IFLA_PORT_MAX+1];
1456 struct nlattr *attr;
1461 if (!ops->ndo_set_vf_port)
1464 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
1465 if (nla_type(attr) != IFLA_VF_PORT)
1467 err = nla_parse_nested(port, IFLA_PORT_MAX,
1468 attr, ifla_port_policy);
1471 if (!port[IFLA_PORT_VF]) {
1475 vf = nla_get_u32(port[IFLA_PORT_VF]);
1476 err = ops->ndo_set_vf_port(dev, vf, port);
1484 if (tb[IFLA_PORT_SELF]) {
1485 struct nlattr *port[IFLA_PORT_MAX+1];
1487 err = nla_parse_nested(port, IFLA_PORT_MAX,
1488 tb[IFLA_PORT_SELF], ifla_port_policy);
1493 if (ops->ndo_set_vf_port)
1494 err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
1500 if (tb[IFLA_AF_SPEC]) {
1504 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1505 const struct rtnl_af_ops *af_ops;
1507 if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1510 err = af_ops->set_link_af(dev, af);
1520 if (err < 0 && modified && net_ratelimit())
1521 printk(KERN_WARNING "A link change request failed with "
1522 "some changes committed already. Interface %s may "
1523 "have been left with an inconsistent configuration, "
1524 "please check.\n", dev->name);
1526 if (send_addr_notify)
1527 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
1532 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1534 struct net *net = sock_net(skb->sk);
1535 struct ifinfomsg *ifm;
1536 struct net_device *dev;
1538 struct nlattr *tb[IFLA_MAX+1];
1539 char ifname[IFNAMSIZ];
1541 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1545 if (tb[IFLA_IFNAME])
1546 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1551 ifm = nlmsg_data(nlh);
1552 if (ifm->ifi_index > 0)
1553 dev = __dev_get_by_index(net, ifm->ifi_index);
1554 else if (tb[IFLA_IFNAME])
1555 dev = __dev_get_by_name(net, ifname);
1564 err = validate_linkmsg(dev, tb);
1568 err = do_setlink(dev, ifm, tb, ifname, 0);
1573 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1575 struct net *net = sock_net(skb->sk);
1576 const struct rtnl_link_ops *ops;
1577 struct net_device *dev;
1578 struct ifinfomsg *ifm;
1579 char ifname[IFNAMSIZ];
1580 struct nlattr *tb[IFLA_MAX+1];
1582 LIST_HEAD(list_kill);
1584 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1588 if (tb[IFLA_IFNAME])
1589 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1591 ifm = nlmsg_data(nlh);
1592 if (ifm->ifi_index > 0)
1593 dev = __dev_get_by_index(net, ifm->ifi_index);
1594 else if (tb[IFLA_IFNAME])
1595 dev = __dev_get_by_name(net, ifname);
1602 ops = dev->rtnl_link_ops;
1606 ops->dellink(dev, &list_kill);
1607 unregister_netdevice_many(&list_kill);
1608 list_del(&list_kill);
1612 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
1614 unsigned int old_flags;
1617 old_flags = dev->flags;
1618 if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
1619 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1624 dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
1625 rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U);
1627 __dev_notify_flags(dev, old_flags);
1630 EXPORT_SYMBOL(rtnl_configure_link);
1632 struct net_device *rtnl_create_link(struct net *src_net, struct net *net,
1633 char *ifname, const struct rtnl_link_ops *ops, struct nlattr *tb[])
1636 struct net_device *dev;
1637 unsigned int num_queues = 1;
1638 unsigned int real_num_queues = 1;
1640 if (ops->get_tx_queues) {
1641 err = ops->get_tx_queues(src_net, tb, &num_queues,
1647 dev = alloc_netdev_mq(ops->priv_size, ifname, ops->setup, num_queues);
1651 dev_net_set(dev, net);
1652 dev->rtnl_link_ops = ops;
1653 dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
1656 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
1657 if (tb[IFLA_ADDRESS])
1658 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
1659 nla_len(tb[IFLA_ADDRESS]));
1660 if (tb[IFLA_BROADCAST])
1661 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
1662 nla_len(tb[IFLA_BROADCAST]));
1663 if (tb[IFLA_TXQLEN])
1664 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1665 if (tb[IFLA_OPERSTATE])
1666 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1667 if (tb[IFLA_LINKMODE])
1668 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1670 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1675 return ERR_PTR(err);
1677 EXPORT_SYMBOL(rtnl_create_link);
1679 static int rtnl_group_changelink(struct net *net, int group,
1680 struct ifinfomsg *ifm,
1683 struct net_device *dev;
1686 for_each_netdev(net, dev) {
1687 if (dev->group == group) {
1688 err = do_setlink(dev, ifm, tb, NULL, 0);
1697 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1699 struct net *net = sock_net(skb->sk);
1700 const struct rtnl_link_ops *ops;
1701 struct net_device *dev;
1702 struct ifinfomsg *ifm;
1703 char kind[MODULE_NAME_LEN];
1704 char ifname[IFNAMSIZ];
1705 struct nlattr *tb[IFLA_MAX+1];
1706 struct nlattr *linkinfo[IFLA_INFO_MAX+1];
1709 #ifdef CONFIG_MODULES
1712 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1716 if (tb[IFLA_IFNAME])
1717 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1721 ifm = nlmsg_data(nlh);
1722 if (ifm->ifi_index > 0)
1723 dev = __dev_get_by_index(net, ifm->ifi_index);
1726 dev = __dev_get_by_name(net, ifname);
1731 err = validate_linkmsg(dev, tb);
1735 if (tb[IFLA_LINKINFO]) {
1736 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
1737 tb[IFLA_LINKINFO], ifla_info_policy);
1741 memset(linkinfo, 0, sizeof(linkinfo));
1743 if (linkinfo[IFLA_INFO_KIND]) {
1744 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
1745 ops = rtnl_link_ops_get(kind);
1752 struct nlattr *attr[ops ? ops->maxtype + 1 : 0], **data = NULL;
1753 struct net *dest_net;
1756 if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
1757 err = nla_parse_nested(attr, ops->maxtype,
1758 linkinfo[IFLA_INFO_DATA],
1764 if (ops->validate) {
1765 err = ops->validate(tb, data);
1774 if (nlh->nlmsg_flags & NLM_F_EXCL)
1776 if (nlh->nlmsg_flags & NLM_F_REPLACE)
1779 if (linkinfo[IFLA_INFO_DATA]) {
1780 if (!ops || ops != dev->rtnl_link_ops ||
1784 err = ops->changelink(dev, tb, data);
1790 return do_setlink(dev, ifm, tb, ifname, modified);
1793 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
1794 if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
1795 return rtnl_group_changelink(net,
1796 nla_get_u32(tb[IFLA_GROUP]),
1803 if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
1807 #ifdef CONFIG_MODULES
1810 request_module("rtnl-link-%s", kind);
1812 ops = rtnl_link_ops_get(kind);
1821 snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
1823 dest_net = rtnl_link_get_net(net, tb);
1824 if (IS_ERR(dest_net))
1825 return PTR_ERR(dest_net);
1827 dev = rtnl_create_link(net, dest_net, ifname, ops, tb);
1831 else if (ops->newlink)
1832 err = ops->newlink(net, dev, tb, data);
1834 err = register_netdevice(dev);
1836 if (err < 0 && !IS_ERR(dev))
1841 err = rtnl_configure_link(dev, ifm);
1843 unregister_netdevice(dev);
1850 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
1852 struct net *net = sock_net(skb->sk);
1853 struct ifinfomsg *ifm;
1854 char ifname[IFNAMSIZ];
1855 struct nlattr *tb[IFLA_MAX+1];
1856 struct net_device *dev = NULL;
1857 struct sk_buff *nskb;
1859 u32 ext_filter_mask = 0;
1861 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1865 if (tb[IFLA_IFNAME])
1866 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1868 if (tb[IFLA_EXT_MASK])
1869 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1871 ifm = nlmsg_data(nlh);
1872 if (ifm->ifi_index > 0)
1873 dev = __dev_get_by_index(net, ifm->ifi_index);
1874 else if (tb[IFLA_IFNAME])
1875 dev = __dev_get_by_name(net, ifname);
1882 nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
1886 err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).pid,
1887 nlh->nlmsg_seq, 0, 0, ext_filter_mask);
1889 /* -EMSGSIZE implies BUG in if_nlmsg_size */
1890 WARN_ON(err == -EMSGSIZE);
1893 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).pid);
1898 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
1900 struct net *net = sock_net(skb->sk);
1901 struct net_device *dev;
1902 struct nlattr *tb[IFLA_MAX+1];
1903 u32 ext_filter_mask = 0;
1904 u16 min_ifinfo_dump_size = 0;
1906 if (nlmsg_parse(nlh, sizeof(struct rtgenmsg), tb, IFLA_MAX,
1907 ifla_policy) >= 0) {
1908 if (tb[IFLA_EXT_MASK])
1909 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1912 if (!ext_filter_mask)
1913 return NLMSG_GOODSIZE;
1915 * traverse the list of net devices and compute the minimum
1916 * buffer size based upon the filter mask.
1918 list_for_each_entry(dev, &net->dev_base_head, dev_list) {
1919 min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
1924 return min_ifinfo_dump_size;
1927 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
1930 int s_idx = cb->family;
1934 for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
1935 int type = cb->nlh->nlmsg_type-RTM_BASE;
1936 if (idx < s_idx || idx == PF_PACKET)
1938 if (rtnl_msg_handlers[idx] == NULL ||
1939 rtnl_msg_handlers[idx][type].dumpit == NULL)
1942 memset(&cb->args[0], 0, sizeof(cb->args));
1943 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
1951 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned change)
1953 struct net *net = dev_net(dev);
1954 struct sk_buff *skb;
1956 size_t if_info_size;
1958 skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), GFP_KERNEL);
1962 err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0);
1964 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
1965 WARN_ON(err == -EMSGSIZE);
1969 rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_KERNEL);
1973 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
1976 /* Protected by RTNL sempahore. */
1977 static struct rtattr **rta_buf;
1978 static int rtattr_max;
1980 /* Process one rtnetlink message. */
1982 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
1984 struct net *net = sock_net(skb->sk);
1985 rtnl_doit_func doit;
1992 type = nlh->nlmsg_type;
1998 /* All the messages must have at least 1 byte length */
1999 if (nlh->nlmsg_len < NLMSG_LENGTH(sizeof(struct rtgenmsg)))
2002 family = ((struct rtgenmsg *)NLMSG_DATA(nlh))->rtgen_family;
2006 if (kind != 2 && !capable(CAP_NET_ADMIN))
2009 if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
2011 rtnl_dumpit_func dumpit;
2012 rtnl_calcit_func calcit;
2013 u16 min_dump_alloc = 0;
2015 dumpit = rtnl_get_dumpit(family, type);
2018 calcit = rtnl_get_calcit(family, type);
2020 min_dump_alloc = calcit(skb, nlh);
2025 struct netlink_dump_control c = {
2027 .min_dump_alloc = min_dump_alloc,
2029 err = netlink_dump_start(rtnl, skb, nlh, &c);
2035 memset(rta_buf, 0, (rtattr_max * sizeof(struct rtattr *)));
2037 min_len = rtm_min[sz_idx];
2038 if (nlh->nlmsg_len < min_len)
2041 if (nlh->nlmsg_len > min_len) {
2042 int attrlen = nlh->nlmsg_len - NLMSG_ALIGN(min_len);
2043 struct rtattr *attr = (void *)nlh + NLMSG_ALIGN(min_len);
2045 while (RTA_OK(attr, attrlen)) {
2046 unsigned flavor = attr->rta_type;
2048 if (flavor > rta_max[sz_idx])
2050 rta_buf[flavor-1] = attr;
2052 attr = RTA_NEXT(attr, attrlen);
2056 doit = rtnl_get_doit(family, type);
2060 return doit(skb, nlh, (void *)&rta_buf[0]);
2063 static void rtnetlink_rcv(struct sk_buff *skb)
2066 netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
2070 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
2072 struct net_device *dev = ptr;
2078 case NETDEV_POST_INIT:
2079 case NETDEV_REGISTER:
2081 case NETDEV_PRE_TYPE_CHANGE:
2082 case NETDEV_GOING_DOWN:
2083 case NETDEV_UNREGISTER:
2084 case NETDEV_UNREGISTER_BATCH:
2085 case NETDEV_RELEASE:
2089 rtmsg_ifinfo(RTM_NEWLINK, dev, 0);
2095 static struct notifier_block rtnetlink_dev_notifier = {
2096 .notifier_call = rtnetlink_event,
2100 static int __net_init rtnetlink_net_init(struct net *net)
2103 sk = netlink_kernel_create(net, NETLINK_ROUTE, RTNLGRP_MAX,
2104 rtnetlink_rcv, &rtnl_mutex, THIS_MODULE);
2111 static void __net_exit rtnetlink_net_exit(struct net *net)
2113 netlink_kernel_release(net->rtnl);
2117 static struct pernet_operations rtnetlink_net_ops = {
2118 .init = rtnetlink_net_init,
2119 .exit = rtnetlink_net_exit,
2122 void __init rtnetlink_init(void)
2127 for (i = 0; i < ARRAY_SIZE(rta_max); i++)
2128 if (rta_max[i] > rtattr_max)
2129 rtattr_max = rta_max[i];
2130 rta_buf = kmalloc(rtattr_max * sizeof(struct rtattr *), GFP_KERNEL);
2132 panic("rtnetlink_init: cannot allocate rta_buf\n");
2134 if (register_pernet_subsys(&rtnetlink_net_ops))
2135 panic("rtnetlink_init: cannot initialize rtnetlink\n");
2137 netlink_set_nonroot(NETLINK_ROUTE, NL_NONROOT_RECV);
2138 register_netdevice_notifier(&rtnetlink_dev_notifier);
2140 rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
2141 rtnl_dump_ifinfo, rtnl_calcit);
2142 rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
2143 rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
2144 rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
2146 rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
2147 rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);