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>
42 #include <linux/inet.h>
43 #include <linux/netdevice.h>
45 #include <net/protocol.h>
47 #include <net/route.h>
50 #include <net/pkt_sched.h>
51 #include <net/fib_rules.h>
52 #include <net/rtnetlink.h>
53 #include <net/net_namespace.h>
57 rtnl_dumpit_func dumpit;
58 rtnl_calcit_func calcit;
61 static DEFINE_MUTEX(rtnl_mutex);
65 mutex_lock(&rtnl_mutex);
67 EXPORT_SYMBOL(rtnl_lock);
69 void __rtnl_unlock(void)
71 mutex_unlock(&rtnl_mutex);
74 void rtnl_unlock(void)
76 /* This fellow will unlock it for us. */
79 EXPORT_SYMBOL(rtnl_unlock);
81 int rtnl_trylock(void)
83 return mutex_trylock(&rtnl_mutex);
85 EXPORT_SYMBOL(rtnl_trylock);
87 int rtnl_is_locked(void)
89 return mutex_is_locked(&rtnl_mutex);
91 EXPORT_SYMBOL(rtnl_is_locked);
93 #ifdef CONFIG_PROVE_LOCKING
94 int lockdep_rtnl_is_held(void)
96 return lockdep_is_held(&rtnl_mutex);
98 EXPORT_SYMBOL(lockdep_rtnl_is_held);
99 #endif /* #ifdef CONFIG_PROVE_LOCKING */
101 static struct rtnl_link *rtnl_msg_handlers[RTNL_FAMILY_MAX + 1];
103 static inline int rtm_msgindex(int msgtype)
105 int msgindex = msgtype - RTM_BASE;
108 * msgindex < 0 implies someone tried to register a netlink
109 * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
110 * the message type has not been added to linux/rtnetlink.h
112 BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
117 static rtnl_doit_func rtnl_get_doit(int protocol, int msgindex)
119 struct rtnl_link *tab;
121 if (protocol <= RTNL_FAMILY_MAX)
122 tab = rtnl_msg_handlers[protocol];
126 if (tab == NULL || tab[msgindex].doit == NULL)
127 tab = rtnl_msg_handlers[PF_UNSPEC];
129 return tab ? tab[msgindex].doit : NULL;
132 static rtnl_dumpit_func rtnl_get_dumpit(int protocol, int msgindex)
134 struct rtnl_link *tab;
136 if (protocol <= RTNL_FAMILY_MAX)
137 tab = rtnl_msg_handlers[protocol];
141 if (tab == NULL || tab[msgindex].dumpit == NULL)
142 tab = rtnl_msg_handlers[PF_UNSPEC];
144 return tab ? tab[msgindex].dumpit : NULL;
147 static rtnl_calcit_func rtnl_get_calcit(int protocol, int msgindex)
149 struct rtnl_link *tab;
151 if (protocol <= RTNL_FAMILY_MAX)
152 tab = rtnl_msg_handlers[protocol];
156 if (tab == NULL || tab[msgindex].calcit == NULL)
157 tab = rtnl_msg_handlers[PF_UNSPEC];
159 return tab ? tab[msgindex].calcit : NULL;
163 * __rtnl_register - Register a rtnetlink message type
164 * @protocol: Protocol family or PF_UNSPEC
165 * @msgtype: rtnetlink message type
166 * @doit: Function pointer called for each request message
167 * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
168 * @calcit: Function pointer to calc size of dump message
170 * Registers the specified function pointers (at least one of them has
171 * to be non-NULL) to be called whenever a request message for the
172 * specified protocol family and message type is received.
174 * The special protocol family PF_UNSPEC may be used to define fallback
175 * function pointers for the case when no entry for the specific protocol
178 * Returns 0 on success or a negative error code.
180 int __rtnl_register(int protocol, int msgtype,
181 rtnl_doit_func doit, rtnl_dumpit_func dumpit,
182 rtnl_calcit_func calcit)
184 struct rtnl_link *tab;
187 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
188 msgindex = rtm_msgindex(msgtype);
190 tab = rtnl_msg_handlers[protocol];
192 tab = kcalloc(RTM_NR_MSGTYPES, sizeof(*tab), GFP_KERNEL);
196 rtnl_msg_handlers[protocol] = tab;
200 tab[msgindex].doit = doit;
203 tab[msgindex].dumpit = dumpit;
206 tab[msgindex].calcit = calcit;
210 EXPORT_SYMBOL_GPL(__rtnl_register);
213 * rtnl_register - Register a rtnetlink message type
215 * Identical to __rtnl_register() but panics on failure. This is useful
216 * as failure of this function is very unlikely, it can only happen due
217 * to lack of memory when allocating the chain to store all message
218 * handlers for a protocol. Meant for use in init functions where lack
219 * of memory implies no sense in continuing.
221 void rtnl_register(int protocol, int msgtype,
222 rtnl_doit_func doit, rtnl_dumpit_func dumpit,
223 rtnl_calcit_func calcit)
225 if (__rtnl_register(protocol, msgtype, doit, dumpit, calcit) < 0)
226 panic("Unable to register rtnetlink message handler, "
227 "protocol = %d, message type = %d\n",
230 EXPORT_SYMBOL_GPL(rtnl_register);
233 * rtnl_unregister - Unregister a rtnetlink message type
234 * @protocol: Protocol family or PF_UNSPEC
235 * @msgtype: rtnetlink message type
237 * Returns 0 on success or a negative error code.
239 int rtnl_unregister(int protocol, int msgtype)
243 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
244 msgindex = rtm_msgindex(msgtype);
246 if (rtnl_msg_handlers[protocol] == NULL)
249 rtnl_msg_handlers[protocol][msgindex].doit = NULL;
250 rtnl_msg_handlers[protocol][msgindex].dumpit = NULL;
254 EXPORT_SYMBOL_GPL(rtnl_unregister);
257 * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
258 * @protocol : Protocol family or PF_UNSPEC
260 * Identical to calling rtnl_unregster() for all registered message types
261 * of a certain protocol family.
263 void rtnl_unregister_all(int protocol)
265 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
267 kfree(rtnl_msg_handlers[protocol]);
268 rtnl_msg_handlers[protocol] = NULL;
270 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
272 static LIST_HEAD(link_ops);
274 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
276 const struct rtnl_link_ops *ops;
278 list_for_each_entry(ops, &link_ops, list) {
279 if (!strcmp(ops->kind, kind))
286 * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
287 * @ops: struct rtnl_link_ops * to register
289 * The caller must hold the rtnl_mutex. This function should be used
290 * by drivers that create devices during module initialization. It
291 * must be called before registering the devices.
293 * Returns 0 on success or a negative error code.
295 int __rtnl_link_register(struct rtnl_link_ops *ops)
297 if (rtnl_link_ops_get(ops->kind))
301 ops->dellink = unregister_netdevice_queue;
303 list_add_tail(&ops->list, &link_ops);
306 EXPORT_SYMBOL_GPL(__rtnl_link_register);
309 * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
310 * @ops: struct rtnl_link_ops * to register
312 * Returns 0 on success or a negative error code.
314 int rtnl_link_register(struct rtnl_link_ops *ops)
319 err = __rtnl_link_register(ops);
323 EXPORT_SYMBOL_GPL(rtnl_link_register);
325 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
327 struct net_device *dev;
328 LIST_HEAD(list_kill);
330 for_each_netdev(net, dev) {
331 if (dev->rtnl_link_ops == ops)
332 ops->dellink(dev, &list_kill);
334 unregister_netdevice_many(&list_kill);
338 * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
339 * @ops: struct rtnl_link_ops * to unregister
341 * The caller must hold the rtnl_mutex.
343 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
348 __rtnl_kill_links(net, ops);
350 list_del(&ops->list);
352 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
355 * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
356 * @ops: struct rtnl_link_ops * to unregister
358 void rtnl_link_unregister(struct rtnl_link_ops *ops)
361 __rtnl_link_unregister(ops);
364 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
366 static size_t rtnl_link_get_size(const struct net_device *dev)
368 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
374 size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
375 nla_total_size(strlen(ops->kind) + 1); /* IFLA_INFO_KIND */
378 /* IFLA_INFO_DATA + nested data */
379 size += nla_total_size(sizeof(struct nlattr)) +
382 if (ops->get_xstats_size)
383 /* IFLA_INFO_XSTATS */
384 size += nla_total_size(ops->get_xstats_size(dev));
389 static LIST_HEAD(rtnl_af_ops);
391 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
393 const struct rtnl_af_ops *ops;
395 list_for_each_entry(ops, &rtnl_af_ops, list) {
396 if (ops->family == family)
404 * __rtnl_af_register - Register rtnl_af_ops with rtnetlink.
405 * @ops: struct rtnl_af_ops * to register
407 * The caller must hold the rtnl_mutex.
409 * Returns 0 on success or a negative error code.
411 int __rtnl_af_register(struct rtnl_af_ops *ops)
413 list_add_tail(&ops->list, &rtnl_af_ops);
416 EXPORT_SYMBOL_GPL(__rtnl_af_register);
419 * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
420 * @ops: struct rtnl_af_ops * to register
422 * Returns 0 on success or a negative error code.
424 int rtnl_af_register(struct rtnl_af_ops *ops)
429 err = __rtnl_af_register(ops);
433 EXPORT_SYMBOL_GPL(rtnl_af_register);
436 * __rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
437 * @ops: struct rtnl_af_ops * to unregister
439 * The caller must hold the rtnl_mutex.
441 void __rtnl_af_unregister(struct rtnl_af_ops *ops)
443 list_del(&ops->list);
445 EXPORT_SYMBOL_GPL(__rtnl_af_unregister);
448 * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
449 * @ops: struct rtnl_af_ops * to unregister
451 void rtnl_af_unregister(struct rtnl_af_ops *ops)
454 __rtnl_af_unregister(ops);
457 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
459 static size_t rtnl_link_get_af_size(const struct net_device *dev)
461 struct rtnl_af_ops *af_ops;
465 size = nla_total_size(sizeof(struct nlattr));
467 list_for_each_entry(af_ops, &rtnl_af_ops, list) {
468 if (af_ops->get_link_af_size) {
469 /* AF_* + nested data */
470 size += nla_total_size(sizeof(struct nlattr)) +
471 af_ops->get_link_af_size(dev);
478 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
480 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
481 struct nlattr *linkinfo, *data;
484 linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
485 if (linkinfo == NULL)
488 if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
489 goto err_cancel_link;
490 if (ops->fill_xstats) {
491 err = ops->fill_xstats(skb, dev);
493 goto err_cancel_link;
495 if (ops->fill_info) {
496 data = nla_nest_start(skb, IFLA_INFO_DATA);
498 goto err_cancel_link;
499 err = ops->fill_info(skb, dev);
501 goto err_cancel_data;
502 nla_nest_end(skb, data);
505 nla_nest_end(skb, linkinfo);
509 nla_nest_cancel(skb, data);
511 nla_nest_cancel(skb, linkinfo);
516 static const int rtm_min[RTM_NR_FAMILIES] =
518 [RTM_FAM(RTM_NEWLINK)] = NLMSG_LENGTH(sizeof(struct ifinfomsg)),
519 [RTM_FAM(RTM_NEWADDR)] = NLMSG_LENGTH(sizeof(struct ifaddrmsg)),
520 [RTM_FAM(RTM_NEWROUTE)] = NLMSG_LENGTH(sizeof(struct rtmsg)),
521 [RTM_FAM(RTM_NEWRULE)] = NLMSG_LENGTH(sizeof(struct fib_rule_hdr)),
522 [RTM_FAM(RTM_NEWQDISC)] = NLMSG_LENGTH(sizeof(struct tcmsg)),
523 [RTM_FAM(RTM_NEWTCLASS)] = NLMSG_LENGTH(sizeof(struct tcmsg)),
524 [RTM_FAM(RTM_NEWTFILTER)] = NLMSG_LENGTH(sizeof(struct tcmsg)),
525 [RTM_FAM(RTM_NEWACTION)] = NLMSG_LENGTH(sizeof(struct tcamsg)),
526 [RTM_FAM(RTM_GETMULTICAST)] = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
527 [RTM_FAM(RTM_GETANYCAST)] = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
530 static const int rta_max[RTM_NR_FAMILIES] =
532 [RTM_FAM(RTM_NEWLINK)] = IFLA_MAX,
533 [RTM_FAM(RTM_NEWADDR)] = IFA_MAX,
534 [RTM_FAM(RTM_NEWROUTE)] = RTA_MAX,
535 [RTM_FAM(RTM_NEWRULE)] = FRA_MAX,
536 [RTM_FAM(RTM_NEWQDISC)] = TCA_MAX,
537 [RTM_FAM(RTM_NEWTCLASS)] = TCA_MAX,
538 [RTM_FAM(RTM_NEWTFILTER)] = TCA_MAX,
539 [RTM_FAM(RTM_NEWACTION)] = TCAA_MAX,
542 void __rta_fill(struct sk_buff *skb, int attrtype, int attrlen, const void *data)
545 int size = RTA_LENGTH(attrlen);
547 rta = (struct rtattr *)skb_put(skb, RTA_ALIGN(size));
548 rta->rta_type = attrtype;
550 memcpy(RTA_DATA(rta), data, attrlen);
551 memset(RTA_DATA(rta) + attrlen, 0, RTA_ALIGN(size) - size);
553 EXPORT_SYMBOL(__rta_fill);
555 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned group, int echo)
557 struct sock *rtnl = net->rtnl;
560 NETLINK_CB(skb).dst_group = group;
562 atomic_inc(&skb->users);
563 netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
565 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
569 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
571 struct sock *rtnl = net->rtnl;
573 return nlmsg_unicast(rtnl, skb, pid);
575 EXPORT_SYMBOL(rtnl_unicast);
577 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
578 struct nlmsghdr *nlh, gfp_t flags)
580 struct sock *rtnl = net->rtnl;
584 report = nlmsg_report(nlh);
586 nlmsg_notify(rtnl, skb, pid, group, report, flags);
588 EXPORT_SYMBOL(rtnl_notify);
590 void rtnl_set_sk_err(struct net *net, u32 group, int error)
592 struct sock *rtnl = net->rtnl;
594 netlink_set_err(rtnl, 0, group, error);
596 EXPORT_SYMBOL(rtnl_set_sk_err);
598 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
603 mx = nla_nest_start(skb, RTA_METRICS);
607 for (i = 0; i < RTAX_MAX; i++) {
610 NLA_PUT_U32(skb, i+1, metrics[i]);
615 nla_nest_cancel(skb, mx);
619 return nla_nest_end(skb, mx);
622 nla_nest_cancel(skb, mx);
625 EXPORT_SYMBOL(rtnetlink_put_metrics);
627 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
628 u32 ts, u32 tsage, long expires, u32 error)
630 struct rta_cacheinfo ci = {
631 .rta_lastuse = jiffies_to_clock_t(jiffies - dst->lastuse),
632 .rta_used = dst->__use,
633 .rta_clntref = atomic_read(&(dst->__refcnt)),
641 ci.rta_expires = jiffies_to_clock_t(expires);
643 return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
645 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
647 static void set_operstate(struct net_device *dev, unsigned char transition)
649 unsigned char operstate = dev->operstate;
651 switch (transition) {
653 if ((operstate == IF_OPER_DORMANT ||
654 operstate == IF_OPER_UNKNOWN) &&
656 operstate = IF_OPER_UP;
659 case IF_OPER_DORMANT:
660 if (operstate == IF_OPER_UP ||
661 operstate == IF_OPER_UNKNOWN)
662 operstate = IF_OPER_DORMANT;
666 if (dev->operstate != operstate) {
667 write_lock_bh(&dev_base_lock);
668 dev->operstate = operstate;
669 write_unlock_bh(&dev_base_lock);
670 netdev_state_change(dev);
674 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
675 const struct ifinfomsg *ifm)
677 unsigned int flags = ifm->ifi_flags;
679 /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
681 flags = (flags & ifm->ifi_change) |
682 (dev->flags & ~ifm->ifi_change);
687 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
688 const struct rtnl_link_stats64 *b)
690 a->rx_packets = b->rx_packets;
691 a->tx_packets = b->tx_packets;
692 a->rx_bytes = b->rx_bytes;
693 a->tx_bytes = b->tx_bytes;
694 a->rx_errors = b->rx_errors;
695 a->tx_errors = b->tx_errors;
696 a->rx_dropped = b->rx_dropped;
697 a->tx_dropped = b->tx_dropped;
699 a->multicast = b->multicast;
700 a->collisions = b->collisions;
702 a->rx_length_errors = b->rx_length_errors;
703 a->rx_over_errors = b->rx_over_errors;
704 a->rx_crc_errors = b->rx_crc_errors;
705 a->rx_frame_errors = b->rx_frame_errors;
706 a->rx_fifo_errors = b->rx_fifo_errors;
707 a->rx_missed_errors = b->rx_missed_errors;
709 a->tx_aborted_errors = b->tx_aborted_errors;
710 a->tx_carrier_errors = b->tx_carrier_errors;
711 a->tx_fifo_errors = b->tx_fifo_errors;
712 a->tx_heartbeat_errors = b->tx_heartbeat_errors;
713 a->tx_window_errors = b->tx_window_errors;
715 a->rx_compressed = b->rx_compressed;
716 a->tx_compressed = b->tx_compressed;
719 static void copy_rtnl_link_stats64(void *v, const struct rtnl_link_stats64 *b)
721 memcpy(v, b, sizeof(*b));
725 static inline int rtnl_vfinfo_size(const struct net_device *dev,
728 if (dev->dev.parent && dev_is_pci(dev->dev.parent) &&
729 (ext_filter_mask & RTEXT_FILTER_VF)) {
730 int num_vfs = dev_num_vf(dev->dev.parent);
731 size_t size = nla_total_size(sizeof(struct nlattr));
732 size += nla_total_size(num_vfs * sizeof(struct nlattr));
734 (nla_total_size(sizeof(struct ifla_vf_mac)) +
735 nla_total_size(sizeof(struct ifla_vf_vlan)) +
736 nla_total_size(sizeof(struct ifla_vf_tx_rate)) +
737 nla_total_size(sizeof(struct ifla_vf_spoofchk)));
743 static size_t rtnl_port_size(const struct net_device *dev)
745 size_t port_size = nla_total_size(4) /* PORT_VF */
746 + nla_total_size(PORT_PROFILE_MAX) /* PORT_PROFILE */
747 + nla_total_size(sizeof(struct ifla_port_vsi))
749 + nla_total_size(PORT_UUID_MAX) /* PORT_INSTANCE_UUID */
750 + nla_total_size(PORT_UUID_MAX) /* PORT_HOST_UUID */
751 + nla_total_size(1) /* PROT_VDP_REQUEST */
752 + nla_total_size(2); /* PORT_VDP_RESPONSE */
753 size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
754 size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
756 size_t port_self_size = nla_total_size(sizeof(struct nlattr))
759 if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent)
761 if (dev_num_vf(dev->dev.parent))
762 return port_self_size + vf_ports_size +
763 vf_port_size * dev_num_vf(dev->dev.parent);
765 return port_self_size;
768 static noinline size_t if_nlmsg_size(const struct net_device *dev,
771 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
772 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
773 + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
774 + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
775 + nla_total_size(sizeof(struct rtnl_link_ifmap))
776 + nla_total_size(sizeof(struct rtnl_link_stats))
777 + nla_total_size(sizeof(struct rtnl_link_stats64))
778 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
779 + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
780 + nla_total_size(4) /* IFLA_TXQLEN */
781 + nla_total_size(4) /* IFLA_WEIGHT */
782 + nla_total_size(4) /* IFLA_MTU */
783 + nla_total_size(4) /* IFLA_LINK */
784 + nla_total_size(4) /* IFLA_MASTER */
785 + nla_total_size(1) /* IFLA_OPERSTATE */
786 + nla_total_size(1) /* IFLA_LINKMODE */
787 + nla_total_size(ext_filter_mask
788 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
789 + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
790 + rtnl_port_size(dev) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
791 + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
792 + rtnl_link_get_af_size(dev); /* IFLA_AF_SPEC */
795 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
797 struct nlattr *vf_ports;
798 struct nlattr *vf_port;
802 vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
806 for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
807 vf_port = nla_nest_start(skb, IFLA_VF_PORT);
809 goto nla_put_failure;
810 NLA_PUT_U32(skb, IFLA_PORT_VF, vf);
811 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
812 if (err == -EMSGSIZE)
813 goto nla_put_failure;
815 nla_nest_cancel(skb, vf_port);
818 nla_nest_end(skb, vf_port);
821 nla_nest_end(skb, vf_ports);
826 nla_nest_cancel(skb, vf_ports);
830 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
832 struct nlattr *port_self;
835 port_self = nla_nest_start(skb, IFLA_PORT_SELF);
839 err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
841 nla_nest_cancel(skb, port_self);
842 return (err == -EMSGSIZE) ? err : 0;
845 nla_nest_end(skb, port_self);
850 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev)
854 if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent)
857 err = rtnl_port_self_fill(skb, dev);
861 if (dev_num_vf(dev->dev.parent)) {
862 err = rtnl_vf_ports_fill(skb, dev);
870 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
871 int type, u32 pid, u32 seq, u32 change,
872 unsigned int flags, u32 ext_filter_mask)
874 struct ifinfomsg *ifm;
875 struct nlmsghdr *nlh;
876 struct rtnl_link_stats64 temp;
877 const struct rtnl_link_stats64 *stats;
878 struct nlattr *attr, *af_spec;
879 struct rtnl_af_ops *af_ops;
882 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
886 ifm = nlmsg_data(nlh);
887 ifm->ifi_family = AF_UNSPEC;
889 ifm->ifi_type = dev->type;
890 ifm->ifi_index = dev->ifindex;
891 ifm->ifi_flags = dev_get_flags(dev);
892 ifm->ifi_change = change;
894 NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
895 NLA_PUT_U32(skb, IFLA_TXQLEN, dev->tx_queue_len);
896 NLA_PUT_U8(skb, IFLA_OPERSTATE,
897 netif_running(dev) ? dev->operstate : IF_OPER_DOWN);
898 NLA_PUT_U8(skb, IFLA_LINKMODE, dev->link_mode);
899 NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
900 NLA_PUT_U32(skb, IFLA_GROUP, dev->group);
902 if (dev->ifindex != dev->iflink)
903 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
906 NLA_PUT_U32(skb, IFLA_MASTER, dev->master->ifindex);
909 NLA_PUT_STRING(skb, IFLA_QDISC, dev->qdisc->ops->id);
912 NLA_PUT_STRING(skb, IFLA_IFALIAS, dev->ifalias);
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 NLA_PUT(skb, IFLA_MAP, sizeof(map), &map);
927 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
928 NLA_PUT(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast);
931 attr = nla_reserve(skb, IFLA_STATS,
932 sizeof(struct rtnl_link_stats));
934 goto nla_put_failure;
936 stats = dev_get_stats(dev, &temp);
937 copy_rtnl_link_stats(nla_data(attr), stats);
939 attr = nla_reserve(skb, IFLA_STATS64,
940 sizeof(struct rtnl_link_stats64));
942 goto nla_put_failure;
943 copy_rtnl_link_stats64(nla_data(attr), stats);
945 if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF))
946 NLA_PUT_U32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent));
948 if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent
949 && (ext_filter_mask & RTEXT_FILTER_VF)) {
952 struct nlattr *vfinfo, *vf;
953 int num_vfs = dev_num_vf(dev->dev.parent);
955 vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
957 goto nla_put_failure;
958 for (i = 0; i < num_vfs; i++) {
959 struct ifla_vf_info ivi;
960 struct ifla_vf_mac vf_mac;
961 struct ifla_vf_vlan vf_vlan;
962 struct ifla_vf_tx_rate vf_tx_rate;
963 struct ifla_vf_spoofchk vf_spoofchk;
966 * Not all SR-IOV capable drivers support the
967 * spoofcheck query. Preset to -1 so the user
968 * space tool can detect that the driver didn't
972 if (dev->netdev_ops->ndo_get_vf_config(dev, i, &ivi))
977 vf_spoofchk.vf = ivi.vf;
979 memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
980 vf_vlan.vlan = ivi.vlan;
981 vf_vlan.qos = ivi.qos;
982 vf_tx_rate.rate = ivi.tx_rate;
983 vf_spoofchk.setting = ivi.spoofchk;
984 vf = nla_nest_start(skb, IFLA_VF_INFO);
986 nla_nest_cancel(skb, vfinfo);
987 goto nla_put_failure;
989 NLA_PUT(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac);
990 NLA_PUT(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan);
991 NLA_PUT(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
993 NLA_PUT(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
995 nla_nest_end(skb, vf);
997 nla_nest_end(skb, vfinfo);
1000 if (rtnl_port_fill(skb, dev))
1001 goto nla_put_failure;
1003 if (dev->rtnl_link_ops) {
1004 if (rtnl_link_fill(skb, dev) < 0)
1005 goto nla_put_failure;
1008 if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1009 goto nla_put_failure;
1011 list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1012 if (af_ops->fill_link_af) {
1016 if (!(af = nla_nest_start(skb, af_ops->family)))
1017 goto nla_put_failure;
1019 err = af_ops->fill_link_af(skb, dev);
1022 * Caller may return ENODATA to indicate that there
1023 * was no data to be dumped. This is not an error, it
1024 * means we should trim the attribute header and
1027 if (err == -ENODATA)
1028 nla_nest_cancel(skb, af);
1030 goto nla_put_failure;
1032 nla_nest_end(skb, af);
1036 nla_nest_end(skb, af_spec);
1038 return nlmsg_end(skb, nlh);
1041 nlmsg_cancel(skb, nlh);
1045 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1047 struct net *net = sock_net(skb->sk);
1050 struct net_device *dev;
1051 struct hlist_head *head;
1052 struct hlist_node *node;
1053 struct nlattr *tb[IFLA_MAX+1];
1054 u32 ext_filter_mask = 0;
1057 s_idx = cb->args[1];
1060 cb->seq = net->dev_base_seq;
1062 if (nlmsg_parse(cb->nlh, sizeof(struct rtgenmsg), tb, IFLA_MAX,
1063 ifla_policy) >= 0) {
1065 if (tb[IFLA_EXT_MASK])
1066 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1069 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1071 head = &net->dev_index_head[h];
1072 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
1075 if (rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1076 NETLINK_CB(cb->skb).pid,
1077 cb->nlh->nlmsg_seq, 0,
1079 ext_filter_mask) <= 0)
1082 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1095 const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1096 [IFLA_IFNAME] = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1097 [IFLA_ADDRESS] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1098 [IFLA_BROADCAST] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1099 [IFLA_MAP] = { .len = sizeof(struct rtnl_link_ifmap) },
1100 [IFLA_MTU] = { .type = NLA_U32 },
1101 [IFLA_LINK] = { .type = NLA_U32 },
1102 [IFLA_MASTER] = { .type = NLA_U32 },
1103 [IFLA_TXQLEN] = { .type = NLA_U32 },
1104 [IFLA_WEIGHT] = { .type = NLA_U32 },
1105 [IFLA_OPERSTATE] = { .type = NLA_U8 },
1106 [IFLA_LINKMODE] = { .type = NLA_U8 },
1107 [IFLA_LINKINFO] = { .type = NLA_NESTED },
1108 [IFLA_NET_NS_PID] = { .type = NLA_U32 },
1109 [IFLA_NET_NS_FD] = { .type = NLA_U32 },
1110 [IFLA_IFALIAS] = { .type = NLA_STRING, .len = IFALIASZ-1 },
1111 [IFLA_VFINFO_LIST] = {. type = NLA_NESTED },
1112 [IFLA_VF_PORTS] = { .type = NLA_NESTED },
1113 [IFLA_PORT_SELF] = { .type = NLA_NESTED },
1114 [IFLA_AF_SPEC] = { .type = NLA_NESTED },
1115 [IFLA_EXT_MASK] = { .type = NLA_U32 },
1117 EXPORT_SYMBOL(ifla_policy);
1119 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1120 [IFLA_INFO_KIND] = { .type = NLA_STRING },
1121 [IFLA_INFO_DATA] = { .type = NLA_NESTED },
1124 static const struct nla_policy ifla_vfinfo_policy[IFLA_VF_INFO_MAX+1] = {
1125 [IFLA_VF_INFO] = { .type = NLA_NESTED },
1128 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1129 [IFLA_VF_MAC] = { .type = NLA_BINARY,
1130 .len = sizeof(struct ifla_vf_mac) },
1131 [IFLA_VF_VLAN] = { .type = NLA_BINARY,
1132 .len = sizeof(struct ifla_vf_vlan) },
1133 [IFLA_VF_TX_RATE] = { .type = NLA_BINARY,
1134 .len = sizeof(struct ifla_vf_tx_rate) },
1135 [IFLA_VF_SPOOFCHK] = { .type = NLA_BINARY,
1136 .len = sizeof(struct ifla_vf_spoofchk) },
1139 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1140 [IFLA_PORT_VF] = { .type = NLA_U32 },
1141 [IFLA_PORT_PROFILE] = { .type = NLA_STRING,
1142 .len = PORT_PROFILE_MAX },
1143 [IFLA_PORT_VSI_TYPE] = { .type = NLA_BINARY,
1144 .len = sizeof(struct ifla_port_vsi)},
1145 [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1146 .len = PORT_UUID_MAX },
1147 [IFLA_PORT_HOST_UUID] = { .type = NLA_STRING,
1148 .len = PORT_UUID_MAX },
1149 [IFLA_PORT_REQUEST] = { .type = NLA_U8, },
1150 [IFLA_PORT_RESPONSE] = { .type = NLA_U16, },
1153 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1156 /* Examine the link attributes and figure out which
1157 * network namespace we are talking about.
1159 if (tb[IFLA_NET_NS_PID])
1160 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1161 else if (tb[IFLA_NET_NS_FD])
1162 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1164 net = get_net(src_net);
1167 EXPORT_SYMBOL(rtnl_link_get_net);
1169 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1172 if (tb[IFLA_ADDRESS] &&
1173 nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1176 if (tb[IFLA_BROADCAST] &&
1177 nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1181 if (tb[IFLA_AF_SPEC]) {
1185 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1186 const struct rtnl_af_ops *af_ops;
1188 if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1189 return -EAFNOSUPPORT;
1191 if (!af_ops->set_link_af)
1194 if (af_ops->validate_link_af) {
1195 err = af_ops->validate_link_af(dev, af);
1205 static int do_setvfinfo(struct net_device *dev, struct nlattr *attr)
1207 int rem, err = -EINVAL;
1209 const struct net_device_ops *ops = dev->netdev_ops;
1211 nla_for_each_nested(vf, attr, rem) {
1212 switch (nla_type(vf)) {
1214 struct ifla_vf_mac *ivm;
1217 if (ops->ndo_set_vf_mac)
1218 err = ops->ndo_set_vf_mac(dev, ivm->vf,
1222 case IFLA_VF_VLAN: {
1223 struct ifla_vf_vlan *ivv;
1226 if (ops->ndo_set_vf_vlan)
1227 err = ops->ndo_set_vf_vlan(dev, ivv->vf,
1232 case IFLA_VF_TX_RATE: {
1233 struct ifla_vf_tx_rate *ivt;
1236 if (ops->ndo_set_vf_tx_rate)
1237 err = ops->ndo_set_vf_tx_rate(dev, ivt->vf,
1241 case IFLA_VF_SPOOFCHK: {
1242 struct ifla_vf_spoofchk *ivs;
1245 if (ops->ndo_set_vf_spoofchk)
1246 err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1260 static int do_set_master(struct net_device *dev, int ifindex)
1262 struct net_device *master_dev;
1263 const struct net_device_ops *ops;
1267 if (dev->master->ifindex == ifindex)
1269 ops = dev->master->netdev_ops;
1270 if (ops->ndo_del_slave) {
1271 err = ops->ndo_del_slave(dev->master, dev);
1280 master_dev = __dev_get_by_index(dev_net(dev), ifindex);
1283 ops = master_dev->netdev_ops;
1284 if (ops->ndo_add_slave) {
1285 err = ops->ndo_add_slave(master_dev, dev);
1295 static int do_setlink(struct net_device *dev, struct ifinfomsg *ifm,
1296 struct nlattr **tb, char *ifname, int modified)
1298 const struct net_device_ops *ops = dev->netdev_ops;
1299 int send_addr_notify = 0;
1302 if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1303 struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1308 err = dev_change_net_namespace(dev, net, ifname);
1316 struct rtnl_link_ifmap *u_map;
1319 if (!ops->ndo_set_config) {
1324 if (!netif_device_present(dev)) {
1329 u_map = nla_data(tb[IFLA_MAP]);
1330 k_map.mem_start = (unsigned long) u_map->mem_start;
1331 k_map.mem_end = (unsigned long) u_map->mem_end;
1332 k_map.base_addr = (unsigned short) u_map->base_addr;
1333 k_map.irq = (unsigned char) u_map->irq;
1334 k_map.dma = (unsigned char) u_map->dma;
1335 k_map.port = (unsigned char) u_map->port;
1337 err = ops->ndo_set_config(dev, &k_map);
1344 if (tb[IFLA_ADDRESS]) {
1345 struct sockaddr *sa;
1348 if (!ops->ndo_set_mac_address) {
1353 if (!netif_device_present(dev)) {
1358 len = sizeof(sa_family_t) + dev->addr_len;
1359 sa = kmalloc(len, GFP_KERNEL);
1364 sa->sa_family = dev->type;
1365 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1367 err = ops->ndo_set_mac_address(dev, sa);
1371 send_addr_notify = 1;
1376 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1382 if (tb[IFLA_GROUP]) {
1383 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1388 * Interface selected by interface index but interface
1389 * name provided implies that a name change has been
1392 if (ifm->ifi_index > 0 && ifname[0]) {
1393 err = dev_change_name(dev, ifname);
1399 if (tb[IFLA_IFALIAS]) {
1400 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
1401 nla_len(tb[IFLA_IFALIAS]));
1407 if (tb[IFLA_BROADCAST]) {
1408 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
1409 send_addr_notify = 1;
1412 if (ifm->ifi_flags || ifm->ifi_change) {
1413 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1418 if (tb[IFLA_MASTER]) {
1419 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
1425 if (tb[IFLA_TXQLEN])
1426 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1428 if (tb[IFLA_OPERSTATE])
1429 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1431 if (tb[IFLA_LINKMODE]) {
1432 write_lock_bh(&dev_base_lock);
1433 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1434 write_unlock_bh(&dev_base_lock);
1437 if (tb[IFLA_VFINFO_LIST]) {
1438 struct nlattr *attr;
1440 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
1441 if (nla_type(attr) != IFLA_VF_INFO) {
1445 err = do_setvfinfo(dev, attr);
1453 if (tb[IFLA_VF_PORTS]) {
1454 struct nlattr *port[IFLA_PORT_MAX+1];
1455 struct nlattr *attr;
1460 if (!ops->ndo_set_vf_port)
1463 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
1464 if (nla_type(attr) != IFLA_VF_PORT)
1466 err = nla_parse_nested(port, IFLA_PORT_MAX,
1467 attr, ifla_port_policy);
1470 if (!port[IFLA_PORT_VF]) {
1474 vf = nla_get_u32(port[IFLA_PORT_VF]);
1475 err = ops->ndo_set_vf_port(dev, vf, port);
1483 if (tb[IFLA_PORT_SELF]) {
1484 struct nlattr *port[IFLA_PORT_MAX+1];
1486 err = nla_parse_nested(port, IFLA_PORT_MAX,
1487 tb[IFLA_PORT_SELF], ifla_port_policy);
1492 if (ops->ndo_set_vf_port)
1493 err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
1499 if (tb[IFLA_AF_SPEC]) {
1503 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1504 const struct rtnl_af_ops *af_ops;
1506 if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1509 err = af_ops->set_link_af(dev, af);
1519 if (err < 0 && modified && net_ratelimit())
1520 printk(KERN_WARNING "A link change request failed with "
1521 "some changes committed already. Interface %s may "
1522 "have been left with an inconsistent configuration, "
1523 "please check.\n", dev->name);
1525 if (send_addr_notify)
1526 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
1531 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1533 struct net *net = sock_net(skb->sk);
1534 struct ifinfomsg *ifm;
1535 struct net_device *dev;
1537 struct nlattr *tb[IFLA_MAX+1];
1538 char ifname[IFNAMSIZ];
1540 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1544 if (tb[IFLA_IFNAME])
1545 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1550 ifm = nlmsg_data(nlh);
1551 if (ifm->ifi_index > 0)
1552 dev = __dev_get_by_index(net, ifm->ifi_index);
1553 else if (tb[IFLA_IFNAME])
1554 dev = __dev_get_by_name(net, ifname);
1563 err = validate_linkmsg(dev, tb);
1567 err = do_setlink(dev, ifm, tb, ifname, 0);
1572 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1574 struct net *net = sock_net(skb->sk);
1575 const struct rtnl_link_ops *ops;
1576 struct net_device *dev;
1577 struct ifinfomsg *ifm;
1578 char ifname[IFNAMSIZ];
1579 struct nlattr *tb[IFLA_MAX+1];
1581 LIST_HEAD(list_kill);
1583 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1587 if (tb[IFLA_IFNAME])
1588 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1590 ifm = nlmsg_data(nlh);
1591 if (ifm->ifi_index > 0)
1592 dev = __dev_get_by_index(net, ifm->ifi_index);
1593 else if (tb[IFLA_IFNAME])
1594 dev = __dev_get_by_name(net, ifname);
1601 ops = dev->rtnl_link_ops;
1605 ops->dellink(dev, &list_kill);
1606 unregister_netdevice_many(&list_kill);
1607 list_del(&list_kill);
1611 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
1613 unsigned int old_flags;
1616 old_flags = dev->flags;
1617 if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
1618 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1623 dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
1624 rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U);
1626 __dev_notify_flags(dev, old_flags);
1629 EXPORT_SYMBOL(rtnl_configure_link);
1631 struct net_device *rtnl_create_link(struct net *src_net, struct net *net,
1632 char *ifname, const struct rtnl_link_ops *ops, struct nlattr *tb[])
1635 struct net_device *dev;
1636 unsigned int num_queues = 1;
1637 unsigned int real_num_queues = 1;
1639 if (ops->get_tx_queues) {
1640 err = ops->get_tx_queues(src_net, tb, &num_queues,
1646 dev = alloc_netdev_mq(ops->priv_size, ifname, ops->setup, num_queues);
1650 dev_net_set(dev, net);
1651 dev->rtnl_link_ops = ops;
1652 dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
1655 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
1656 if (tb[IFLA_ADDRESS])
1657 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
1658 nla_len(tb[IFLA_ADDRESS]));
1659 if (tb[IFLA_BROADCAST])
1660 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
1661 nla_len(tb[IFLA_BROADCAST]));
1662 if (tb[IFLA_TXQLEN])
1663 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1664 if (tb[IFLA_OPERSTATE])
1665 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1666 if (tb[IFLA_LINKMODE])
1667 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1669 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
1674 return ERR_PTR(err);
1676 EXPORT_SYMBOL(rtnl_create_link);
1678 static int rtnl_group_changelink(struct net *net, int group,
1679 struct ifinfomsg *ifm,
1682 struct net_device *dev;
1685 for_each_netdev(net, dev) {
1686 if (dev->group == group) {
1687 err = do_setlink(dev, ifm, tb, NULL, 0);
1696 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1698 struct net *net = sock_net(skb->sk);
1699 const struct rtnl_link_ops *ops;
1700 struct net_device *dev;
1701 struct ifinfomsg *ifm;
1702 char kind[MODULE_NAME_LEN];
1703 char ifname[IFNAMSIZ];
1704 struct nlattr *tb[IFLA_MAX+1];
1705 struct nlattr *linkinfo[IFLA_INFO_MAX+1];
1708 #ifdef CONFIG_MODULES
1711 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1715 if (tb[IFLA_IFNAME])
1716 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1720 ifm = nlmsg_data(nlh);
1721 if (ifm->ifi_index > 0)
1722 dev = __dev_get_by_index(net, ifm->ifi_index);
1725 dev = __dev_get_by_name(net, ifname);
1730 err = validate_linkmsg(dev, tb);
1734 if (tb[IFLA_LINKINFO]) {
1735 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
1736 tb[IFLA_LINKINFO], ifla_info_policy);
1740 memset(linkinfo, 0, sizeof(linkinfo));
1742 if (linkinfo[IFLA_INFO_KIND]) {
1743 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
1744 ops = rtnl_link_ops_get(kind);
1751 struct nlattr *attr[ops ? ops->maxtype + 1 : 0], **data = NULL;
1752 struct net *dest_net;
1755 if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
1756 err = nla_parse_nested(attr, ops->maxtype,
1757 linkinfo[IFLA_INFO_DATA],
1763 if (ops->validate) {
1764 err = ops->validate(tb, data);
1773 if (nlh->nlmsg_flags & NLM_F_EXCL)
1775 if (nlh->nlmsg_flags & NLM_F_REPLACE)
1778 if (linkinfo[IFLA_INFO_DATA]) {
1779 if (!ops || ops != dev->rtnl_link_ops ||
1783 err = ops->changelink(dev, tb, data);
1789 return do_setlink(dev, ifm, tb, ifname, modified);
1792 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
1793 if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
1794 return rtnl_group_changelink(net,
1795 nla_get_u32(tb[IFLA_GROUP]),
1802 if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
1806 #ifdef CONFIG_MODULES
1809 request_module("rtnl-link-%s", kind);
1811 ops = rtnl_link_ops_get(kind);
1820 snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
1822 dest_net = rtnl_link_get_net(net, tb);
1823 if (IS_ERR(dest_net))
1824 return PTR_ERR(dest_net);
1826 dev = rtnl_create_link(net, dest_net, ifname, ops, tb);
1830 else if (ops->newlink)
1831 err = ops->newlink(net, dev, tb, data);
1833 err = register_netdevice(dev);
1835 if (err < 0 && !IS_ERR(dev))
1840 err = rtnl_configure_link(dev, ifm);
1842 unregister_netdevice(dev);
1849 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
1851 struct net *net = sock_net(skb->sk);
1852 struct ifinfomsg *ifm;
1853 char ifname[IFNAMSIZ];
1854 struct nlattr *tb[IFLA_MAX+1];
1855 struct net_device *dev = NULL;
1856 struct sk_buff *nskb;
1858 u32 ext_filter_mask = 0;
1860 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1864 if (tb[IFLA_IFNAME])
1865 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1867 if (tb[IFLA_EXT_MASK])
1868 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1870 ifm = nlmsg_data(nlh);
1871 if (ifm->ifi_index > 0)
1872 dev = __dev_get_by_index(net, ifm->ifi_index);
1873 else if (tb[IFLA_IFNAME])
1874 dev = __dev_get_by_name(net, ifname);
1881 nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
1885 err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).pid,
1886 nlh->nlmsg_seq, 0, 0, ext_filter_mask);
1888 /* -EMSGSIZE implies BUG in if_nlmsg_size */
1889 WARN_ON(err == -EMSGSIZE);
1892 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).pid);
1897 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
1899 struct net *net = sock_net(skb->sk);
1900 struct net_device *dev;
1901 struct nlattr *tb[IFLA_MAX+1];
1902 u32 ext_filter_mask = 0;
1903 u16 min_ifinfo_dump_size = 0;
1905 if (nlmsg_parse(nlh, sizeof(struct rtgenmsg), tb, IFLA_MAX,
1906 ifla_policy) >= 0) {
1907 if (tb[IFLA_EXT_MASK])
1908 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1911 if (!ext_filter_mask)
1912 return NLMSG_GOODSIZE;
1914 * traverse the list of net devices and compute the minimum
1915 * buffer size based upon the filter mask.
1917 list_for_each_entry(dev, &net->dev_base_head, dev_list) {
1918 min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
1923 return min_ifinfo_dump_size;
1926 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
1929 int s_idx = cb->family;
1933 for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
1934 int type = cb->nlh->nlmsg_type-RTM_BASE;
1935 if (idx < s_idx || idx == PF_PACKET)
1937 if (rtnl_msg_handlers[idx] == NULL ||
1938 rtnl_msg_handlers[idx][type].dumpit == NULL)
1941 memset(&cb->args[0], 0, sizeof(cb->args));
1942 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
1950 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned change)
1952 struct net *net = dev_net(dev);
1953 struct sk_buff *skb;
1955 size_t if_info_size;
1957 skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), GFP_KERNEL);
1961 err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0);
1963 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
1964 WARN_ON(err == -EMSGSIZE);
1968 rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_KERNEL);
1972 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
1975 /* Protected by RTNL sempahore. */
1976 static struct rtattr **rta_buf;
1977 static int rtattr_max;
1979 /* Process one rtnetlink message. */
1981 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
1983 struct net *net = sock_net(skb->sk);
1984 rtnl_doit_func doit;
1991 type = nlh->nlmsg_type;
1997 /* All the messages must have at least 1 byte length */
1998 if (nlh->nlmsg_len < NLMSG_LENGTH(sizeof(struct rtgenmsg)))
2001 family = ((struct rtgenmsg *)NLMSG_DATA(nlh))->rtgen_family;
2005 if (kind != 2 && !capable(CAP_NET_ADMIN))
2008 if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
2010 rtnl_dumpit_func dumpit;
2011 rtnl_calcit_func calcit;
2012 u16 min_dump_alloc = 0;
2014 dumpit = rtnl_get_dumpit(family, type);
2017 calcit = rtnl_get_calcit(family, type);
2019 min_dump_alloc = calcit(skb, nlh);
2024 struct netlink_dump_control c = {
2026 .min_dump_alloc = min_dump_alloc,
2028 err = netlink_dump_start(rtnl, skb, nlh, &c);
2034 memset(rta_buf, 0, (rtattr_max * sizeof(struct rtattr *)));
2036 min_len = rtm_min[sz_idx];
2037 if (nlh->nlmsg_len < min_len)
2040 if (nlh->nlmsg_len > min_len) {
2041 int attrlen = nlh->nlmsg_len - NLMSG_ALIGN(min_len);
2042 struct rtattr *attr = (void *)nlh + NLMSG_ALIGN(min_len);
2044 while (RTA_OK(attr, attrlen)) {
2045 unsigned flavor = attr->rta_type;
2047 if (flavor > rta_max[sz_idx])
2049 rta_buf[flavor-1] = attr;
2051 attr = RTA_NEXT(attr, attrlen);
2055 doit = rtnl_get_doit(family, type);
2059 return doit(skb, nlh, (void *)&rta_buf[0]);
2062 static void rtnetlink_rcv(struct sk_buff *skb)
2065 netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
2069 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
2071 struct net_device *dev = ptr;
2077 case NETDEV_POST_INIT:
2078 case NETDEV_REGISTER:
2080 case NETDEV_PRE_TYPE_CHANGE:
2081 case NETDEV_GOING_DOWN:
2082 case NETDEV_UNREGISTER:
2083 case NETDEV_UNREGISTER_BATCH:
2084 case NETDEV_RELEASE:
2088 rtmsg_ifinfo(RTM_NEWLINK, dev, 0);
2094 static struct notifier_block rtnetlink_dev_notifier = {
2095 .notifier_call = rtnetlink_event,
2099 static int __net_init rtnetlink_net_init(struct net *net)
2102 sk = netlink_kernel_create(net, NETLINK_ROUTE, RTNLGRP_MAX,
2103 rtnetlink_rcv, &rtnl_mutex, THIS_MODULE);
2110 static void __net_exit rtnetlink_net_exit(struct net *net)
2112 netlink_kernel_release(net->rtnl);
2116 static struct pernet_operations rtnetlink_net_ops = {
2117 .init = rtnetlink_net_init,
2118 .exit = rtnetlink_net_exit,
2121 void __init rtnetlink_init(void)
2126 for (i = 0; i < ARRAY_SIZE(rta_max); i++)
2127 if (rta_max[i] > rtattr_max)
2128 rtattr_max = rta_max[i];
2129 rta_buf = kmalloc(rtattr_max * sizeof(struct rtattr *), GFP_KERNEL);
2131 panic("rtnetlink_init: cannot allocate rta_buf\n");
2133 if (register_pernet_subsys(&rtnetlink_net_ops))
2134 panic("rtnetlink_init: cannot initialize rtnetlink\n");
2136 netlink_set_nonroot(NETLINK_ROUTE, NL_NONROOT_RECV);
2137 register_netdevice_notifier(&rtnetlink_dev_notifier);
2139 rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
2140 rtnl_dump_ifinfo, rtnl_calcit);
2141 rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, NULL);
2142 rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, NULL);
2143 rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, NULL);
2145 rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, NULL);
2146 rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, NULL);