writeback: split writeback_inodes_wb
[linux-2.6/next.git] / net / core / rtnetlink.c
blob1a2af24e9e3d2b3754fa032e261a9480b4d0b7ac
1 /*
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.
15 * Fixes:
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>
29 #include <linux/mm.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>
45 #include <net/ip.h>
46 #include <net/protocol.h>
47 #include <net/arp.h>
48 #include <net/route.h>
49 #include <net/udp.h>
50 #include <net/sock.h>
51 #include <net/pkt_sched.h>
52 #include <net/fib_rules.h>
53 #include <net/rtnetlink.h>
54 #include <net/net_namespace.h>
56 struct rtnl_link {
57 rtnl_doit_func doit;
58 rtnl_dumpit_func dumpit;
61 static DEFINE_MUTEX(rtnl_mutex);
63 void rtnl_lock(void)
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. */
77 netdev_run_todo();
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);
114 return msgindex;
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];
123 else
124 tab = NULL;
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];
138 else
139 tab = NULL;
141 if (tab == NULL || tab[msgindex].dumpit == NULL)
142 tab = rtnl_msg_handlers[PF_UNSPEC];
144 return tab ? tab[msgindex].dumpit : NULL;
148 * __rtnl_register - Register a rtnetlink message type
149 * @protocol: Protocol family or PF_UNSPEC
150 * @msgtype: rtnetlink message type
151 * @doit: Function pointer called for each request message
152 * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
154 * Registers the specified function pointers (at least one of them has
155 * to be non-NULL) to be called whenever a request message for the
156 * specified protocol family and message type is received.
158 * The special protocol family PF_UNSPEC may be used to define fallback
159 * function pointers for the case when no entry for the specific protocol
160 * family exists.
162 * Returns 0 on success or a negative error code.
164 int __rtnl_register(int protocol, int msgtype,
165 rtnl_doit_func doit, rtnl_dumpit_func dumpit)
167 struct rtnl_link *tab;
168 int msgindex;
170 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
171 msgindex = rtm_msgindex(msgtype);
173 tab = rtnl_msg_handlers[protocol];
174 if (tab == NULL) {
175 tab = kcalloc(RTM_NR_MSGTYPES, sizeof(*tab), GFP_KERNEL);
176 if (tab == NULL)
177 return -ENOBUFS;
179 rtnl_msg_handlers[protocol] = tab;
182 if (doit)
183 tab[msgindex].doit = doit;
185 if (dumpit)
186 tab[msgindex].dumpit = dumpit;
188 return 0;
190 EXPORT_SYMBOL_GPL(__rtnl_register);
193 * rtnl_register - Register a rtnetlink message type
195 * Identical to __rtnl_register() but panics on failure. This is useful
196 * as failure of this function is very unlikely, it can only happen due
197 * to lack of memory when allocating the chain to store all message
198 * handlers for a protocol. Meant for use in init functions where lack
199 * of memory implies no sense in continueing.
201 void rtnl_register(int protocol, int msgtype,
202 rtnl_doit_func doit, rtnl_dumpit_func dumpit)
204 if (__rtnl_register(protocol, msgtype, doit, dumpit) < 0)
205 panic("Unable to register rtnetlink message handler, "
206 "protocol = %d, message type = %d\n",
207 protocol, msgtype);
209 EXPORT_SYMBOL_GPL(rtnl_register);
212 * rtnl_unregister - Unregister a rtnetlink message type
213 * @protocol: Protocol family or PF_UNSPEC
214 * @msgtype: rtnetlink message type
216 * Returns 0 on success or a negative error code.
218 int rtnl_unregister(int protocol, int msgtype)
220 int msgindex;
222 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
223 msgindex = rtm_msgindex(msgtype);
225 if (rtnl_msg_handlers[protocol] == NULL)
226 return -ENOENT;
228 rtnl_msg_handlers[protocol][msgindex].doit = NULL;
229 rtnl_msg_handlers[protocol][msgindex].dumpit = NULL;
231 return 0;
233 EXPORT_SYMBOL_GPL(rtnl_unregister);
236 * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
237 * @protocol : Protocol family or PF_UNSPEC
239 * Identical to calling rtnl_unregster() for all registered message types
240 * of a certain protocol family.
242 void rtnl_unregister_all(int protocol)
244 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
246 kfree(rtnl_msg_handlers[protocol]);
247 rtnl_msg_handlers[protocol] = NULL;
249 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
251 static LIST_HEAD(link_ops);
254 * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
255 * @ops: struct rtnl_link_ops * to register
257 * The caller must hold the rtnl_mutex. This function should be used
258 * by drivers that create devices during module initialization. It
259 * must be called before registering the devices.
261 * Returns 0 on success or a negative error code.
263 int __rtnl_link_register(struct rtnl_link_ops *ops)
265 if (!ops->dellink)
266 ops->dellink = unregister_netdevice_queue;
268 list_add_tail(&ops->list, &link_ops);
269 return 0;
271 EXPORT_SYMBOL_GPL(__rtnl_link_register);
274 * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
275 * @ops: struct rtnl_link_ops * to register
277 * Returns 0 on success or a negative error code.
279 int rtnl_link_register(struct rtnl_link_ops *ops)
281 int err;
283 rtnl_lock();
284 err = __rtnl_link_register(ops);
285 rtnl_unlock();
286 return err;
288 EXPORT_SYMBOL_GPL(rtnl_link_register);
290 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
292 struct net_device *dev;
293 LIST_HEAD(list_kill);
295 for_each_netdev(net, dev) {
296 if (dev->rtnl_link_ops == ops)
297 ops->dellink(dev, &list_kill);
299 unregister_netdevice_many(&list_kill);
302 void rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
304 rtnl_lock();
305 __rtnl_kill_links(net, ops);
306 rtnl_unlock();
308 EXPORT_SYMBOL_GPL(rtnl_kill_links);
311 * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
312 * @ops: struct rtnl_link_ops * to unregister
314 * The caller must hold the rtnl_mutex.
316 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
318 struct net *net;
320 for_each_net(net) {
321 __rtnl_kill_links(net, ops);
323 list_del(&ops->list);
325 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
328 * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
329 * @ops: struct rtnl_link_ops * to unregister
331 void rtnl_link_unregister(struct rtnl_link_ops *ops)
333 rtnl_lock();
334 __rtnl_link_unregister(ops);
335 rtnl_unlock();
337 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
339 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
341 const struct rtnl_link_ops *ops;
343 list_for_each_entry(ops, &link_ops, list) {
344 if (!strcmp(ops->kind, kind))
345 return ops;
347 return NULL;
350 static size_t rtnl_link_get_size(const struct net_device *dev)
352 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
353 size_t size;
355 if (!ops)
356 return 0;
358 size = nlmsg_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
359 nlmsg_total_size(strlen(ops->kind) + 1); /* IFLA_INFO_KIND */
361 if (ops->get_size)
362 /* IFLA_INFO_DATA + nested data */
363 size += nlmsg_total_size(sizeof(struct nlattr)) +
364 ops->get_size(dev);
366 if (ops->get_xstats_size)
367 size += ops->get_xstats_size(dev); /* IFLA_INFO_XSTATS */
369 return size;
372 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
374 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
375 struct nlattr *linkinfo, *data;
376 int err = -EMSGSIZE;
378 linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
379 if (linkinfo == NULL)
380 goto out;
382 if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
383 goto err_cancel_link;
384 if (ops->fill_xstats) {
385 err = ops->fill_xstats(skb, dev);
386 if (err < 0)
387 goto err_cancel_link;
389 if (ops->fill_info) {
390 data = nla_nest_start(skb, IFLA_INFO_DATA);
391 if (data == NULL)
392 goto err_cancel_link;
393 err = ops->fill_info(skb, dev);
394 if (err < 0)
395 goto err_cancel_data;
396 nla_nest_end(skb, data);
399 nla_nest_end(skb, linkinfo);
400 return 0;
402 err_cancel_data:
403 nla_nest_cancel(skb, data);
404 err_cancel_link:
405 nla_nest_cancel(skb, linkinfo);
406 out:
407 return err;
410 static const int rtm_min[RTM_NR_FAMILIES] =
412 [RTM_FAM(RTM_NEWLINK)] = NLMSG_LENGTH(sizeof(struct ifinfomsg)),
413 [RTM_FAM(RTM_NEWADDR)] = NLMSG_LENGTH(sizeof(struct ifaddrmsg)),
414 [RTM_FAM(RTM_NEWROUTE)] = NLMSG_LENGTH(sizeof(struct rtmsg)),
415 [RTM_FAM(RTM_NEWRULE)] = NLMSG_LENGTH(sizeof(struct fib_rule_hdr)),
416 [RTM_FAM(RTM_NEWQDISC)] = NLMSG_LENGTH(sizeof(struct tcmsg)),
417 [RTM_FAM(RTM_NEWTCLASS)] = NLMSG_LENGTH(sizeof(struct tcmsg)),
418 [RTM_FAM(RTM_NEWTFILTER)] = NLMSG_LENGTH(sizeof(struct tcmsg)),
419 [RTM_FAM(RTM_NEWACTION)] = NLMSG_LENGTH(sizeof(struct tcamsg)),
420 [RTM_FAM(RTM_GETMULTICAST)] = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
421 [RTM_FAM(RTM_GETANYCAST)] = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
424 static const int rta_max[RTM_NR_FAMILIES] =
426 [RTM_FAM(RTM_NEWLINK)] = IFLA_MAX,
427 [RTM_FAM(RTM_NEWADDR)] = IFA_MAX,
428 [RTM_FAM(RTM_NEWROUTE)] = RTA_MAX,
429 [RTM_FAM(RTM_NEWRULE)] = FRA_MAX,
430 [RTM_FAM(RTM_NEWQDISC)] = TCA_MAX,
431 [RTM_FAM(RTM_NEWTCLASS)] = TCA_MAX,
432 [RTM_FAM(RTM_NEWTFILTER)] = TCA_MAX,
433 [RTM_FAM(RTM_NEWACTION)] = TCAA_MAX,
436 void __rta_fill(struct sk_buff *skb, int attrtype, int attrlen, const void *data)
438 struct rtattr *rta;
439 int size = RTA_LENGTH(attrlen);
441 rta = (struct rtattr *)skb_put(skb, RTA_ALIGN(size));
442 rta->rta_type = attrtype;
443 rta->rta_len = size;
444 memcpy(RTA_DATA(rta), data, attrlen);
445 memset(RTA_DATA(rta) + attrlen, 0, RTA_ALIGN(size) - size);
447 EXPORT_SYMBOL(__rta_fill);
449 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned group, int echo)
451 struct sock *rtnl = net->rtnl;
452 int err = 0;
454 NETLINK_CB(skb).dst_group = group;
455 if (echo)
456 atomic_inc(&skb->users);
457 netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
458 if (echo)
459 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
460 return err;
463 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
465 struct sock *rtnl = net->rtnl;
467 return nlmsg_unicast(rtnl, skb, pid);
469 EXPORT_SYMBOL(rtnl_unicast);
471 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
472 struct nlmsghdr *nlh, gfp_t flags)
474 struct sock *rtnl = net->rtnl;
475 int report = 0;
477 if (nlh)
478 report = nlmsg_report(nlh);
480 nlmsg_notify(rtnl, skb, pid, group, report, flags);
482 EXPORT_SYMBOL(rtnl_notify);
484 void rtnl_set_sk_err(struct net *net, u32 group, int error)
486 struct sock *rtnl = net->rtnl;
488 netlink_set_err(rtnl, 0, group, error);
490 EXPORT_SYMBOL(rtnl_set_sk_err);
492 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
494 struct nlattr *mx;
495 int i, valid = 0;
497 mx = nla_nest_start(skb, RTA_METRICS);
498 if (mx == NULL)
499 return -ENOBUFS;
501 for (i = 0; i < RTAX_MAX; i++) {
502 if (metrics[i]) {
503 valid++;
504 NLA_PUT_U32(skb, i+1, metrics[i]);
508 if (!valid) {
509 nla_nest_cancel(skb, mx);
510 return 0;
513 return nla_nest_end(skb, mx);
515 nla_put_failure:
516 nla_nest_cancel(skb, mx);
517 return -EMSGSIZE;
519 EXPORT_SYMBOL(rtnetlink_put_metrics);
521 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
522 u32 ts, u32 tsage, long expires, u32 error)
524 struct rta_cacheinfo ci = {
525 .rta_lastuse = jiffies_to_clock_t(jiffies - dst->lastuse),
526 .rta_used = dst->__use,
527 .rta_clntref = atomic_read(&(dst->__refcnt)),
528 .rta_error = error,
529 .rta_id = id,
530 .rta_ts = ts,
531 .rta_tsage = tsage,
534 if (expires)
535 ci.rta_expires = jiffies_to_clock_t(expires);
537 return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
539 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
541 static void set_operstate(struct net_device *dev, unsigned char transition)
543 unsigned char operstate = dev->operstate;
545 switch (transition) {
546 case IF_OPER_UP:
547 if ((operstate == IF_OPER_DORMANT ||
548 operstate == IF_OPER_UNKNOWN) &&
549 !netif_dormant(dev))
550 operstate = IF_OPER_UP;
551 break;
553 case IF_OPER_DORMANT:
554 if (operstate == IF_OPER_UP ||
555 operstate == IF_OPER_UNKNOWN)
556 operstate = IF_OPER_DORMANT;
557 break;
560 if (dev->operstate != operstate) {
561 write_lock_bh(&dev_base_lock);
562 dev->operstate = operstate;
563 write_unlock_bh(&dev_base_lock);
564 netdev_state_change(dev);
568 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
569 const struct ifinfomsg *ifm)
571 unsigned int flags = ifm->ifi_flags;
573 /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
574 if (ifm->ifi_change)
575 flags = (flags & ifm->ifi_change) |
576 (dev->flags & ~ifm->ifi_change);
578 return flags;
581 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
582 const struct net_device_stats *b)
584 a->rx_packets = b->rx_packets;
585 a->tx_packets = b->tx_packets;
586 a->rx_bytes = b->rx_bytes;
587 a->tx_bytes = b->tx_bytes;
588 a->rx_errors = b->rx_errors;
589 a->tx_errors = b->tx_errors;
590 a->rx_dropped = b->rx_dropped;
591 a->tx_dropped = b->tx_dropped;
593 a->multicast = b->multicast;
594 a->collisions = b->collisions;
596 a->rx_length_errors = b->rx_length_errors;
597 a->rx_over_errors = b->rx_over_errors;
598 a->rx_crc_errors = b->rx_crc_errors;
599 a->rx_frame_errors = b->rx_frame_errors;
600 a->rx_fifo_errors = b->rx_fifo_errors;
601 a->rx_missed_errors = b->rx_missed_errors;
603 a->tx_aborted_errors = b->tx_aborted_errors;
604 a->tx_carrier_errors = b->tx_carrier_errors;
605 a->tx_fifo_errors = b->tx_fifo_errors;
606 a->tx_heartbeat_errors = b->tx_heartbeat_errors;
607 a->tx_window_errors = b->tx_window_errors;
609 a->rx_compressed = b->rx_compressed;
610 a->tx_compressed = b->tx_compressed;
613 static void copy_rtnl_link_stats64(void *v, const struct net_device_stats *b)
615 struct rtnl_link_stats64 a;
617 a.rx_packets = b->rx_packets;
618 a.tx_packets = b->tx_packets;
619 a.rx_bytes = b->rx_bytes;
620 a.tx_bytes = b->tx_bytes;
621 a.rx_errors = b->rx_errors;
622 a.tx_errors = b->tx_errors;
623 a.rx_dropped = b->rx_dropped;
624 a.tx_dropped = b->tx_dropped;
626 a.multicast = b->multicast;
627 a.collisions = b->collisions;
629 a.rx_length_errors = b->rx_length_errors;
630 a.rx_over_errors = b->rx_over_errors;
631 a.rx_crc_errors = b->rx_crc_errors;
632 a.rx_frame_errors = b->rx_frame_errors;
633 a.rx_fifo_errors = b->rx_fifo_errors;
634 a.rx_missed_errors = b->rx_missed_errors;
636 a.tx_aborted_errors = b->tx_aborted_errors;
637 a.tx_carrier_errors = b->tx_carrier_errors;
638 a.tx_fifo_errors = b->tx_fifo_errors;
639 a.tx_heartbeat_errors = b->tx_heartbeat_errors;
640 a.tx_window_errors = b->tx_window_errors;
642 a.rx_compressed = b->rx_compressed;
643 a.tx_compressed = b->tx_compressed;
644 memcpy(v, &a, sizeof(a));
647 /* All VF info */
648 static inline int rtnl_vfinfo_size(const struct net_device *dev)
650 if (dev->dev.parent && dev_is_pci(dev->dev.parent)) {
652 int num_vfs = dev_num_vf(dev->dev.parent);
653 size_t size = nla_total_size(sizeof(struct nlattr));
654 size += nla_total_size(num_vfs * sizeof(struct nlattr));
655 size += num_vfs *
656 (nla_total_size(sizeof(struct ifla_vf_mac)) +
657 nla_total_size(sizeof(struct ifla_vf_vlan)) +
658 nla_total_size(sizeof(struct ifla_vf_tx_rate)));
659 return size;
660 } else
661 return 0;
664 static size_t rtnl_port_size(const struct net_device *dev)
666 size_t port_size = nla_total_size(4) /* PORT_VF */
667 + nla_total_size(PORT_PROFILE_MAX) /* PORT_PROFILE */
668 + nla_total_size(sizeof(struct ifla_port_vsi))
669 /* PORT_VSI_TYPE */
670 + nla_total_size(PORT_UUID_MAX) /* PORT_INSTANCE_UUID */
671 + nla_total_size(PORT_UUID_MAX) /* PORT_HOST_UUID */
672 + nla_total_size(1) /* PROT_VDP_REQUEST */
673 + nla_total_size(2); /* PORT_VDP_RESPONSE */
674 size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
675 size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
676 + port_size;
677 size_t port_self_size = nla_total_size(sizeof(struct nlattr))
678 + port_size;
680 if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent)
681 return 0;
682 if (dev_num_vf(dev->dev.parent))
683 return port_self_size + vf_ports_size +
684 vf_port_size * dev_num_vf(dev->dev.parent);
685 else
686 return port_self_size;
689 static inline size_t if_nlmsg_size(const struct net_device *dev)
691 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
692 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
693 + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
694 + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
695 + nla_total_size(sizeof(struct rtnl_link_ifmap))
696 + nla_total_size(sizeof(struct rtnl_link_stats))
697 + nla_total_size(sizeof(struct rtnl_link_stats64))
698 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
699 + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
700 + nla_total_size(4) /* IFLA_TXQLEN */
701 + nla_total_size(4) /* IFLA_WEIGHT */
702 + nla_total_size(4) /* IFLA_MTU */
703 + nla_total_size(4) /* IFLA_LINK */
704 + nla_total_size(4) /* IFLA_MASTER */
705 + nla_total_size(1) /* IFLA_OPERSTATE */
706 + nla_total_size(1) /* IFLA_LINKMODE */
707 + nla_total_size(4) /* IFLA_NUM_VF */
708 + rtnl_vfinfo_size(dev) /* IFLA_VFINFO_LIST */
709 + rtnl_port_size(dev) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
710 + rtnl_link_get_size(dev); /* IFLA_LINKINFO */
713 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
715 struct nlattr *vf_ports;
716 struct nlattr *vf_port;
717 int vf;
718 int err;
720 vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
721 if (!vf_ports)
722 return -EMSGSIZE;
724 for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
725 vf_port = nla_nest_start(skb, IFLA_VF_PORT);
726 if (!vf_port)
727 goto nla_put_failure;
728 NLA_PUT_U32(skb, IFLA_PORT_VF, vf);
729 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
730 if (err == -EMSGSIZE)
731 goto nla_put_failure;
732 if (err) {
733 nla_nest_cancel(skb, vf_port);
734 continue;
736 nla_nest_end(skb, vf_port);
739 nla_nest_end(skb, vf_ports);
741 return 0;
743 nla_put_failure:
744 nla_nest_cancel(skb, vf_ports);
745 return -EMSGSIZE;
748 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
750 struct nlattr *port_self;
751 int err;
753 port_self = nla_nest_start(skb, IFLA_PORT_SELF);
754 if (!port_self)
755 return -EMSGSIZE;
757 err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
758 if (err) {
759 nla_nest_cancel(skb, port_self);
760 return (err == -EMSGSIZE) ? err : 0;
763 nla_nest_end(skb, port_self);
765 return 0;
768 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev)
770 int err;
772 if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent)
773 return 0;
775 err = rtnl_port_self_fill(skb, dev);
776 if (err)
777 return err;
779 if (dev_num_vf(dev->dev.parent)) {
780 err = rtnl_vf_ports_fill(skb, dev);
781 if (err)
782 return err;
785 return 0;
788 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
789 int type, u32 pid, u32 seq, u32 change,
790 unsigned int flags)
792 struct ifinfomsg *ifm;
793 struct nlmsghdr *nlh;
794 const struct net_device_stats *stats;
795 struct nlattr *attr;
797 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
798 if (nlh == NULL)
799 return -EMSGSIZE;
801 ifm = nlmsg_data(nlh);
802 ifm->ifi_family = AF_UNSPEC;
803 ifm->__ifi_pad = 0;
804 ifm->ifi_type = dev->type;
805 ifm->ifi_index = dev->ifindex;
806 ifm->ifi_flags = dev_get_flags(dev);
807 ifm->ifi_change = change;
809 NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
810 NLA_PUT_U32(skb, IFLA_TXQLEN, dev->tx_queue_len);
811 NLA_PUT_U8(skb, IFLA_OPERSTATE,
812 netif_running(dev) ? dev->operstate : IF_OPER_DOWN);
813 NLA_PUT_U8(skb, IFLA_LINKMODE, dev->link_mode);
814 NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
816 if (dev->ifindex != dev->iflink)
817 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
819 if (dev->master)
820 NLA_PUT_U32(skb, IFLA_MASTER, dev->master->ifindex);
822 if (dev->qdisc)
823 NLA_PUT_STRING(skb, IFLA_QDISC, dev->qdisc->ops->id);
825 if (dev->ifalias)
826 NLA_PUT_STRING(skb, IFLA_IFALIAS, dev->ifalias);
828 if (1) {
829 struct rtnl_link_ifmap map = {
830 .mem_start = dev->mem_start,
831 .mem_end = dev->mem_end,
832 .base_addr = dev->base_addr,
833 .irq = dev->irq,
834 .dma = dev->dma,
835 .port = dev->if_port,
837 NLA_PUT(skb, IFLA_MAP, sizeof(map), &map);
840 if (dev->addr_len) {
841 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
842 NLA_PUT(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast);
845 attr = nla_reserve(skb, IFLA_STATS,
846 sizeof(struct rtnl_link_stats));
847 if (attr == NULL)
848 goto nla_put_failure;
850 stats = dev_get_stats(dev);
851 copy_rtnl_link_stats(nla_data(attr), stats);
853 attr = nla_reserve(skb, IFLA_STATS64,
854 sizeof(struct rtnl_link_stats64));
855 if (attr == NULL)
856 goto nla_put_failure;
857 copy_rtnl_link_stats64(nla_data(attr), stats);
859 if (dev->dev.parent)
860 NLA_PUT_U32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent));
862 if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent) {
863 int i;
865 struct nlattr *vfinfo, *vf;
866 int num_vfs = dev_num_vf(dev->dev.parent);
868 vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
869 if (!vfinfo)
870 goto nla_put_failure;
871 for (i = 0; i < num_vfs; i++) {
872 struct ifla_vf_info ivi;
873 struct ifla_vf_mac vf_mac;
874 struct ifla_vf_vlan vf_vlan;
875 struct ifla_vf_tx_rate vf_tx_rate;
876 if (dev->netdev_ops->ndo_get_vf_config(dev, i, &ivi))
877 break;
878 vf_mac.vf = vf_vlan.vf = vf_tx_rate.vf = ivi.vf;
879 memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
880 vf_vlan.vlan = ivi.vlan;
881 vf_vlan.qos = ivi.qos;
882 vf_tx_rate.rate = ivi.tx_rate;
883 vf = nla_nest_start(skb, IFLA_VF_INFO);
884 if (!vf) {
885 nla_nest_cancel(skb, vfinfo);
886 goto nla_put_failure;
888 NLA_PUT(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac);
889 NLA_PUT(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan);
890 NLA_PUT(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate), &vf_tx_rate);
891 nla_nest_end(skb, vf);
893 nla_nest_end(skb, vfinfo);
896 if (rtnl_port_fill(skb, dev))
897 goto nla_put_failure;
899 if (dev->rtnl_link_ops) {
900 if (rtnl_link_fill(skb, dev) < 0)
901 goto nla_put_failure;
904 return nlmsg_end(skb, nlh);
906 nla_put_failure:
907 nlmsg_cancel(skb, nlh);
908 return -EMSGSIZE;
911 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
913 struct net *net = sock_net(skb->sk);
914 int h, s_h;
915 int idx = 0, s_idx;
916 struct net_device *dev;
917 struct hlist_head *head;
918 struct hlist_node *node;
920 s_h = cb->args[0];
921 s_idx = cb->args[1];
923 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
924 idx = 0;
925 head = &net->dev_index_head[h];
926 hlist_for_each_entry(dev, node, head, index_hlist) {
927 if (idx < s_idx)
928 goto cont;
929 if (rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
930 NETLINK_CB(cb->skb).pid,
931 cb->nlh->nlmsg_seq, 0,
932 NLM_F_MULTI) <= 0)
933 goto out;
934 cont:
935 idx++;
938 out:
939 cb->args[1] = idx;
940 cb->args[0] = h;
942 return skb->len;
945 const struct nla_policy ifla_policy[IFLA_MAX+1] = {
946 [IFLA_IFNAME] = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
947 [IFLA_ADDRESS] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
948 [IFLA_BROADCAST] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
949 [IFLA_MAP] = { .len = sizeof(struct rtnl_link_ifmap) },
950 [IFLA_MTU] = { .type = NLA_U32 },
951 [IFLA_LINK] = { .type = NLA_U32 },
952 [IFLA_TXQLEN] = { .type = NLA_U32 },
953 [IFLA_WEIGHT] = { .type = NLA_U32 },
954 [IFLA_OPERSTATE] = { .type = NLA_U8 },
955 [IFLA_LINKMODE] = { .type = NLA_U8 },
956 [IFLA_LINKINFO] = { .type = NLA_NESTED },
957 [IFLA_NET_NS_PID] = { .type = NLA_U32 },
958 [IFLA_IFALIAS] = { .type = NLA_STRING, .len = IFALIASZ-1 },
959 [IFLA_VFINFO_LIST] = {. type = NLA_NESTED },
960 [IFLA_VF_PORTS] = { .type = NLA_NESTED },
961 [IFLA_PORT_SELF] = { .type = NLA_NESTED },
963 EXPORT_SYMBOL(ifla_policy);
965 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
966 [IFLA_INFO_KIND] = { .type = NLA_STRING },
967 [IFLA_INFO_DATA] = { .type = NLA_NESTED },
970 static const struct nla_policy ifla_vfinfo_policy[IFLA_VF_INFO_MAX+1] = {
971 [IFLA_VF_INFO] = { .type = NLA_NESTED },
974 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
975 [IFLA_VF_MAC] = { .type = NLA_BINARY,
976 .len = sizeof(struct ifla_vf_mac) },
977 [IFLA_VF_VLAN] = { .type = NLA_BINARY,
978 .len = sizeof(struct ifla_vf_vlan) },
979 [IFLA_VF_TX_RATE] = { .type = NLA_BINARY,
980 .len = sizeof(struct ifla_vf_tx_rate) },
983 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
984 [IFLA_PORT_VF] = { .type = NLA_U32 },
985 [IFLA_PORT_PROFILE] = { .type = NLA_STRING,
986 .len = PORT_PROFILE_MAX },
987 [IFLA_PORT_VSI_TYPE] = { .type = NLA_BINARY,
988 .len = sizeof(struct ifla_port_vsi)},
989 [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
990 .len = PORT_UUID_MAX },
991 [IFLA_PORT_HOST_UUID] = { .type = NLA_STRING,
992 .len = PORT_UUID_MAX },
993 [IFLA_PORT_REQUEST] = { .type = NLA_U8, },
994 [IFLA_PORT_RESPONSE] = { .type = NLA_U16, },
997 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
999 struct net *net;
1000 /* Examine the link attributes and figure out which
1001 * network namespace we are talking about.
1003 if (tb[IFLA_NET_NS_PID])
1004 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1005 else
1006 net = get_net(src_net);
1007 return net;
1009 EXPORT_SYMBOL(rtnl_link_get_net);
1011 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1013 if (dev) {
1014 if (tb[IFLA_ADDRESS] &&
1015 nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1016 return -EINVAL;
1018 if (tb[IFLA_BROADCAST] &&
1019 nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1020 return -EINVAL;
1023 return 0;
1026 static int do_setvfinfo(struct net_device *dev, struct nlattr *attr)
1028 int rem, err = -EINVAL;
1029 struct nlattr *vf;
1030 const struct net_device_ops *ops = dev->netdev_ops;
1032 nla_for_each_nested(vf, attr, rem) {
1033 switch (nla_type(vf)) {
1034 case IFLA_VF_MAC: {
1035 struct ifla_vf_mac *ivm;
1036 ivm = nla_data(vf);
1037 err = -EOPNOTSUPP;
1038 if (ops->ndo_set_vf_mac)
1039 err = ops->ndo_set_vf_mac(dev, ivm->vf,
1040 ivm->mac);
1041 break;
1043 case IFLA_VF_VLAN: {
1044 struct ifla_vf_vlan *ivv;
1045 ivv = nla_data(vf);
1046 err = -EOPNOTSUPP;
1047 if (ops->ndo_set_vf_vlan)
1048 err = ops->ndo_set_vf_vlan(dev, ivv->vf,
1049 ivv->vlan,
1050 ivv->qos);
1051 break;
1053 case IFLA_VF_TX_RATE: {
1054 struct ifla_vf_tx_rate *ivt;
1055 ivt = nla_data(vf);
1056 err = -EOPNOTSUPP;
1057 if (ops->ndo_set_vf_tx_rate)
1058 err = ops->ndo_set_vf_tx_rate(dev, ivt->vf,
1059 ivt->rate);
1060 break;
1062 default:
1063 err = -EINVAL;
1064 break;
1066 if (err)
1067 break;
1069 return err;
1072 static int do_setlink(struct net_device *dev, struct ifinfomsg *ifm,
1073 struct nlattr **tb, char *ifname, int modified)
1075 const struct net_device_ops *ops = dev->netdev_ops;
1076 int send_addr_notify = 0;
1077 int err;
1079 if (tb[IFLA_NET_NS_PID]) {
1080 struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1081 if (IS_ERR(net)) {
1082 err = PTR_ERR(net);
1083 goto errout;
1085 err = dev_change_net_namespace(dev, net, ifname);
1086 put_net(net);
1087 if (err)
1088 goto errout;
1089 modified = 1;
1092 if (tb[IFLA_MAP]) {
1093 struct rtnl_link_ifmap *u_map;
1094 struct ifmap k_map;
1096 if (!ops->ndo_set_config) {
1097 err = -EOPNOTSUPP;
1098 goto errout;
1101 if (!netif_device_present(dev)) {
1102 err = -ENODEV;
1103 goto errout;
1106 u_map = nla_data(tb[IFLA_MAP]);
1107 k_map.mem_start = (unsigned long) u_map->mem_start;
1108 k_map.mem_end = (unsigned long) u_map->mem_end;
1109 k_map.base_addr = (unsigned short) u_map->base_addr;
1110 k_map.irq = (unsigned char) u_map->irq;
1111 k_map.dma = (unsigned char) u_map->dma;
1112 k_map.port = (unsigned char) u_map->port;
1114 err = ops->ndo_set_config(dev, &k_map);
1115 if (err < 0)
1116 goto errout;
1118 modified = 1;
1121 if (tb[IFLA_ADDRESS]) {
1122 struct sockaddr *sa;
1123 int len;
1125 if (!ops->ndo_set_mac_address) {
1126 err = -EOPNOTSUPP;
1127 goto errout;
1130 if (!netif_device_present(dev)) {
1131 err = -ENODEV;
1132 goto errout;
1135 len = sizeof(sa_family_t) + dev->addr_len;
1136 sa = kmalloc(len, GFP_KERNEL);
1137 if (!sa) {
1138 err = -ENOMEM;
1139 goto errout;
1141 sa->sa_family = dev->type;
1142 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1143 dev->addr_len);
1144 err = ops->ndo_set_mac_address(dev, sa);
1145 kfree(sa);
1146 if (err)
1147 goto errout;
1148 send_addr_notify = 1;
1149 modified = 1;
1152 if (tb[IFLA_MTU]) {
1153 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1154 if (err < 0)
1155 goto errout;
1156 modified = 1;
1160 * Interface selected by interface index but interface
1161 * name provided implies that a name change has been
1162 * requested.
1164 if (ifm->ifi_index > 0 && ifname[0]) {
1165 err = dev_change_name(dev, ifname);
1166 if (err < 0)
1167 goto errout;
1168 modified = 1;
1171 if (tb[IFLA_IFALIAS]) {
1172 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
1173 nla_len(tb[IFLA_IFALIAS]));
1174 if (err < 0)
1175 goto errout;
1176 modified = 1;
1179 if (tb[IFLA_BROADCAST]) {
1180 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
1181 send_addr_notify = 1;
1184 if (ifm->ifi_flags || ifm->ifi_change) {
1185 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1186 if (err < 0)
1187 goto errout;
1190 if (tb[IFLA_TXQLEN])
1191 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1193 if (tb[IFLA_OPERSTATE])
1194 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1196 if (tb[IFLA_LINKMODE]) {
1197 write_lock_bh(&dev_base_lock);
1198 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1199 write_unlock_bh(&dev_base_lock);
1202 if (tb[IFLA_VFINFO_LIST]) {
1203 struct nlattr *attr;
1204 int rem;
1205 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
1206 if (nla_type(attr) != IFLA_VF_INFO) {
1207 err = -EINVAL;
1208 goto errout;
1210 err = do_setvfinfo(dev, attr);
1211 if (err < 0)
1212 goto errout;
1213 modified = 1;
1216 err = 0;
1218 if (tb[IFLA_VF_PORTS]) {
1219 struct nlattr *port[IFLA_PORT_MAX+1];
1220 struct nlattr *attr;
1221 int vf;
1222 int rem;
1224 err = -EOPNOTSUPP;
1225 if (!ops->ndo_set_vf_port)
1226 goto errout;
1228 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
1229 if (nla_type(attr) != IFLA_VF_PORT)
1230 continue;
1231 err = nla_parse_nested(port, IFLA_PORT_MAX,
1232 attr, ifla_port_policy);
1233 if (err < 0)
1234 goto errout;
1235 if (!port[IFLA_PORT_VF]) {
1236 err = -EOPNOTSUPP;
1237 goto errout;
1239 vf = nla_get_u32(port[IFLA_PORT_VF]);
1240 err = ops->ndo_set_vf_port(dev, vf, port);
1241 if (err < 0)
1242 goto errout;
1243 modified = 1;
1246 err = 0;
1248 if (tb[IFLA_PORT_SELF]) {
1249 struct nlattr *port[IFLA_PORT_MAX+1];
1251 err = nla_parse_nested(port, IFLA_PORT_MAX,
1252 tb[IFLA_PORT_SELF], ifla_port_policy);
1253 if (err < 0)
1254 goto errout;
1256 err = -EOPNOTSUPP;
1257 if (ops->ndo_set_vf_port)
1258 err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
1259 if (err < 0)
1260 goto errout;
1261 modified = 1;
1263 err = 0;
1265 errout:
1266 if (err < 0 && modified && net_ratelimit())
1267 printk(KERN_WARNING "A link change request failed with "
1268 "some changes comitted already. Interface %s may "
1269 "have been left with an inconsistent configuration, "
1270 "please check.\n", dev->name);
1272 if (send_addr_notify)
1273 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
1274 return err;
1277 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1279 struct net *net = sock_net(skb->sk);
1280 struct ifinfomsg *ifm;
1281 struct net_device *dev;
1282 int err;
1283 struct nlattr *tb[IFLA_MAX+1];
1284 char ifname[IFNAMSIZ];
1286 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1287 if (err < 0)
1288 goto errout;
1290 if (tb[IFLA_IFNAME])
1291 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1292 else
1293 ifname[0] = '\0';
1295 err = -EINVAL;
1296 ifm = nlmsg_data(nlh);
1297 if (ifm->ifi_index > 0)
1298 dev = __dev_get_by_index(net, ifm->ifi_index);
1299 else if (tb[IFLA_IFNAME])
1300 dev = __dev_get_by_name(net, ifname);
1301 else
1302 goto errout;
1304 if (dev == NULL) {
1305 err = -ENODEV;
1306 goto errout;
1309 err = validate_linkmsg(dev, tb);
1310 if (err < 0)
1311 goto errout;
1313 err = do_setlink(dev, ifm, tb, ifname, 0);
1314 errout:
1315 return err;
1318 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1320 struct net *net = sock_net(skb->sk);
1321 const struct rtnl_link_ops *ops;
1322 struct net_device *dev;
1323 struct ifinfomsg *ifm;
1324 char ifname[IFNAMSIZ];
1325 struct nlattr *tb[IFLA_MAX+1];
1326 int err;
1328 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1329 if (err < 0)
1330 return err;
1332 if (tb[IFLA_IFNAME])
1333 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1335 ifm = nlmsg_data(nlh);
1336 if (ifm->ifi_index > 0)
1337 dev = __dev_get_by_index(net, ifm->ifi_index);
1338 else if (tb[IFLA_IFNAME])
1339 dev = __dev_get_by_name(net, ifname);
1340 else
1341 return -EINVAL;
1343 if (!dev)
1344 return -ENODEV;
1346 ops = dev->rtnl_link_ops;
1347 if (!ops)
1348 return -EOPNOTSUPP;
1350 ops->dellink(dev, NULL);
1351 return 0;
1354 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
1356 unsigned int old_flags;
1357 int err;
1359 old_flags = dev->flags;
1360 if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
1361 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1362 if (err < 0)
1363 return err;
1366 dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
1367 rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U);
1369 __dev_notify_flags(dev, old_flags);
1370 return 0;
1372 EXPORT_SYMBOL(rtnl_configure_link);
1374 struct net_device *rtnl_create_link(struct net *src_net, struct net *net,
1375 char *ifname, const struct rtnl_link_ops *ops, struct nlattr *tb[])
1377 int err;
1378 struct net_device *dev;
1379 unsigned int num_queues = 1;
1380 unsigned int real_num_queues = 1;
1382 if (ops->get_tx_queues) {
1383 err = ops->get_tx_queues(src_net, tb, &num_queues,
1384 &real_num_queues);
1385 if (err)
1386 goto err;
1388 err = -ENOMEM;
1389 dev = alloc_netdev_mq(ops->priv_size, ifname, ops->setup, num_queues);
1390 if (!dev)
1391 goto err;
1393 dev_net_set(dev, net);
1394 dev->rtnl_link_ops = ops;
1395 dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
1396 dev->real_num_tx_queues = real_num_queues;
1398 if (strchr(dev->name, '%')) {
1399 err = dev_alloc_name(dev, dev->name);
1400 if (err < 0)
1401 goto err_free;
1404 if (tb[IFLA_MTU])
1405 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
1406 if (tb[IFLA_ADDRESS])
1407 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
1408 nla_len(tb[IFLA_ADDRESS]));
1409 if (tb[IFLA_BROADCAST])
1410 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
1411 nla_len(tb[IFLA_BROADCAST]));
1412 if (tb[IFLA_TXQLEN])
1413 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1414 if (tb[IFLA_OPERSTATE])
1415 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1416 if (tb[IFLA_LINKMODE])
1417 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1419 return dev;
1421 err_free:
1422 free_netdev(dev);
1423 err:
1424 return ERR_PTR(err);
1426 EXPORT_SYMBOL(rtnl_create_link);
1428 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1430 struct net *net = sock_net(skb->sk);
1431 const struct rtnl_link_ops *ops;
1432 struct net_device *dev;
1433 struct ifinfomsg *ifm;
1434 char kind[MODULE_NAME_LEN];
1435 char ifname[IFNAMSIZ];
1436 struct nlattr *tb[IFLA_MAX+1];
1437 struct nlattr *linkinfo[IFLA_INFO_MAX+1];
1438 int err;
1440 #ifdef CONFIG_MODULES
1441 replay:
1442 #endif
1443 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1444 if (err < 0)
1445 return err;
1447 if (tb[IFLA_IFNAME])
1448 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1449 else
1450 ifname[0] = '\0';
1452 ifm = nlmsg_data(nlh);
1453 if (ifm->ifi_index > 0)
1454 dev = __dev_get_by_index(net, ifm->ifi_index);
1455 else if (ifname[0])
1456 dev = __dev_get_by_name(net, ifname);
1457 else
1458 dev = NULL;
1460 err = validate_linkmsg(dev, tb);
1461 if (err < 0)
1462 return err;
1464 if (tb[IFLA_LINKINFO]) {
1465 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
1466 tb[IFLA_LINKINFO], ifla_info_policy);
1467 if (err < 0)
1468 return err;
1469 } else
1470 memset(linkinfo, 0, sizeof(linkinfo));
1472 if (linkinfo[IFLA_INFO_KIND]) {
1473 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
1474 ops = rtnl_link_ops_get(kind);
1475 } else {
1476 kind[0] = '\0';
1477 ops = NULL;
1480 if (1) {
1481 struct nlattr *attr[ops ? ops->maxtype + 1 : 0], **data = NULL;
1482 struct net *dest_net;
1484 if (ops) {
1485 if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
1486 err = nla_parse_nested(attr, ops->maxtype,
1487 linkinfo[IFLA_INFO_DATA],
1488 ops->policy);
1489 if (err < 0)
1490 return err;
1491 data = attr;
1493 if (ops->validate) {
1494 err = ops->validate(tb, data);
1495 if (err < 0)
1496 return err;
1500 if (dev) {
1501 int modified = 0;
1503 if (nlh->nlmsg_flags & NLM_F_EXCL)
1504 return -EEXIST;
1505 if (nlh->nlmsg_flags & NLM_F_REPLACE)
1506 return -EOPNOTSUPP;
1508 if (linkinfo[IFLA_INFO_DATA]) {
1509 if (!ops || ops != dev->rtnl_link_ops ||
1510 !ops->changelink)
1511 return -EOPNOTSUPP;
1513 err = ops->changelink(dev, tb, data);
1514 if (err < 0)
1515 return err;
1516 modified = 1;
1519 return do_setlink(dev, ifm, tb, ifname, modified);
1522 if (!(nlh->nlmsg_flags & NLM_F_CREATE))
1523 return -ENODEV;
1525 if (ifm->ifi_index)
1526 return -EOPNOTSUPP;
1527 if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
1528 return -EOPNOTSUPP;
1530 if (!ops) {
1531 #ifdef CONFIG_MODULES
1532 if (kind[0]) {
1533 __rtnl_unlock();
1534 request_module("rtnl-link-%s", kind);
1535 rtnl_lock();
1536 ops = rtnl_link_ops_get(kind);
1537 if (ops)
1538 goto replay;
1540 #endif
1541 return -EOPNOTSUPP;
1544 if (!ifname[0])
1545 snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
1547 dest_net = rtnl_link_get_net(net, tb);
1548 dev = rtnl_create_link(net, dest_net, ifname, ops, tb);
1550 if (IS_ERR(dev))
1551 err = PTR_ERR(dev);
1552 else if (ops->newlink)
1553 err = ops->newlink(net, dev, tb, data);
1554 else
1555 err = register_netdevice(dev);
1557 if (err < 0 && !IS_ERR(dev))
1558 free_netdev(dev);
1559 if (err < 0)
1560 goto out;
1562 err = rtnl_configure_link(dev, ifm);
1563 if (err < 0)
1564 unregister_netdevice(dev);
1565 out:
1566 put_net(dest_net);
1567 return err;
1571 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
1573 struct net *net = sock_net(skb->sk);
1574 struct ifinfomsg *ifm;
1575 char ifname[IFNAMSIZ];
1576 struct nlattr *tb[IFLA_MAX+1];
1577 struct net_device *dev = NULL;
1578 struct sk_buff *nskb;
1579 int err;
1581 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1582 if (err < 0)
1583 return err;
1585 if (tb[IFLA_IFNAME])
1586 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1588 ifm = nlmsg_data(nlh);
1589 if (ifm->ifi_index > 0)
1590 dev = __dev_get_by_index(net, ifm->ifi_index);
1591 else if (tb[IFLA_IFNAME])
1592 dev = __dev_get_by_name(net, ifname);
1593 else
1594 return -EINVAL;
1596 if (dev == NULL)
1597 return -ENODEV;
1599 nskb = nlmsg_new(if_nlmsg_size(dev), GFP_KERNEL);
1600 if (nskb == NULL)
1601 return -ENOBUFS;
1603 err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).pid,
1604 nlh->nlmsg_seq, 0, 0);
1605 if (err < 0) {
1606 /* -EMSGSIZE implies BUG in if_nlmsg_size */
1607 WARN_ON(err == -EMSGSIZE);
1608 kfree_skb(nskb);
1609 } else
1610 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).pid);
1612 return err;
1615 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
1617 int idx;
1618 int s_idx = cb->family;
1620 if (s_idx == 0)
1621 s_idx = 1;
1622 for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
1623 int type = cb->nlh->nlmsg_type-RTM_BASE;
1624 if (idx < s_idx || idx == PF_PACKET)
1625 continue;
1626 if (rtnl_msg_handlers[idx] == NULL ||
1627 rtnl_msg_handlers[idx][type].dumpit == NULL)
1628 continue;
1629 if (idx > s_idx)
1630 memset(&cb->args[0], 0, sizeof(cb->args));
1631 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
1632 break;
1634 cb->family = idx;
1636 return skb->len;
1639 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned change)
1641 struct net *net = dev_net(dev);
1642 struct sk_buff *skb;
1643 int err = -ENOBUFS;
1645 skb = nlmsg_new(if_nlmsg_size(dev), GFP_KERNEL);
1646 if (skb == NULL)
1647 goto errout;
1649 err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0);
1650 if (err < 0) {
1651 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
1652 WARN_ON(err == -EMSGSIZE);
1653 kfree_skb(skb);
1654 goto errout;
1656 rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_KERNEL);
1657 return;
1658 errout:
1659 if (err < 0)
1660 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
1663 /* Protected by RTNL sempahore. */
1664 static struct rtattr **rta_buf;
1665 static int rtattr_max;
1667 /* Process one rtnetlink message. */
1669 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
1671 struct net *net = sock_net(skb->sk);
1672 rtnl_doit_func doit;
1673 int sz_idx, kind;
1674 int min_len;
1675 int family;
1676 int type;
1677 int err;
1679 type = nlh->nlmsg_type;
1680 if (type > RTM_MAX)
1681 return -EOPNOTSUPP;
1683 type -= RTM_BASE;
1685 /* All the messages must have at least 1 byte length */
1686 if (nlh->nlmsg_len < NLMSG_LENGTH(sizeof(struct rtgenmsg)))
1687 return 0;
1689 family = ((struct rtgenmsg *)NLMSG_DATA(nlh))->rtgen_family;
1690 sz_idx = type>>2;
1691 kind = type&3;
1693 if (kind != 2 && security_netlink_recv(skb, CAP_NET_ADMIN))
1694 return -EPERM;
1696 if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
1697 struct sock *rtnl;
1698 rtnl_dumpit_func dumpit;
1700 dumpit = rtnl_get_dumpit(family, type);
1701 if (dumpit == NULL)
1702 return -EOPNOTSUPP;
1704 __rtnl_unlock();
1705 rtnl = net->rtnl;
1706 err = netlink_dump_start(rtnl, skb, nlh, dumpit, NULL);
1707 rtnl_lock();
1708 return err;
1711 memset(rta_buf, 0, (rtattr_max * sizeof(struct rtattr *)));
1713 min_len = rtm_min[sz_idx];
1714 if (nlh->nlmsg_len < min_len)
1715 return -EINVAL;
1717 if (nlh->nlmsg_len > min_len) {
1718 int attrlen = nlh->nlmsg_len - NLMSG_ALIGN(min_len);
1719 struct rtattr *attr = (void *)nlh + NLMSG_ALIGN(min_len);
1721 while (RTA_OK(attr, attrlen)) {
1722 unsigned flavor = attr->rta_type;
1723 if (flavor) {
1724 if (flavor > rta_max[sz_idx])
1725 return -EINVAL;
1726 rta_buf[flavor-1] = attr;
1728 attr = RTA_NEXT(attr, attrlen);
1732 doit = rtnl_get_doit(family, type);
1733 if (doit == NULL)
1734 return -EOPNOTSUPP;
1736 return doit(skb, nlh, (void *)&rta_buf[0]);
1739 static void rtnetlink_rcv(struct sk_buff *skb)
1741 rtnl_lock();
1742 netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
1743 rtnl_unlock();
1746 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
1748 struct net_device *dev = ptr;
1750 switch (event) {
1751 case NETDEV_UP:
1752 case NETDEV_DOWN:
1753 case NETDEV_PRE_UP:
1754 case NETDEV_POST_INIT:
1755 case NETDEV_REGISTER:
1756 case NETDEV_CHANGE:
1757 case NETDEV_PRE_TYPE_CHANGE:
1758 case NETDEV_GOING_DOWN:
1759 case NETDEV_UNREGISTER:
1760 case NETDEV_UNREGISTER_BATCH:
1761 break;
1762 default:
1763 rtmsg_ifinfo(RTM_NEWLINK, dev, 0);
1764 break;
1766 return NOTIFY_DONE;
1769 static struct notifier_block rtnetlink_dev_notifier = {
1770 .notifier_call = rtnetlink_event,
1774 static int __net_init rtnetlink_net_init(struct net *net)
1776 struct sock *sk;
1777 sk = netlink_kernel_create(net, NETLINK_ROUTE, RTNLGRP_MAX,
1778 rtnetlink_rcv, &rtnl_mutex, THIS_MODULE);
1779 if (!sk)
1780 return -ENOMEM;
1781 net->rtnl = sk;
1782 return 0;
1785 static void __net_exit rtnetlink_net_exit(struct net *net)
1787 netlink_kernel_release(net->rtnl);
1788 net->rtnl = NULL;
1791 static struct pernet_operations rtnetlink_net_ops = {
1792 .init = rtnetlink_net_init,
1793 .exit = rtnetlink_net_exit,
1796 void __init rtnetlink_init(void)
1798 int i;
1800 rtattr_max = 0;
1801 for (i = 0; i < ARRAY_SIZE(rta_max); i++)
1802 if (rta_max[i] > rtattr_max)
1803 rtattr_max = rta_max[i];
1804 rta_buf = kmalloc(rtattr_max * sizeof(struct rtattr *), GFP_KERNEL);
1805 if (!rta_buf)
1806 panic("rtnetlink_init: cannot allocate rta_buf\n");
1808 if (register_pernet_subsys(&rtnetlink_net_ops))
1809 panic("rtnetlink_init: cannot initialize rtnetlink\n");
1811 netlink_set_nonroot(NETLINK_ROUTE, NL_NONROOT_RECV);
1812 register_netdevice_notifier(&rtnetlink_dev_notifier);
1814 rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink, rtnl_dump_ifinfo);
1815 rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL);
1816 rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL);
1817 rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL);
1819 rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all);
1820 rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all);