Documentation: remove anticipatory scheduler info
[linux-2.6/cjktty.git] / net / core / rtnetlink.c
blob8121268ddbddfcd2da9845c8254fe0406b031be9
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);
303 * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
304 * @ops: struct rtnl_link_ops * to unregister
306 * The caller must hold the rtnl_mutex.
308 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
310 struct net *net;
312 for_each_net(net) {
313 __rtnl_kill_links(net, ops);
315 list_del(&ops->list);
317 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
320 * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
321 * @ops: struct rtnl_link_ops * to unregister
323 void rtnl_link_unregister(struct rtnl_link_ops *ops)
325 rtnl_lock();
326 __rtnl_link_unregister(ops);
327 rtnl_unlock();
329 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
331 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
333 const struct rtnl_link_ops *ops;
335 list_for_each_entry(ops, &link_ops, list) {
336 if (!strcmp(ops->kind, kind))
337 return ops;
339 return NULL;
342 static size_t rtnl_link_get_size(const struct net_device *dev)
344 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
345 size_t size;
347 if (!ops)
348 return 0;
350 size = nlmsg_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
351 nlmsg_total_size(strlen(ops->kind) + 1); /* IFLA_INFO_KIND */
353 if (ops->get_size)
354 /* IFLA_INFO_DATA + nested data */
355 size += nlmsg_total_size(sizeof(struct nlattr)) +
356 ops->get_size(dev);
358 if (ops->get_xstats_size)
359 size += ops->get_xstats_size(dev); /* IFLA_INFO_XSTATS */
361 return size;
364 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
366 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
367 struct nlattr *linkinfo, *data;
368 int err = -EMSGSIZE;
370 linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
371 if (linkinfo == NULL)
372 goto out;
374 if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
375 goto err_cancel_link;
376 if (ops->fill_xstats) {
377 err = ops->fill_xstats(skb, dev);
378 if (err < 0)
379 goto err_cancel_link;
381 if (ops->fill_info) {
382 data = nla_nest_start(skb, IFLA_INFO_DATA);
383 if (data == NULL)
384 goto err_cancel_link;
385 err = ops->fill_info(skb, dev);
386 if (err < 0)
387 goto err_cancel_data;
388 nla_nest_end(skb, data);
391 nla_nest_end(skb, linkinfo);
392 return 0;
394 err_cancel_data:
395 nla_nest_cancel(skb, data);
396 err_cancel_link:
397 nla_nest_cancel(skb, linkinfo);
398 out:
399 return err;
402 static const int rtm_min[RTM_NR_FAMILIES] =
404 [RTM_FAM(RTM_NEWLINK)] = NLMSG_LENGTH(sizeof(struct ifinfomsg)),
405 [RTM_FAM(RTM_NEWADDR)] = NLMSG_LENGTH(sizeof(struct ifaddrmsg)),
406 [RTM_FAM(RTM_NEWROUTE)] = NLMSG_LENGTH(sizeof(struct rtmsg)),
407 [RTM_FAM(RTM_NEWRULE)] = NLMSG_LENGTH(sizeof(struct fib_rule_hdr)),
408 [RTM_FAM(RTM_NEWQDISC)] = NLMSG_LENGTH(sizeof(struct tcmsg)),
409 [RTM_FAM(RTM_NEWTCLASS)] = NLMSG_LENGTH(sizeof(struct tcmsg)),
410 [RTM_FAM(RTM_NEWTFILTER)] = NLMSG_LENGTH(sizeof(struct tcmsg)),
411 [RTM_FAM(RTM_NEWACTION)] = NLMSG_LENGTH(sizeof(struct tcamsg)),
412 [RTM_FAM(RTM_GETMULTICAST)] = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
413 [RTM_FAM(RTM_GETANYCAST)] = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
416 static const int rta_max[RTM_NR_FAMILIES] =
418 [RTM_FAM(RTM_NEWLINK)] = IFLA_MAX,
419 [RTM_FAM(RTM_NEWADDR)] = IFA_MAX,
420 [RTM_FAM(RTM_NEWROUTE)] = RTA_MAX,
421 [RTM_FAM(RTM_NEWRULE)] = FRA_MAX,
422 [RTM_FAM(RTM_NEWQDISC)] = TCA_MAX,
423 [RTM_FAM(RTM_NEWTCLASS)] = TCA_MAX,
424 [RTM_FAM(RTM_NEWTFILTER)] = TCA_MAX,
425 [RTM_FAM(RTM_NEWACTION)] = TCAA_MAX,
428 void __rta_fill(struct sk_buff *skb, int attrtype, int attrlen, const void *data)
430 struct rtattr *rta;
431 int size = RTA_LENGTH(attrlen);
433 rta = (struct rtattr *)skb_put(skb, RTA_ALIGN(size));
434 rta->rta_type = attrtype;
435 rta->rta_len = size;
436 memcpy(RTA_DATA(rta), data, attrlen);
437 memset(RTA_DATA(rta) + attrlen, 0, RTA_ALIGN(size) - size);
439 EXPORT_SYMBOL(__rta_fill);
441 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned group, int echo)
443 struct sock *rtnl = net->rtnl;
444 int err = 0;
446 NETLINK_CB(skb).dst_group = group;
447 if (echo)
448 atomic_inc(&skb->users);
449 netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
450 if (echo)
451 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
452 return err;
455 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
457 struct sock *rtnl = net->rtnl;
459 return nlmsg_unicast(rtnl, skb, pid);
461 EXPORT_SYMBOL(rtnl_unicast);
463 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
464 struct nlmsghdr *nlh, gfp_t flags)
466 struct sock *rtnl = net->rtnl;
467 int report = 0;
469 if (nlh)
470 report = nlmsg_report(nlh);
472 nlmsg_notify(rtnl, skb, pid, group, report, flags);
474 EXPORT_SYMBOL(rtnl_notify);
476 void rtnl_set_sk_err(struct net *net, u32 group, int error)
478 struct sock *rtnl = net->rtnl;
480 netlink_set_err(rtnl, 0, group, error);
482 EXPORT_SYMBOL(rtnl_set_sk_err);
484 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
486 struct nlattr *mx;
487 int i, valid = 0;
489 mx = nla_nest_start(skb, RTA_METRICS);
490 if (mx == NULL)
491 return -ENOBUFS;
493 for (i = 0; i < RTAX_MAX; i++) {
494 if (metrics[i]) {
495 valid++;
496 NLA_PUT_U32(skb, i+1, metrics[i]);
500 if (!valid) {
501 nla_nest_cancel(skb, mx);
502 return 0;
505 return nla_nest_end(skb, mx);
507 nla_put_failure:
508 nla_nest_cancel(skb, mx);
509 return -EMSGSIZE;
511 EXPORT_SYMBOL(rtnetlink_put_metrics);
513 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
514 u32 ts, u32 tsage, long expires, u32 error)
516 struct rta_cacheinfo ci = {
517 .rta_lastuse = jiffies_to_clock_t(jiffies - dst->lastuse),
518 .rta_used = dst->__use,
519 .rta_clntref = atomic_read(&(dst->__refcnt)),
520 .rta_error = error,
521 .rta_id = id,
522 .rta_ts = ts,
523 .rta_tsage = tsage,
526 if (expires)
527 ci.rta_expires = jiffies_to_clock_t(expires);
529 return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
531 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
533 static void set_operstate(struct net_device *dev, unsigned char transition)
535 unsigned char operstate = dev->operstate;
537 switch (transition) {
538 case IF_OPER_UP:
539 if ((operstate == IF_OPER_DORMANT ||
540 operstate == IF_OPER_UNKNOWN) &&
541 !netif_dormant(dev))
542 operstate = IF_OPER_UP;
543 break;
545 case IF_OPER_DORMANT:
546 if (operstate == IF_OPER_UP ||
547 operstate == IF_OPER_UNKNOWN)
548 operstate = IF_OPER_DORMANT;
549 break;
552 if (dev->operstate != operstate) {
553 write_lock_bh(&dev_base_lock);
554 dev->operstate = operstate;
555 write_unlock_bh(&dev_base_lock);
556 netdev_state_change(dev);
560 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
561 const struct ifinfomsg *ifm)
563 unsigned int flags = ifm->ifi_flags;
565 /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
566 if (ifm->ifi_change)
567 flags = (flags & ifm->ifi_change) |
568 (dev->flags & ~ifm->ifi_change);
570 return flags;
573 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
574 const struct rtnl_link_stats64 *b)
576 a->rx_packets = b->rx_packets;
577 a->tx_packets = b->tx_packets;
578 a->rx_bytes = b->rx_bytes;
579 a->tx_bytes = b->tx_bytes;
580 a->rx_errors = b->rx_errors;
581 a->tx_errors = b->tx_errors;
582 a->rx_dropped = b->rx_dropped;
583 a->tx_dropped = b->tx_dropped;
585 a->multicast = b->multicast;
586 a->collisions = b->collisions;
588 a->rx_length_errors = b->rx_length_errors;
589 a->rx_over_errors = b->rx_over_errors;
590 a->rx_crc_errors = b->rx_crc_errors;
591 a->rx_frame_errors = b->rx_frame_errors;
592 a->rx_fifo_errors = b->rx_fifo_errors;
593 a->rx_missed_errors = b->rx_missed_errors;
595 a->tx_aborted_errors = b->tx_aborted_errors;
596 a->tx_carrier_errors = b->tx_carrier_errors;
597 a->tx_fifo_errors = b->tx_fifo_errors;
598 a->tx_heartbeat_errors = b->tx_heartbeat_errors;
599 a->tx_window_errors = b->tx_window_errors;
601 a->rx_compressed = b->rx_compressed;
602 a->tx_compressed = b->tx_compressed;
605 static void copy_rtnl_link_stats64(void *v, const struct rtnl_link_stats64 *b)
607 memcpy(v, b, sizeof(*b));
610 /* All VF info */
611 static inline int rtnl_vfinfo_size(const struct net_device *dev)
613 if (dev->dev.parent && dev_is_pci(dev->dev.parent)) {
615 int num_vfs = dev_num_vf(dev->dev.parent);
616 size_t size = nla_total_size(sizeof(struct nlattr));
617 size += nla_total_size(num_vfs * sizeof(struct nlattr));
618 size += num_vfs *
619 (nla_total_size(sizeof(struct ifla_vf_mac)) +
620 nla_total_size(sizeof(struct ifla_vf_vlan)) +
621 nla_total_size(sizeof(struct ifla_vf_tx_rate)));
622 return size;
623 } else
624 return 0;
627 static size_t rtnl_port_size(const struct net_device *dev)
629 size_t port_size = nla_total_size(4) /* PORT_VF */
630 + nla_total_size(PORT_PROFILE_MAX) /* PORT_PROFILE */
631 + nla_total_size(sizeof(struct ifla_port_vsi))
632 /* PORT_VSI_TYPE */
633 + nla_total_size(PORT_UUID_MAX) /* PORT_INSTANCE_UUID */
634 + nla_total_size(PORT_UUID_MAX) /* PORT_HOST_UUID */
635 + nla_total_size(1) /* PROT_VDP_REQUEST */
636 + nla_total_size(2); /* PORT_VDP_RESPONSE */
637 size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
638 size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
639 + port_size;
640 size_t port_self_size = nla_total_size(sizeof(struct nlattr))
641 + port_size;
643 if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent)
644 return 0;
645 if (dev_num_vf(dev->dev.parent))
646 return port_self_size + vf_ports_size +
647 vf_port_size * dev_num_vf(dev->dev.parent);
648 else
649 return port_self_size;
652 static noinline size_t if_nlmsg_size(const struct net_device *dev)
654 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
655 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
656 + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
657 + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
658 + nla_total_size(sizeof(struct rtnl_link_ifmap))
659 + nla_total_size(sizeof(struct rtnl_link_stats))
660 + nla_total_size(sizeof(struct rtnl_link_stats64))
661 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
662 + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
663 + nla_total_size(4) /* IFLA_TXQLEN */
664 + nla_total_size(4) /* IFLA_WEIGHT */
665 + nla_total_size(4) /* IFLA_MTU */
666 + nla_total_size(4) /* IFLA_LINK */
667 + nla_total_size(4) /* IFLA_MASTER */
668 + nla_total_size(1) /* IFLA_OPERSTATE */
669 + nla_total_size(1) /* IFLA_LINKMODE */
670 + nla_total_size(4) /* IFLA_NUM_VF */
671 + rtnl_vfinfo_size(dev) /* IFLA_VFINFO_LIST */
672 + rtnl_port_size(dev) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
673 + rtnl_link_get_size(dev); /* IFLA_LINKINFO */
676 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
678 struct nlattr *vf_ports;
679 struct nlattr *vf_port;
680 int vf;
681 int err;
683 vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
684 if (!vf_ports)
685 return -EMSGSIZE;
687 for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
688 vf_port = nla_nest_start(skb, IFLA_VF_PORT);
689 if (!vf_port)
690 goto nla_put_failure;
691 NLA_PUT_U32(skb, IFLA_PORT_VF, vf);
692 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
693 if (err == -EMSGSIZE)
694 goto nla_put_failure;
695 if (err) {
696 nla_nest_cancel(skb, vf_port);
697 continue;
699 nla_nest_end(skb, vf_port);
702 nla_nest_end(skb, vf_ports);
704 return 0;
706 nla_put_failure:
707 nla_nest_cancel(skb, vf_ports);
708 return -EMSGSIZE;
711 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
713 struct nlattr *port_self;
714 int err;
716 port_self = nla_nest_start(skb, IFLA_PORT_SELF);
717 if (!port_self)
718 return -EMSGSIZE;
720 err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
721 if (err) {
722 nla_nest_cancel(skb, port_self);
723 return (err == -EMSGSIZE) ? err : 0;
726 nla_nest_end(skb, port_self);
728 return 0;
731 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev)
733 int err;
735 if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent)
736 return 0;
738 err = rtnl_port_self_fill(skb, dev);
739 if (err)
740 return err;
742 if (dev_num_vf(dev->dev.parent)) {
743 err = rtnl_vf_ports_fill(skb, dev);
744 if (err)
745 return err;
748 return 0;
751 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
752 int type, u32 pid, u32 seq, u32 change,
753 unsigned int flags)
755 struct ifinfomsg *ifm;
756 struct nlmsghdr *nlh;
757 struct rtnl_link_stats64 temp;
758 const struct rtnl_link_stats64 *stats;
759 struct nlattr *attr;
761 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
762 if (nlh == NULL)
763 return -EMSGSIZE;
765 ifm = nlmsg_data(nlh);
766 ifm->ifi_family = AF_UNSPEC;
767 ifm->__ifi_pad = 0;
768 ifm->ifi_type = dev->type;
769 ifm->ifi_index = dev->ifindex;
770 ifm->ifi_flags = dev_get_flags(dev);
771 ifm->ifi_change = change;
773 NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
774 NLA_PUT_U32(skb, IFLA_TXQLEN, dev->tx_queue_len);
775 NLA_PUT_U8(skb, IFLA_OPERSTATE,
776 netif_running(dev) ? dev->operstate : IF_OPER_DOWN);
777 NLA_PUT_U8(skb, IFLA_LINKMODE, dev->link_mode);
778 NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
780 if (dev->ifindex != dev->iflink)
781 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
783 if (dev->master)
784 NLA_PUT_U32(skb, IFLA_MASTER, dev->master->ifindex);
786 if (dev->qdisc)
787 NLA_PUT_STRING(skb, IFLA_QDISC, dev->qdisc->ops->id);
789 if (dev->ifalias)
790 NLA_PUT_STRING(skb, IFLA_IFALIAS, dev->ifalias);
792 if (1) {
793 struct rtnl_link_ifmap map = {
794 .mem_start = dev->mem_start,
795 .mem_end = dev->mem_end,
796 .base_addr = dev->base_addr,
797 .irq = dev->irq,
798 .dma = dev->dma,
799 .port = dev->if_port,
801 NLA_PUT(skb, IFLA_MAP, sizeof(map), &map);
804 if (dev->addr_len) {
805 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
806 NLA_PUT(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast);
809 attr = nla_reserve(skb, IFLA_STATS,
810 sizeof(struct rtnl_link_stats));
811 if (attr == NULL)
812 goto nla_put_failure;
814 stats = dev_get_stats(dev, &temp);
815 copy_rtnl_link_stats(nla_data(attr), stats);
817 attr = nla_reserve(skb, IFLA_STATS64,
818 sizeof(struct rtnl_link_stats64));
819 if (attr == NULL)
820 goto nla_put_failure;
821 copy_rtnl_link_stats64(nla_data(attr), stats);
823 if (dev->dev.parent)
824 NLA_PUT_U32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent));
826 if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent) {
827 int i;
829 struct nlattr *vfinfo, *vf;
830 int num_vfs = dev_num_vf(dev->dev.parent);
832 vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
833 if (!vfinfo)
834 goto nla_put_failure;
835 for (i = 0; i < num_vfs; i++) {
836 struct ifla_vf_info ivi;
837 struct ifla_vf_mac vf_mac;
838 struct ifla_vf_vlan vf_vlan;
839 struct ifla_vf_tx_rate vf_tx_rate;
840 if (dev->netdev_ops->ndo_get_vf_config(dev, i, &ivi))
841 break;
842 vf_mac.vf = vf_vlan.vf = vf_tx_rate.vf = ivi.vf;
843 memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
844 vf_vlan.vlan = ivi.vlan;
845 vf_vlan.qos = ivi.qos;
846 vf_tx_rate.rate = ivi.tx_rate;
847 vf = nla_nest_start(skb, IFLA_VF_INFO);
848 if (!vf) {
849 nla_nest_cancel(skb, vfinfo);
850 goto nla_put_failure;
852 NLA_PUT(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac);
853 NLA_PUT(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan);
854 NLA_PUT(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate), &vf_tx_rate);
855 nla_nest_end(skb, vf);
857 nla_nest_end(skb, vfinfo);
860 if (rtnl_port_fill(skb, dev))
861 goto nla_put_failure;
863 if (dev->rtnl_link_ops) {
864 if (rtnl_link_fill(skb, dev) < 0)
865 goto nla_put_failure;
868 return nlmsg_end(skb, nlh);
870 nla_put_failure:
871 nlmsg_cancel(skb, nlh);
872 return -EMSGSIZE;
875 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
877 struct net *net = sock_net(skb->sk);
878 int h, s_h;
879 int idx = 0, s_idx;
880 struct net_device *dev;
881 struct hlist_head *head;
882 struct hlist_node *node;
884 s_h = cb->args[0];
885 s_idx = cb->args[1];
887 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
888 idx = 0;
889 head = &net->dev_index_head[h];
890 hlist_for_each_entry(dev, node, head, index_hlist) {
891 if (idx < s_idx)
892 goto cont;
893 if (rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
894 NETLINK_CB(cb->skb).pid,
895 cb->nlh->nlmsg_seq, 0,
896 NLM_F_MULTI) <= 0)
897 goto out;
898 cont:
899 idx++;
902 out:
903 cb->args[1] = idx;
904 cb->args[0] = h;
906 return skb->len;
909 const struct nla_policy ifla_policy[IFLA_MAX+1] = {
910 [IFLA_IFNAME] = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
911 [IFLA_ADDRESS] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
912 [IFLA_BROADCAST] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
913 [IFLA_MAP] = { .len = sizeof(struct rtnl_link_ifmap) },
914 [IFLA_MTU] = { .type = NLA_U32 },
915 [IFLA_LINK] = { .type = NLA_U32 },
916 [IFLA_TXQLEN] = { .type = NLA_U32 },
917 [IFLA_WEIGHT] = { .type = NLA_U32 },
918 [IFLA_OPERSTATE] = { .type = NLA_U8 },
919 [IFLA_LINKMODE] = { .type = NLA_U8 },
920 [IFLA_LINKINFO] = { .type = NLA_NESTED },
921 [IFLA_NET_NS_PID] = { .type = NLA_U32 },
922 [IFLA_IFALIAS] = { .type = NLA_STRING, .len = IFALIASZ-1 },
923 [IFLA_VFINFO_LIST] = {. type = NLA_NESTED },
924 [IFLA_VF_PORTS] = { .type = NLA_NESTED },
925 [IFLA_PORT_SELF] = { .type = NLA_NESTED },
927 EXPORT_SYMBOL(ifla_policy);
929 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
930 [IFLA_INFO_KIND] = { .type = NLA_STRING },
931 [IFLA_INFO_DATA] = { .type = NLA_NESTED },
934 static const struct nla_policy ifla_vfinfo_policy[IFLA_VF_INFO_MAX+1] = {
935 [IFLA_VF_INFO] = { .type = NLA_NESTED },
938 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
939 [IFLA_VF_MAC] = { .type = NLA_BINARY,
940 .len = sizeof(struct ifla_vf_mac) },
941 [IFLA_VF_VLAN] = { .type = NLA_BINARY,
942 .len = sizeof(struct ifla_vf_vlan) },
943 [IFLA_VF_TX_RATE] = { .type = NLA_BINARY,
944 .len = sizeof(struct ifla_vf_tx_rate) },
947 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
948 [IFLA_PORT_VF] = { .type = NLA_U32 },
949 [IFLA_PORT_PROFILE] = { .type = NLA_STRING,
950 .len = PORT_PROFILE_MAX },
951 [IFLA_PORT_VSI_TYPE] = { .type = NLA_BINARY,
952 .len = sizeof(struct ifla_port_vsi)},
953 [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
954 .len = PORT_UUID_MAX },
955 [IFLA_PORT_HOST_UUID] = { .type = NLA_STRING,
956 .len = PORT_UUID_MAX },
957 [IFLA_PORT_REQUEST] = { .type = NLA_U8, },
958 [IFLA_PORT_RESPONSE] = { .type = NLA_U16, },
961 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
963 struct net *net;
964 /* Examine the link attributes and figure out which
965 * network namespace we are talking about.
967 if (tb[IFLA_NET_NS_PID])
968 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
969 else
970 net = get_net(src_net);
971 return net;
973 EXPORT_SYMBOL(rtnl_link_get_net);
975 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
977 if (dev) {
978 if (tb[IFLA_ADDRESS] &&
979 nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
980 return -EINVAL;
982 if (tb[IFLA_BROADCAST] &&
983 nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
984 return -EINVAL;
987 return 0;
990 static int do_setvfinfo(struct net_device *dev, struct nlattr *attr)
992 int rem, err = -EINVAL;
993 struct nlattr *vf;
994 const struct net_device_ops *ops = dev->netdev_ops;
996 nla_for_each_nested(vf, attr, rem) {
997 switch (nla_type(vf)) {
998 case IFLA_VF_MAC: {
999 struct ifla_vf_mac *ivm;
1000 ivm = nla_data(vf);
1001 err = -EOPNOTSUPP;
1002 if (ops->ndo_set_vf_mac)
1003 err = ops->ndo_set_vf_mac(dev, ivm->vf,
1004 ivm->mac);
1005 break;
1007 case IFLA_VF_VLAN: {
1008 struct ifla_vf_vlan *ivv;
1009 ivv = nla_data(vf);
1010 err = -EOPNOTSUPP;
1011 if (ops->ndo_set_vf_vlan)
1012 err = ops->ndo_set_vf_vlan(dev, ivv->vf,
1013 ivv->vlan,
1014 ivv->qos);
1015 break;
1017 case IFLA_VF_TX_RATE: {
1018 struct ifla_vf_tx_rate *ivt;
1019 ivt = nla_data(vf);
1020 err = -EOPNOTSUPP;
1021 if (ops->ndo_set_vf_tx_rate)
1022 err = ops->ndo_set_vf_tx_rate(dev, ivt->vf,
1023 ivt->rate);
1024 break;
1026 default:
1027 err = -EINVAL;
1028 break;
1030 if (err)
1031 break;
1033 return err;
1036 static int do_setlink(struct net_device *dev, struct ifinfomsg *ifm,
1037 struct nlattr **tb, char *ifname, int modified)
1039 const struct net_device_ops *ops = dev->netdev_ops;
1040 int send_addr_notify = 0;
1041 int err;
1043 if (tb[IFLA_NET_NS_PID]) {
1044 struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1045 if (IS_ERR(net)) {
1046 err = PTR_ERR(net);
1047 goto errout;
1049 err = dev_change_net_namespace(dev, net, ifname);
1050 put_net(net);
1051 if (err)
1052 goto errout;
1053 modified = 1;
1056 if (tb[IFLA_MAP]) {
1057 struct rtnl_link_ifmap *u_map;
1058 struct ifmap k_map;
1060 if (!ops->ndo_set_config) {
1061 err = -EOPNOTSUPP;
1062 goto errout;
1065 if (!netif_device_present(dev)) {
1066 err = -ENODEV;
1067 goto errout;
1070 u_map = nla_data(tb[IFLA_MAP]);
1071 k_map.mem_start = (unsigned long) u_map->mem_start;
1072 k_map.mem_end = (unsigned long) u_map->mem_end;
1073 k_map.base_addr = (unsigned short) u_map->base_addr;
1074 k_map.irq = (unsigned char) u_map->irq;
1075 k_map.dma = (unsigned char) u_map->dma;
1076 k_map.port = (unsigned char) u_map->port;
1078 err = ops->ndo_set_config(dev, &k_map);
1079 if (err < 0)
1080 goto errout;
1082 modified = 1;
1085 if (tb[IFLA_ADDRESS]) {
1086 struct sockaddr *sa;
1087 int len;
1089 if (!ops->ndo_set_mac_address) {
1090 err = -EOPNOTSUPP;
1091 goto errout;
1094 if (!netif_device_present(dev)) {
1095 err = -ENODEV;
1096 goto errout;
1099 len = sizeof(sa_family_t) + dev->addr_len;
1100 sa = kmalloc(len, GFP_KERNEL);
1101 if (!sa) {
1102 err = -ENOMEM;
1103 goto errout;
1105 sa->sa_family = dev->type;
1106 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
1107 dev->addr_len);
1108 err = ops->ndo_set_mac_address(dev, sa);
1109 kfree(sa);
1110 if (err)
1111 goto errout;
1112 send_addr_notify = 1;
1113 modified = 1;
1116 if (tb[IFLA_MTU]) {
1117 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
1118 if (err < 0)
1119 goto errout;
1120 modified = 1;
1124 * Interface selected by interface index but interface
1125 * name provided implies that a name change has been
1126 * requested.
1128 if (ifm->ifi_index > 0 && ifname[0]) {
1129 err = dev_change_name(dev, ifname);
1130 if (err < 0)
1131 goto errout;
1132 modified = 1;
1135 if (tb[IFLA_IFALIAS]) {
1136 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
1137 nla_len(tb[IFLA_IFALIAS]));
1138 if (err < 0)
1139 goto errout;
1140 modified = 1;
1143 if (tb[IFLA_BROADCAST]) {
1144 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
1145 send_addr_notify = 1;
1148 if (ifm->ifi_flags || ifm->ifi_change) {
1149 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1150 if (err < 0)
1151 goto errout;
1154 if (tb[IFLA_TXQLEN])
1155 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1157 if (tb[IFLA_OPERSTATE])
1158 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1160 if (tb[IFLA_LINKMODE]) {
1161 write_lock_bh(&dev_base_lock);
1162 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1163 write_unlock_bh(&dev_base_lock);
1166 if (tb[IFLA_VFINFO_LIST]) {
1167 struct nlattr *attr;
1168 int rem;
1169 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
1170 if (nla_type(attr) != IFLA_VF_INFO) {
1171 err = -EINVAL;
1172 goto errout;
1174 err = do_setvfinfo(dev, attr);
1175 if (err < 0)
1176 goto errout;
1177 modified = 1;
1180 err = 0;
1182 if (tb[IFLA_VF_PORTS]) {
1183 struct nlattr *port[IFLA_PORT_MAX+1];
1184 struct nlattr *attr;
1185 int vf;
1186 int rem;
1188 err = -EOPNOTSUPP;
1189 if (!ops->ndo_set_vf_port)
1190 goto errout;
1192 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
1193 if (nla_type(attr) != IFLA_VF_PORT)
1194 continue;
1195 err = nla_parse_nested(port, IFLA_PORT_MAX,
1196 attr, ifla_port_policy);
1197 if (err < 0)
1198 goto errout;
1199 if (!port[IFLA_PORT_VF]) {
1200 err = -EOPNOTSUPP;
1201 goto errout;
1203 vf = nla_get_u32(port[IFLA_PORT_VF]);
1204 err = ops->ndo_set_vf_port(dev, vf, port);
1205 if (err < 0)
1206 goto errout;
1207 modified = 1;
1210 err = 0;
1212 if (tb[IFLA_PORT_SELF]) {
1213 struct nlattr *port[IFLA_PORT_MAX+1];
1215 err = nla_parse_nested(port, IFLA_PORT_MAX,
1216 tb[IFLA_PORT_SELF], ifla_port_policy);
1217 if (err < 0)
1218 goto errout;
1220 err = -EOPNOTSUPP;
1221 if (ops->ndo_set_vf_port)
1222 err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
1223 if (err < 0)
1224 goto errout;
1225 modified = 1;
1227 err = 0;
1229 errout:
1230 if (err < 0 && modified && net_ratelimit())
1231 printk(KERN_WARNING "A link change request failed with "
1232 "some changes comitted already. Interface %s may "
1233 "have been left with an inconsistent configuration, "
1234 "please check.\n", dev->name);
1236 if (send_addr_notify)
1237 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
1238 return err;
1241 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1243 struct net *net = sock_net(skb->sk);
1244 struct ifinfomsg *ifm;
1245 struct net_device *dev;
1246 int err;
1247 struct nlattr *tb[IFLA_MAX+1];
1248 char ifname[IFNAMSIZ];
1250 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1251 if (err < 0)
1252 goto errout;
1254 if (tb[IFLA_IFNAME])
1255 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1256 else
1257 ifname[0] = '\0';
1259 err = -EINVAL;
1260 ifm = nlmsg_data(nlh);
1261 if (ifm->ifi_index > 0)
1262 dev = __dev_get_by_index(net, ifm->ifi_index);
1263 else if (tb[IFLA_IFNAME])
1264 dev = __dev_get_by_name(net, ifname);
1265 else
1266 goto errout;
1268 if (dev == NULL) {
1269 err = -ENODEV;
1270 goto errout;
1273 err = validate_linkmsg(dev, tb);
1274 if (err < 0)
1275 goto errout;
1277 err = do_setlink(dev, ifm, tb, ifname, 0);
1278 errout:
1279 return err;
1282 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1284 struct net *net = sock_net(skb->sk);
1285 const struct rtnl_link_ops *ops;
1286 struct net_device *dev;
1287 struct ifinfomsg *ifm;
1288 char ifname[IFNAMSIZ];
1289 struct nlattr *tb[IFLA_MAX+1];
1290 int err;
1292 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1293 if (err < 0)
1294 return err;
1296 if (tb[IFLA_IFNAME])
1297 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1299 ifm = nlmsg_data(nlh);
1300 if (ifm->ifi_index > 0)
1301 dev = __dev_get_by_index(net, ifm->ifi_index);
1302 else if (tb[IFLA_IFNAME])
1303 dev = __dev_get_by_name(net, ifname);
1304 else
1305 return -EINVAL;
1307 if (!dev)
1308 return -ENODEV;
1310 ops = dev->rtnl_link_ops;
1311 if (!ops)
1312 return -EOPNOTSUPP;
1314 ops->dellink(dev, NULL);
1315 return 0;
1318 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
1320 unsigned int old_flags;
1321 int err;
1323 old_flags = dev->flags;
1324 if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
1325 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
1326 if (err < 0)
1327 return err;
1330 dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
1331 rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U);
1333 __dev_notify_flags(dev, old_flags);
1334 return 0;
1336 EXPORT_SYMBOL(rtnl_configure_link);
1338 struct net_device *rtnl_create_link(struct net *src_net, struct net *net,
1339 char *ifname, const struct rtnl_link_ops *ops, struct nlattr *tb[])
1341 int err;
1342 struct net_device *dev;
1343 unsigned int num_queues = 1;
1344 unsigned int real_num_queues = 1;
1346 if (ops->get_tx_queues) {
1347 err = ops->get_tx_queues(src_net, tb, &num_queues,
1348 &real_num_queues);
1349 if (err)
1350 goto err;
1352 err = -ENOMEM;
1353 dev = alloc_netdev_mq(ops->priv_size, ifname, ops->setup, num_queues);
1354 if (!dev)
1355 goto err;
1357 dev_net_set(dev, net);
1358 dev->rtnl_link_ops = ops;
1359 dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
1360 dev->real_num_tx_queues = real_num_queues;
1362 if (strchr(dev->name, '%')) {
1363 err = dev_alloc_name(dev, dev->name);
1364 if (err < 0)
1365 goto err_free;
1368 if (tb[IFLA_MTU])
1369 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
1370 if (tb[IFLA_ADDRESS])
1371 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
1372 nla_len(tb[IFLA_ADDRESS]));
1373 if (tb[IFLA_BROADCAST])
1374 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
1375 nla_len(tb[IFLA_BROADCAST]));
1376 if (tb[IFLA_TXQLEN])
1377 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1378 if (tb[IFLA_OPERSTATE])
1379 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1380 if (tb[IFLA_LINKMODE])
1381 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1383 return dev;
1385 err_free:
1386 free_netdev(dev);
1387 err:
1388 return ERR_PTR(err);
1390 EXPORT_SYMBOL(rtnl_create_link);
1392 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1394 struct net *net = sock_net(skb->sk);
1395 const struct rtnl_link_ops *ops;
1396 struct net_device *dev;
1397 struct ifinfomsg *ifm;
1398 char kind[MODULE_NAME_LEN];
1399 char ifname[IFNAMSIZ];
1400 struct nlattr *tb[IFLA_MAX+1];
1401 struct nlattr *linkinfo[IFLA_INFO_MAX+1];
1402 int err;
1404 #ifdef CONFIG_MODULES
1405 replay:
1406 #endif
1407 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1408 if (err < 0)
1409 return err;
1411 if (tb[IFLA_IFNAME])
1412 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1413 else
1414 ifname[0] = '\0';
1416 ifm = nlmsg_data(nlh);
1417 if (ifm->ifi_index > 0)
1418 dev = __dev_get_by_index(net, ifm->ifi_index);
1419 else if (ifname[0])
1420 dev = __dev_get_by_name(net, ifname);
1421 else
1422 dev = NULL;
1424 err = validate_linkmsg(dev, tb);
1425 if (err < 0)
1426 return err;
1428 if (tb[IFLA_LINKINFO]) {
1429 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
1430 tb[IFLA_LINKINFO], ifla_info_policy);
1431 if (err < 0)
1432 return err;
1433 } else
1434 memset(linkinfo, 0, sizeof(linkinfo));
1436 if (linkinfo[IFLA_INFO_KIND]) {
1437 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
1438 ops = rtnl_link_ops_get(kind);
1439 } else {
1440 kind[0] = '\0';
1441 ops = NULL;
1444 if (1) {
1445 struct nlattr *attr[ops ? ops->maxtype + 1 : 0], **data = NULL;
1446 struct net *dest_net;
1448 if (ops) {
1449 if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
1450 err = nla_parse_nested(attr, ops->maxtype,
1451 linkinfo[IFLA_INFO_DATA],
1452 ops->policy);
1453 if (err < 0)
1454 return err;
1455 data = attr;
1457 if (ops->validate) {
1458 err = ops->validate(tb, data);
1459 if (err < 0)
1460 return err;
1464 if (dev) {
1465 int modified = 0;
1467 if (nlh->nlmsg_flags & NLM_F_EXCL)
1468 return -EEXIST;
1469 if (nlh->nlmsg_flags & NLM_F_REPLACE)
1470 return -EOPNOTSUPP;
1472 if (linkinfo[IFLA_INFO_DATA]) {
1473 if (!ops || ops != dev->rtnl_link_ops ||
1474 !ops->changelink)
1475 return -EOPNOTSUPP;
1477 err = ops->changelink(dev, tb, data);
1478 if (err < 0)
1479 return err;
1480 modified = 1;
1483 return do_setlink(dev, ifm, tb, ifname, modified);
1486 if (!(nlh->nlmsg_flags & NLM_F_CREATE))
1487 return -ENODEV;
1489 if (ifm->ifi_index)
1490 return -EOPNOTSUPP;
1491 if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
1492 return -EOPNOTSUPP;
1494 if (!ops) {
1495 #ifdef CONFIG_MODULES
1496 if (kind[0]) {
1497 __rtnl_unlock();
1498 request_module("rtnl-link-%s", kind);
1499 rtnl_lock();
1500 ops = rtnl_link_ops_get(kind);
1501 if (ops)
1502 goto replay;
1504 #endif
1505 return -EOPNOTSUPP;
1508 if (!ifname[0])
1509 snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
1511 dest_net = rtnl_link_get_net(net, tb);
1512 dev = rtnl_create_link(net, dest_net, ifname, ops, tb);
1514 if (IS_ERR(dev))
1515 err = PTR_ERR(dev);
1516 else if (ops->newlink)
1517 err = ops->newlink(net, dev, tb, data);
1518 else
1519 err = register_netdevice(dev);
1521 if (err < 0 && !IS_ERR(dev))
1522 free_netdev(dev);
1523 if (err < 0)
1524 goto out;
1526 err = rtnl_configure_link(dev, ifm);
1527 if (err < 0)
1528 unregister_netdevice(dev);
1529 out:
1530 put_net(dest_net);
1531 return err;
1535 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
1537 struct net *net = sock_net(skb->sk);
1538 struct ifinfomsg *ifm;
1539 char ifname[IFNAMSIZ];
1540 struct nlattr *tb[IFLA_MAX+1];
1541 struct net_device *dev = NULL;
1542 struct sk_buff *nskb;
1543 int err;
1545 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1546 if (err < 0)
1547 return err;
1549 if (tb[IFLA_IFNAME])
1550 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1552 ifm = nlmsg_data(nlh);
1553 if (ifm->ifi_index > 0)
1554 dev = __dev_get_by_index(net, ifm->ifi_index);
1555 else if (tb[IFLA_IFNAME])
1556 dev = __dev_get_by_name(net, ifname);
1557 else
1558 return -EINVAL;
1560 if (dev == NULL)
1561 return -ENODEV;
1563 nskb = nlmsg_new(if_nlmsg_size(dev), GFP_KERNEL);
1564 if (nskb == NULL)
1565 return -ENOBUFS;
1567 err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).pid,
1568 nlh->nlmsg_seq, 0, 0);
1569 if (err < 0) {
1570 /* -EMSGSIZE implies BUG in if_nlmsg_size */
1571 WARN_ON(err == -EMSGSIZE);
1572 kfree_skb(nskb);
1573 } else
1574 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).pid);
1576 return err;
1579 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
1581 int idx;
1582 int s_idx = cb->family;
1584 if (s_idx == 0)
1585 s_idx = 1;
1586 for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
1587 int type = cb->nlh->nlmsg_type-RTM_BASE;
1588 if (idx < s_idx || idx == PF_PACKET)
1589 continue;
1590 if (rtnl_msg_handlers[idx] == NULL ||
1591 rtnl_msg_handlers[idx][type].dumpit == NULL)
1592 continue;
1593 if (idx > s_idx)
1594 memset(&cb->args[0], 0, sizeof(cb->args));
1595 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
1596 break;
1598 cb->family = idx;
1600 return skb->len;
1603 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned change)
1605 struct net *net = dev_net(dev);
1606 struct sk_buff *skb;
1607 int err = -ENOBUFS;
1609 skb = nlmsg_new(if_nlmsg_size(dev), GFP_KERNEL);
1610 if (skb == NULL)
1611 goto errout;
1613 err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0);
1614 if (err < 0) {
1615 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
1616 WARN_ON(err == -EMSGSIZE);
1617 kfree_skb(skb);
1618 goto errout;
1620 rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_KERNEL);
1621 return;
1622 errout:
1623 if (err < 0)
1624 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
1627 /* Protected by RTNL sempahore. */
1628 static struct rtattr **rta_buf;
1629 static int rtattr_max;
1631 /* Process one rtnetlink message. */
1633 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
1635 struct net *net = sock_net(skb->sk);
1636 rtnl_doit_func doit;
1637 int sz_idx, kind;
1638 int min_len;
1639 int family;
1640 int type;
1641 int err;
1643 type = nlh->nlmsg_type;
1644 if (type > RTM_MAX)
1645 return -EOPNOTSUPP;
1647 type -= RTM_BASE;
1649 /* All the messages must have at least 1 byte length */
1650 if (nlh->nlmsg_len < NLMSG_LENGTH(sizeof(struct rtgenmsg)))
1651 return 0;
1653 family = ((struct rtgenmsg *)NLMSG_DATA(nlh))->rtgen_family;
1654 sz_idx = type>>2;
1655 kind = type&3;
1657 if (kind != 2 && security_netlink_recv(skb, CAP_NET_ADMIN))
1658 return -EPERM;
1660 if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
1661 struct sock *rtnl;
1662 rtnl_dumpit_func dumpit;
1664 dumpit = rtnl_get_dumpit(family, type);
1665 if (dumpit == NULL)
1666 return -EOPNOTSUPP;
1668 __rtnl_unlock();
1669 rtnl = net->rtnl;
1670 err = netlink_dump_start(rtnl, skb, nlh, dumpit, NULL);
1671 rtnl_lock();
1672 return err;
1675 memset(rta_buf, 0, (rtattr_max * sizeof(struct rtattr *)));
1677 min_len = rtm_min[sz_idx];
1678 if (nlh->nlmsg_len < min_len)
1679 return -EINVAL;
1681 if (nlh->nlmsg_len > min_len) {
1682 int attrlen = nlh->nlmsg_len - NLMSG_ALIGN(min_len);
1683 struct rtattr *attr = (void *)nlh + NLMSG_ALIGN(min_len);
1685 while (RTA_OK(attr, attrlen)) {
1686 unsigned flavor = attr->rta_type;
1687 if (flavor) {
1688 if (flavor > rta_max[sz_idx])
1689 return -EINVAL;
1690 rta_buf[flavor-1] = attr;
1692 attr = RTA_NEXT(attr, attrlen);
1696 doit = rtnl_get_doit(family, type);
1697 if (doit == NULL)
1698 return -EOPNOTSUPP;
1700 return doit(skb, nlh, (void *)&rta_buf[0]);
1703 static void rtnetlink_rcv(struct sk_buff *skb)
1705 rtnl_lock();
1706 netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
1707 rtnl_unlock();
1710 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
1712 struct net_device *dev = ptr;
1714 switch (event) {
1715 case NETDEV_UP:
1716 case NETDEV_DOWN:
1717 case NETDEV_PRE_UP:
1718 case NETDEV_POST_INIT:
1719 case NETDEV_REGISTER:
1720 case NETDEV_CHANGE:
1721 case NETDEV_PRE_TYPE_CHANGE:
1722 case NETDEV_GOING_DOWN:
1723 case NETDEV_UNREGISTER:
1724 case NETDEV_UNREGISTER_BATCH:
1725 break;
1726 default:
1727 rtmsg_ifinfo(RTM_NEWLINK, dev, 0);
1728 break;
1730 return NOTIFY_DONE;
1733 static struct notifier_block rtnetlink_dev_notifier = {
1734 .notifier_call = rtnetlink_event,
1738 static int __net_init rtnetlink_net_init(struct net *net)
1740 struct sock *sk;
1741 sk = netlink_kernel_create(net, NETLINK_ROUTE, RTNLGRP_MAX,
1742 rtnetlink_rcv, &rtnl_mutex, THIS_MODULE);
1743 if (!sk)
1744 return -ENOMEM;
1745 net->rtnl = sk;
1746 return 0;
1749 static void __net_exit rtnetlink_net_exit(struct net *net)
1751 netlink_kernel_release(net->rtnl);
1752 net->rtnl = NULL;
1755 static struct pernet_operations rtnetlink_net_ops = {
1756 .init = rtnetlink_net_init,
1757 .exit = rtnetlink_net_exit,
1760 void __init rtnetlink_init(void)
1762 int i;
1764 rtattr_max = 0;
1765 for (i = 0; i < ARRAY_SIZE(rta_max); i++)
1766 if (rta_max[i] > rtattr_max)
1767 rtattr_max = rta_max[i];
1768 rta_buf = kmalloc(rtattr_max * sizeof(struct rtattr *), GFP_KERNEL);
1769 if (!rta_buf)
1770 panic("rtnetlink_init: cannot allocate rta_buf\n");
1772 if (register_pernet_subsys(&rtnetlink_net_ops))
1773 panic("rtnetlink_init: cannot initialize rtnetlink\n");
1775 netlink_set_nonroot(NETLINK_ROUTE, NL_NONROOT_RECV);
1776 register_netdevice_notifier(&rtnetlink_dev_notifier);
1778 rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink, rtnl_dump_ifinfo);
1779 rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL);
1780 rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL);
1781 rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL);
1783 rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all);
1784 rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all);