mm owner: fix race between swapoff and exit
[linux-2.6/next.git] / net / core / rtnetlink.c
blob71edb8b36341622df50ec48f32127b1be490b19f
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/nsproxy.h>
40 #include <asm/uaccess.h>
41 #include <asm/system.h>
42 #include <asm/string.h>
44 #include <linux/inet.h>
45 #include <linux/netdevice.h>
46 #include <net/ip.h>
47 #include <net/protocol.h>
48 #include <net/arp.h>
49 #include <net/route.h>
50 #include <net/udp.h>
51 #include <net/sock.h>
52 #include <net/pkt_sched.h>
53 #include <net/fib_rules.h>
54 #include <net/rtnetlink.h>
56 struct rtnl_link
58 rtnl_doit_func doit;
59 rtnl_dumpit_func dumpit;
62 static DEFINE_MUTEX(rtnl_mutex);
64 void rtnl_lock(void)
66 mutex_lock(&rtnl_mutex);
69 void __rtnl_unlock(void)
71 mutex_unlock(&rtnl_mutex);
74 void rtnl_unlock(void)
76 mutex_unlock(&rtnl_mutex);
77 netdev_run_todo();
80 int rtnl_trylock(void)
82 return mutex_trylock(&rtnl_mutex);
85 int rtnl_is_locked(void)
87 return mutex_is_locked(&rtnl_mutex);
90 static struct rtnl_link *rtnl_msg_handlers[NPROTO];
92 static inline int rtm_msgindex(int msgtype)
94 int msgindex = msgtype - RTM_BASE;
97 * msgindex < 0 implies someone tried to register a netlink
98 * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
99 * the message type has not been added to linux/rtnetlink.h
101 BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
103 return msgindex;
106 static rtnl_doit_func rtnl_get_doit(int protocol, int msgindex)
108 struct rtnl_link *tab;
110 tab = rtnl_msg_handlers[protocol];
111 if (tab == NULL || tab[msgindex].doit == NULL)
112 tab = rtnl_msg_handlers[PF_UNSPEC];
114 return tab ? tab[msgindex].doit : NULL;
117 static rtnl_dumpit_func rtnl_get_dumpit(int protocol, int msgindex)
119 struct rtnl_link *tab;
121 tab = rtnl_msg_handlers[protocol];
122 if (tab == NULL || tab[msgindex].dumpit == NULL)
123 tab = rtnl_msg_handlers[PF_UNSPEC];
125 return tab ? tab[msgindex].dumpit : NULL;
129 * __rtnl_register - Register a rtnetlink message type
130 * @protocol: Protocol family or PF_UNSPEC
131 * @msgtype: rtnetlink message type
132 * @doit: Function pointer called for each request message
133 * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
135 * Registers the specified function pointers (at least one of them has
136 * to be non-NULL) to be called whenever a request message for the
137 * specified protocol family and message type is received.
139 * The special protocol family PF_UNSPEC may be used to define fallback
140 * function pointers for the case when no entry for the specific protocol
141 * family exists.
143 * Returns 0 on success or a negative error code.
145 int __rtnl_register(int protocol, int msgtype,
146 rtnl_doit_func doit, rtnl_dumpit_func dumpit)
148 struct rtnl_link *tab;
149 int msgindex;
151 BUG_ON(protocol < 0 || protocol >= NPROTO);
152 msgindex = rtm_msgindex(msgtype);
154 tab = rtnl_msg_handlers[protocol];
155 if (tab == NULL) {
156 tab = kcalloc(RTM_NR_MSGTYPES, sizeof(*tab), GFP_KERNEL);
157 if (tab == NULL)
158 return -ENOBUFS;
160 rtnl_msg_handlers[protocol] = tab;
163 if (doit)
164 tab[msgindex].doit = doit;
166 if (dumpit)
167 tab[msgindex].dumpit = dumpit;
169 return 0;
172 EXPORT_SYMBOL_GPL(__rtnl_register);
175 * rtnl_register - Register a rtnetlink message type
177 * Identical to __rtnl_register() but panics on failure. This is useful
178 * as failure of this function is very unlikely, it can only happen due
179 * to lack of memory when allocating the chain to store all message
180 * handlers for a protocol. Meant for use in init functions where lack
181 * of memory implies no sense in continueing.
183 void rtnl_register(int protocol, int msgtype,
184 rtnl_doit_func doit, rtnl_dumpit_func dumpit)
186 if (__rtnl_register(protocol, msgtype, doit, dumpit) < 0)
187 panic("Unable to register rtnetlink message handler, "
188 "protocol = %d, message type = %d\n",
189 protocol, msgtype);
192 EXPORT_SYMBOL_GPL(rtnl_register);
195 * rtnl_unregister - Unregister a rtnetlink message type
196 * @protocol: Protocol family or PF_UNSPEC
197 * @msgtype: rtnetlink message type
199 * Returns 0 on success or a negative error code.
201 int rtnl_unregister(int protocol, int msgtype)
203 int msgindex;
205 BUG_ON(protocol < 0 || protocol >= NPROTO);
206 msgindex = rtm_msgindex(msgtype);
208 if (rtnl_msg_handlers[protocol] == NULL)
209 return -ENOENT;
211 rtnl_msg_handlers[protocol][msgindex].doit = NULL;
212 rtnl_msg_handlers[protocol][msgindex].dumpit = NULL;
214 return 0;
217 EXPORT_SYMBOL_GPL(rtnl_unregister);
220 * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
221 * @protocol : Protocol family or PF_UNSPEC
223 * Identical to calling rtnl_unregster() for all registered message types
224 * of a certain protocol family.
226 void rtnl_unregister_all(int protocol)
228 BUG_ON(protocol < 0 || protocol >= NPROTO);
230 kfree(rtnl_msg_handlers[protocol]);
231 rtnl_msg_handlers[protocol] = NULL;
234 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
236 static LIST_HEAD(link_ops);
239 * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
240 * @ops: struct rtnl_link_ops * to register
242 * The caller must hold the rtnl_mutex. This function should be used
243 * by drivers that create devices during module initialization. It
244 * must be called before registering the devices.
246 * Returns 0 on success or a negative error code.
248 int __rtnl_link_register(struct rtnl_link_ops *ops)
250 if (!ops->dellink)
251 ops->dellink = unregister_netdevice;
253 list_add_tail(&ops->list, &link_ops);
254 return 0;
257 EXPORT_SYMBOL_GPL(__rtnl_link_register);
260 * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
261 * @ops: struct rtnl_link_ops * to register
263 * Returns 0 on success or a negative error code.
265 int rtnl_link_register(struct rtnl_link_ops *ops)
267 int err;
269 rtnl_lock();
270 err = __rtnl_link_register(ops);
271 rtnl_unlock();
272 return err;
275 EXPORT_SYMBOL_GPL(rtnl_link_register);
277 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
279 struct net_device *dev;
280 restart:
281 for_each_netdev(net, dev) {
282 if (dev->rtnl_link_ops == ops) {
283 ops->dellink(dev);
284 goto restart;
289 void rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
291 rtnl_lock();
292 __rtnl_kill_links(net, ops);
293 rtnl_unlock();
295 EXPORT_SYMBOL_GPL(rtnl_kill_links);
298 * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
299 * @ops: struct rtnl_link_ops * to unregister
301 * The caller must hold the rtnl_mutex.
303 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
305 struct net *net;
307 for_each_net(net) {
308 __rtnl_kill_links(net, ops);
310 list_del(&ops->list);
313 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
316 * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
317 * @ops: struct rtnl_link_ops * to unregister
319 void rtnl_link_unregister(struct rtnl_link_ops *ops)
321 rtnl_lock();
322 __rtnl_link_unregister(ops);
323 rtnl_unlock();
326 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
328 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
330 const struct rtnl_link_ops *ops;
332 list_for_each_entry(ops, &link_ops, list) {
333 if (!strcmp(ops->kind, kind))
334 return ops;
336 return NULL;
339 static size_t rtnl_link_get_size(const struct net_device *dev)
341 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
342 size_t size;
344 if (!ops)
345 return 0;
347 size = nlmsg_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
348 nlmsg_total_size(strlen(ops->kind) + 1); /* IFLA_INFO_KIND */
350 if (ops->get_size)
351 /* IFLA_INFO_DATA + nested data */
352 size += nlmsg_total_size(sizeof(struct nlattr)) +
353 ops->get_size(dev);
355 if (ops->get_xstats_size)
356 size += ops->get_xstats_size(dev); /* IFLA_INFO_XSTATS */
358 return size;
361 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
363 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
364 struct nlattr *linkinfo, *data;
365 int err = -EMSGSIZE;
367 linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
368 if (linkinfo == NULL)
369 goto out;
371 if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
372 goto err_cancel_link;
373 if (ops->fill_xstats) {
374 err = ops->fill_xstats(skb, dev);
375 if (err < 0)
376 goto err_cancel_link;
378 if (ops->fill_info) {
379 data = nla_nest_start(skb, IFLA_INFO_DATA);
380 if (data == NULL)
381 goto err_cancel_link;
382 err = ops->fill_info(skb, dev);
383 if (err < 0)
384 goto err_cancel_data;
385 nla_nest_end(skb, data);
388 nla_nest_end(skb, linkinfo);
389 return 0;
391 err_cancel_data:
392 nla_nest_cancel(skb, data);
393 err_cancel_link:
394 nla_nest_cancel(skb, linkinfo);
395 out:
396 return err;
399 static const int rtm_min[RTM_NR_FAMILIES] =
401 [RTM_FAM(RTM_NEWLINK)] = NLMSG_LENGTH(sizeof(struct ifinfomsg)),
402 [RTM_FAM(RTM_NEWADDR)] = NLMSG_LENGTH(sizeof(struct ifaddrmsg)),
403 [RTM_FAM(RTM_NEWROUTE)] = NLMSG_LENGTH(sizeof(struct rtmsg)),
404 [RTM_FAM(RTM_NEWRULE)] = NLMSG_LENGTH(sizeof(struct fib_rule_hdr)),
405 [RTM_FAM(RTM_NEWQDISC)] = NLMSG_LENGTH(sizeof(struct tcmsg)),
406 [RTM_FAM(RTM_NEWTCLASS)] = NLMSG_LENGTH(sizeof(struct tcmsg)),
407 [RTM_FAM(RTM_NEWTFILTER)] = NLMSG_LENGTH(sizeof(struct tcmsg)),
408 [RTM_FAM(RTM_NEWACTION)] = NLMSG_LENGTH(sizeof(struct tcamsg)),
409 [RTM_FAM(RTM_GETMULTICAST)] = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
410 [RTM_FAM(RTM_GETANYCAST)] = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
413 static const int rta_max[RTM_NR_FAMILIES] =
415 [RTM_FAM(RTM_NEWLINK)] = IFLA_MAX,
416 [RTM_FAM(RTM_NEWADDR)] = IFA_MAX,
417 [RTM_FAM(RTM_NEWROUTE)] = RTA_MAX,
418 [RTM_FAM(RTM_NEWRULE)] = FRA_MAX,
419 [RTM_FAM(RTM_NEWQDISC)] = TCA_MAX,
420 [RTM_FAM(RTM_NEWTCLASS)] = TCA_MAX,
421 [RTM_FAM(RTM_NEWTFILTER)] = TCA_MAX,
422 [RTM_FAM(RTM_NEWACTION)] = TCAA_MAX,
425 void __rta_fill(struct sk_buff *skb, int attrtype, int attrlen, const void *data)
427 struct rtattr *rta;
428 int size = RTA_LENGTH(attrlen);
430 rta = (struct rtattr*)skb_put(skb, RTA_ALIGN(size));
431 rta->rta_type = attrtype;
432 rta->rta_len = size;
433 memcpy(RTA_DATA(rta), data, attrlen);
434 memset(RTA_DATA(rta) + attrlen, 0, RTA_ALIGN(size) - size);
437 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned group, int echo)
439 struct sock *rtnl = net->rtnl;
440 int err = 0;
442 NETLINK_CB(skb).dst_group = group;
443 if (echo)
444 atomic_inc(&skb->users);
445 netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
446 if (echo)
447 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
448 return err;
451 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
453 struct sock *rtnl = net->rtnl;
455 return nlmsg_unicast(rtnl, skb, pid);
458 int rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
459 struct nlmsghdr *nlh, gfp_t flags)
461 struct sock *rtnl = net->rtnl;
462 int report = 0;
464 if (nlh)
465 report = nlmsg_report(nlh);
467 return nlmsg_notify(rtnl, skb, pid, group, report, flags);
470 void rtnl_set_sk_err(struct net *net, u32 group, int error)
472 struct sock *rtnl = net->rtnl;
474 netlink_set_err(rtnl, 0, group, error);
477 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
479 struct nlattr *mx;
480 int i, valid = 0;
482 mx = nla_nest_start(skb, RTA_METRICS);
483 if (mx == NULL)
484 return -ENOBUFS;
486 for (i = 0; i < RTAX_MAX; i++) {
487 if (metrics[i]) {
488 valid++;
489 NLA_PUT_U32(skb, i+1, metrics[i]);
493 if (!valid) {
494 nla_nest_cancel(skb, mx);
495 return 0;
498 return nla_nest_end(skb, mx);
500 nla_put_failure:
501 nla_nest_cancel(skb, mx);
502 return -EMSGSIZE;
505 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
506 u32 ts, u32 tsage, long expires, u32 error)
508 struct rta_cacheinfo ci = {
509 .rta_lastuse = jiffies_to_clock_t(jiffies - dst->lastuse),
510 .rta_used = dst->__use,
511 .rta_clntref = atomic_read(&(dst->__refcnt)),
512 .rta_error = error,
513 .rta_id = id,
514 .rta_ts = ts,
515 .rta_tsage = tsage,
518 if (expires)
519 ci.rta_expires = jiffies_to_clock_t(expires);
521 return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
524 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
526 static void set_operstate(struct net_device *dev, unsigned char transition)
528 unsigned char operstate = dev->operstate;
530 switch(transition) {
531 case IF_OPER_UP:
532 if ((operstate == IF_OPER_DORMANT ||
533 operstate == IF_OPER_UNKNOWN) &&
534 !netif_dormant(dev))
535 operstate = IF_OPER_UP;
536 break;
538 case IF_OPER_DORMANT:
539 if (operstate == IF_OPER_UP ||
540 operstate == IF_OPER_UNKNOWN)
541 operstate = IF_OPER_DORMANT;
542 break;
545 if (dev->operstate != operstate) {
546 write_lock_bh(&dev_base_lock);
547 dev->operstate = operstate;
548 write_unlock_bh(&dev_base_lock);
549 netdev_state_change(dev);
553 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
554 struct net_device_stats *b)
556 a->rx_packets = b->rx_packets;
557 a->tx_packets = b->tx_packets;
558 a->rx_bytes = b->rx_bytes;
559 a->tx_bytes = b->tx_bytes;
560 a->rx_errors = b->rx_errors;
561 a->tx_errors = b->tx_errors;
562 a->rx_dropped = b->rx_dropped;
563 a->tx_dropped = b->tx_dropped;
565 a->multicast = b->multicast;
566 a->collisions = b->collisions;
568 a->rx_length_errors = b->rx_length_errors;
569 a->rx_over_errors = b->rx_over_errors;
570 a->rx_crc_errors = b->rx_crc_errors;
571 a->rx_frame_errors = b->rx_frame_errors;
572 a->rx_fifo_errors = b->rx_fifo_errors;
573 a->rx_missed_errors = b->rx_missed_errors;
575 a->tx_aborted_errors = b->tx_aborted_errors;
576 a->tx_carrier_errors = b->tx_carrier_errors;
577 a->tx_fifo_errors = b->tx_fifo_errors;
578 a->tx_heartbeat_errors = b->tx_heartbeat_errors;
579 a->tx_window_errors = b->tx_window_errors;
581 a->rx_compressed = b->rx_compressed;
582 a->tx_compressed = b->tx_compressed;
585 static inline size_t if_nlmsg_size(const struct net_device *dev)
587 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
588 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
589 + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
590 + nla_total_size(sizeof(struct rtnl_link_ifmap))
591 + nla_total_size(sizeof(struct rtnl_link_stats))
592 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
593 + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
594 + nla_total_size(4) /* IFLA_TXQLEN */
595 + nla_total_size(4) /* IFLA_WEIGHT */
596 + nla_total_size(4) /* IFLA_MTU */
597 + nla_total_size(4) /* IFLA_LINK */
598 + nla_total_size(4) /* IFLA_MASTER */
599 + nla_total_size(1) /* IFLA_OPERSTATE */
600 + nla_total_size(1) /* IFLA_LINKMODE */
601 + rtnl_link_get_size(dev); /* IFLA_LINKINFO */
604 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
605 int type, u32 pid, u32 seq, u32 change,
606 unsigned int flags)
608 struct netdev_queue *txq;
609 struct ifinfomsg *ifm;
610 struct nlmsghdr *nlh;
611 struct net_device_stats *stats;
612 struct nlattr *attr;
614 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
615 if (nlh == NULL)
616 return -EMSGSIZE;
618 ifm = nlmsg_data(nlh);
619 ifm->ifi_family = AF_UNSPEC;
620 ifm->__ifi_pad = 0;
621 ifm->ifi_type = dev->type;
622 ifm->ifi_index = dev->ifindex;
623 ifm->ifi_flags = dev_get_flags(dev);
624 ifm->ifi_change = change;
626 NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
627 NLA_PUT_U32(skb, IFLA_TXQLEN, dev->tx_queue_len);
628 NLA_PUT_U8(skb, IFLA_OPERSTATE,
629 netif_running(dev) ? dev->operstate : IF_OPER_DOWN);
630 NLA_PUT_U8(skb, IFLA_LINKMODE, dev->link_mode);
631 NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
633 if (dev->ifindex != dev->iflink)
634 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
636 if (dev->master)
637 NLA_PUT_U32(skb, IFLA_MASTER, dev->master->ifindex);
639 txq = netdev_get_tx_queue(dev, 0);
640 if (txq->qdisc_sleeping)
641 NLA_PUT_STRING(skb, IFLA_QDISC, txq->qdisc_sleeping->ops->id);
643 if (1) {
644 struct rtnl_link_ifmap map = {
645 .mem_start = dev->mem_start,
646 .mem_end = dev->mem_end,
647 .base_addr = dev->base_addr,
648 .irq = dev->irq,
649 .dma = dev->dma,
650 .port = dev->if_port,
652 NLA_PUT(skb, IFLA_MAP, sizeof(map), &map);
655 if (dev->addr_len) {
656 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
657 NLA_PUT(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast);
660 attr = nla_reserve(skb, IFLA_STATS,
661 sizeof(struct rtnl_link_stats));
662 if (attr == NULL)
663 goto nla_put_failure;
665 stats = dev->get_stats(dev);
666 copy_rtnl_link_stats(nla_data(attr), stats);
668 if (dev->rtnl_link_ops) {
669 if (rtnl_link_fill(skb, dev) < 0)
670 goto nla_put_failure;
673 return nlmsg_end(skb, nlh);
675 nla_put_failure:
676 nlmsg_cancel(skb, nlh);
677 return -EMSGSIZE;
680 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
682 struct net *net = sock_net(skb->sk);
683 int idx;
684 int s_idx = cb->args[0];
685 struct net_device *dev;
687 idx = 0;
688 for_each_netdev(net, dev) {
689 if (idx < s_idx)
690 goto cont;
691 if (rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
692 NETLINK_CB(cb->skb).pid,
693 cb->nlh->nlmsg_seq, 0, NLM_F_MULTI) <= 0)
694 break;
695 cont:
696 idx++;
698 cb->args[0] = idx;
700 return skb->len;
703 const struct nla_policy ifla_policy[IFLA_MAX+1] = {
704 [IFLA_IFNAME] = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
705 [IFLA_ADDRESS] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
706 [IFLA_BROADCAST] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
707 [IFLA_MAP] = { .len = sizeof(struct rtnl_link_ifmap) },
708 [IFLA_MTU] = { .type = NLA_U32 },
709 [IFLA_LINK] = { .type = NLA_U32 },
710 [IFLA_TXQLEN] = { .type = NLA_U32 },
711 [IFLA_WEIGHT] = { .type = NLA_U32 },
712 [IFLA_OPERSTATE] = { .type = NLA_U8 },
713 [IFLA_LINKMODE] = { .type = NLA_U8 },
714 [IFLA_LINKINFO] = { .type = NLA_NESTED },
715 [IFLA_NET_NS_PID] = { .type = NLA_U32 },
718 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
719 [IFLA_INFO_KIND] = { .type = NLA_STRING },
720 [IFLA_INFO_DATA] = { .type = NLA_NESTED },
723 static struct net *get_net_ns_by_pid(pid_t pid)
725 struct task_struct *tsk;
726 struct net *net;
728 /* Lookup the network namespace */
729 net = ERR_PTR(-ESRCH);
730 rcu_read_lock();
731 tsk = find_task_by_vpid(pid);
732 if (tsk) {
733 struct nsproxy *nsproxy;
734 nsproxy = task_nsproxy(tsk);
735 if (nsproxy)
736 net = get_net(nsproxy->net_ns);
738 rcu_read_unlock();
739 return net;
742 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
744 if (dev) {
745 if (tb[IFLA_ADDRESS] &&
746 nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
747 return -EINVAL;
749 if (tb[IFLA_BROADCAST] &&
750 nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
751 return -EINVAL;
754 return 0;
757 static int do_setlink(struct net_device *dev, struct ifinfomsg *ifm,
758 struct nlattr **tb, char *ifname, int modified)
760 int send_addr_notify = 0;
761 int err;
763 if (tb[IFLA_NET_NS_PID]) {
764 struct net *net;
765 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
766 if (IS_ERR(net)) {
767 err = PTR_ERR(net);
768 goto errout;
770 err = dev_change_net_namespace(dev, net, ifname);
771 put_net(net);
772 if (err)
773 goto errout;
774 modified = 1;
777 if (tb[IFLA_MAP]) {
778 struct rtnl_link_ifmap *u_map;
779 struct ifmap k_map;
781 if (!dev->set_config) {
782 err = -EOPNOTSUPP;
783 goto errout;
786 if (!netif_device_present(dev)) {
787 err = -ENODEV;
788 goto errout;
791 u_map = nla_data(tb[IFLA_MAP]);
792 k_map.mem_start = (unsigned long) u_map->mem_start;
793 k_map.mem_end = (unsigned long) u_map->mem_end;
794 k_map.base_addr = (unsigned short) u_map->base_addr;
795 k_map.irq = (unsigned char) u_map->irq;
796 k_map.dma = (unsigned char) u_map->dma;
797 k_map.port = (unsigned char) u_map->port;
799 err = dev->set_config(dev, &k_map);
800 if (err < 0)
801 goto errout;
803 modified = 1;
806 if (tb[IFLA_ADDRESS]) {
807 struct sockaddr *sa;
808 int len;
810 if (!dev->set_mac_address) {
811 err = -EOPNOTSUPP;
812 goto errout;
815 if (!netif_device_present(dev)) {
816 err = -ENODEV;
817 goto errout;
820 len = sizeof(sa_family_t) + dev->addr_len;
821 sa = kmalloc(len, GFP_KERNEL);
822 if (!sa) {
823 err = -ENOMEM;
824 goto errout;
826 sa->sa_family = dev->type;
827 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
828 dev->addr_len);
829 err = dev->set_mac_address(dev, sa);
830 kfree(sa);
831 if (err)
832 goto errout;
833 send_addr_notify = 1;
834 modified = 1;
837 if (tb[IFLA_MTU]) {
838 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
839 if (err < 0)
840 goto errout;
841 modified = 1;
845 * Interface selected by interface index but interface
846 * name provided implies that a name change has been
847 * requested.
849 if (ifm->ifi_index > 0 && ifname[0]) {
850 err = dev_change_name(dev, ifname);
851 if (err < 0)
852 goto errout;
853 modified = 1;
856 if (tb[IFLA_BROADCAST]) {
857 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
858 send_addr_notify = 1;
861 if (ifm->ifi_flags || ifm->ifi_change) {
862 unsigned int flags = ifm->ifi_flags;
864 /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
865 if (ifm->ifi_change)
866 flags = (flags & ifm->ifi_change) |
867 (dev->flags & ~ifm->ifi_change);
868 dev_change_flags(dev, flags);
871 if (tb[IFLA_TXQLEN])
872 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
874 if (tb[IFLA_OPERSTATE])
875 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
877 if (tb[IFLA_LINKMODE]) {
878 write_lock_bh(&dev_base_lock);
879 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
880 write_unlock_bh(&dev_base_lock);
883 err = 0;
885 errout:
886 if (err < 0 && modified && net_ratelimit())
887 printk(KERN_WARNING "A link change request failed with "
888 "some changes comitted already. Interface %s may "
889 "have been left with an inconsistent configuration, "
890 "please check.\n", dev->name);
892 if (send_addr_notify)
893 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
894 return err;
897 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
899 struct net *net = sock_net(skb->sk);
900 struct ifinfomsg *ifm;
901 struct net_device *dev;
902 int err;
903 struct nlattr *tb[IFLA_MAX+1];
904 char ifname[IFNAMSIZ];
906 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
907 if (err < 0)
908 goto errout;
910 if (tb[IFLA_IFNAME])
911 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
912 else
913 ifname[0] = '\0';
915 err = -EINVAL;
916 ifm = nlmsg_data(nlh);
917 if (ifm->ifi_index > 0)
918 dev = dev_get_by_index(net, ifm->ifi_index);
919 else if (tb[IFLA_IFNAME])
920 dev = dev_get_by_name(net, ifname);
921 else
922 goto errout;
924 if (dev == NULL) {
925 err = -ENODEV;
926 goto errout;
929 if ((err = validate_linkmsg(dev, tb)) < 0)
930 goto errout_dev;
932 err = do_setlink(dev, ifm, tb, ifname, 0);
933 errout_dev:
934 dev_put(dev);
935 errout:
936 return err;
939 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
941 struct net *net = sock_net(skb->sk);
942 const struct rtnl_link_ops *ops;
943 struct net_device *dev;
944 struct ifinfomsg *ifm;
945 char ifname[IFNAMSIZ];
946 struct nlattr *tb[IFLA_MAX+1];
947 int err;
949 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
950 if (err < 0)
951 return err;
953 if (tb[IFLA_IFNAME])
954 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
956 ifm = nlmsg_data(nlh);
957 if (ifm->ifi_index > 0)
958 dev = __dev_get_by_index(net, ifm->ifi_index);
959 else if (tb[IFLA_IFNAME])
960 dev = __dev_get_by_name(net, ifname);
961 else
962 return -EINVAL;
964 if (!dev)
965 return -ENODEV;
967 ops = dev->rtnl_link_ops;
968 if (!ops)
969 return -EOPNOTSUPP;
971 ops->dellink(dev);
972 return 0;
975 struct net_device *rtnl_create_link(struct net *net, char *ifname,
976 const struct rtnl_link_ops *ops, struct nlattr *tb[])
978 int err;
979 struct net_device *dev;
981 err = -ENOMEM;
982 dev = alloc_netdev(ops->priv_size, ifname, ops->setup);
983 if (!dev)
984 goto err;
986 if (strchr(dev->name, '%')) {
987 err = dev_alloc_name(dev, dev->name);
988 if (err < 0)
989 goto err_free;
992 dev_net_set(dev, net);
993 dev->rtnl_link_ops = ops;
995 if (tb[IFLA_MTU])
996 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
997 if (tb[IFLA_ADDRESS])
998 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
999 nla_len(tb[IFLA_ADDRESS]));
1000 if (tb[IFLA_BROADCAST])
1001 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
1002 nla_len(tb[IFLA_BROADCAST]));
1003 if (tb[IFLA_TXQLEN])
1004 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1005 if (tb[IFLA_OPERSTATE])
1006 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1007 if (tb[IFLA_LINKMODE])
1008 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1010 return dev;
1012 err_free:
1013 free_netdev(dev);
1014 err:
1015 return ERR_PTR(err);
1018 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1020 struct net *net = sock_net(skb->sk);
1021 const struct rtnl_link_ops *ops;
1022 struct net_device *dev;
1023 struct ifinfomsg *ifm;
1024 char kind[MODULE_NAME_LEN];
1025 char ifname[IFNAMSIZ];
1026 struct nlattr *tb[IFLA_MAX+1];
1027 struct nlattr *linkinfo[IFLA_INFO_MAX+1];
1028 int err;
1030 #ifdef CONFIG_KMOD
1031 replay:
1032 #endif
1033 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1034 if (err < 0)
1035 return err;
1037 if (tb[IFLA_IFNAME])
1038 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1039 else
1040 ifname[0] = '\0';
1042 ifm = nlmsg_data(nlh);
1043 if (ifm->ifi_index > 0)
1044 dev = __dev_get_by_index(net, ifm->ifi_index);
1045 else if (ifname[0])
1046 dev = __dev_get_by_name(net, ifname);
1047 else
1048 dev = NULL;
1050 if ((err = validate_linkmsg(dev, tb)) < 0)
1051 return err;
1053 if (tb[IFLA_LINKINFO]) {
1054 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
1055 tb[IFLA_LINKINFO], ifla_info_policy);
1056 if (err < 0)
1057 return err;
1058 } else
1059 memset(linkinfo, 0, sizeof(linkinfo));
1061 if (linkinfo[IFLA_INFO_KIND]) {
1062 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
1063 ops = rtnl_link_ops_get(kind);
1064 } else {
1065 kind[0] = '\0';
1066 ops = NULL;
1069 if (1) {
1070 struct nlattr *attr[ops ? ops->maxtype + 1 : 0], **data = NULL;
1072 if (ops) {
1073 if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
1074 err = nla_parse_nested(attr, ops->maxtype,
1075 linkinfo[IFLA_INFO_DATA],
1076 ops->policy);
1077 if (err < 0)
1078 return err;
1079 data = attr;
1081 if (ops->validate) {
1082 err = ops->validate(tb, data);
1083 if (err < 0)
1084 return err;
1088 if (dev) {
1089 int modified = 0;
1091 if (nlh->nlmsg_flags & NLM_F_EXCL)
1092 return -EEXIST;
1093 if (nlh->nlmsg_flags & NLM_F_REPLACE)
1094 return -EOPNOTSUPP;
1096 if (linkinfo[IFLA_INFO_DATA]) {
1097 if (!ops || ops != dev->rtnl_link_ops ||
1098 !ops->changelink)
1099 return -EOPNOTSUPP;
1101 err = ops->changelink(dev, tb, data);
1102 if (err < 0)
1103 return err;
1104 modified = 1;
1107 return do_setlink(dev, ifm, tb, ifname, modified);
1110 if (!(nlh->nlmsg_flags & NLM_F_CREATE))
1111 return -ENODEV;
1113 if (ifm->ifi_index || ifm->ifi_flags || ifm->ifi_change)
1114 return -EOPNOTSUPP;
1115 if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
1116 return -EOPNOTSUPP;
1118 if (!ops) {
1119 #ifdef CONFIG_KMOD
1120 if (kind[0]) {
1121 __rtnl_unlock();
1122 request_module("rtnl-link-%s", kind);
1123 rtnl_lock();
1124 ops = rtnl_link_ops_get(kind);
1125 if (ops)
1126 goto replay;
1128 #endif
1129 return -EOPNOTSUPP;
1132 if (!ifname[0])
1133 snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
1135 dev = rtnl_create_link(net, ifname, ops, tb);
1137 if (IS_ERR(dev))
1138 err = PTR_ERR(dev);
1139 else if (ops->newlink)
1140 err = ops->newlink(dev, tb, data);
1141 else
1142 err = register_netdevice(dev);
1144 if (err < 0 && !IS_ERR(dev))
1145 free_netdev(dev);
1146 return err;
1150 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
1152 struct net *net = sock_net(skb->sk);
1153 struct ifinfomsg *ifm;
1154 struct nlattr *tb[IFLA_MAX+1];
1155 struct net_device *dev = NULL;
1156 struct sk_buff *nskb;
1157 int err;
1159 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1160 if (err < 0)
1161 return err;
1163 ifm = nlmsg_data(nlh);
1164 if (ifm->ifi_index > 0) {
1165 dev = dev_get_by_index(net, ifm->ifi_index);
1166 if (dev == NULL)
1167 return -ENODEV;
1168 } else
1169 return -EINVAL;
1171 nskb = nlmsg_new(if_nlmsg_size(dev), GFP_KERNEL);
1172 if (nskb == NULL) {
1173 err = -ENOBUFS;
1174 goto errout;
1177 err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).pid,
1178 nlh->nlmsg_seq, 0, 0);
1179 if (err < 0) {
1180 /* -EMSGSIZE implies BUG in if_nlmsg_size */
1181 WARN_ON(err == -EMSGSIZE);
1182 kfree_skb(nskb);
1183 goto errout;
1185 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).pid);
1186 errout:
1187 dev_put(dev);
1189 return err;
1192 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
1194 int idx;
1195 int s_idx = cb->family;
1197 if (s_idx == 0)
1198 s_idx = 1;
1199 for (idx=1; idx<NPROTO; idx++) {
1200 int type = cb->nlh->nlmsg_type-RTM_BASE;
1201 if (idx < s_idx || idx == PF_PACKET)
1202 continue;
1203 if (rtnl_msg_handlers[idx] == NULL ||
1204 rtnl_msg_handlers[idx][type].dumpit == NULL)
1205 continue;
1206 if (idx > s_idx)
1207 memset(&cb->args[0], 0, sizeof(cb->args));
1208 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
1209 break;
1211 cb->family = idx;
1213 return skb->len;
1216 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned change)
1218 struct net *net = dev_net(dev);
1219 struct sk_buff *skb;
1220 int err = -ENOBUFS;
1222 skb = nlmsg_new(if_nlmsg_size(dev), GFP_KERNEL);
1223 if (skb == NULL)
1224 goto errout;
1226 err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0);
1227 if (err < 0) {
1228 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
1229 WARN_ON(err == -EMSGSIZE);
1230 kfree_skb(skb);
1231 goto errout;
1233 err = rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_KERNEL);
1234 errout:
1235 if (err < 0)
1236 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
1239 /* Protected by RTNL sempahore. */
1240 static struct rtattr **rta_buf;
1241 static int rtattr_max;
1243 /* Process one rtnetlink message. */
1245 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
1247 struct net *net = sock_net(skb->sk);
1248 rtnl_doit_func doit;
1249 int sz_idx, kind;
1250 int min_len;
1251 int family;
1252 int type;
1253 int err;
1255 type = nlh->nlmsg_type;
1256 if (type > RTM_MAX)
1257 return -EOPNOTSUPP;
1259 type -= RTM_BASE;
1261 /* All the messages must have at least 1 byte length */
1262 if (nlh->nlmsg_len < NLMSG_LENGTH(sizeof(struct rtgenmsg)))
1263 return 0;
1265 family = ((struct rtgenmsg*)NLMSG_DATA(nlh))->rtgen_family;
1266 if (family >= NPROTO)
1267 return -EAFNOSUPPORT;
1269 sz_idx = type>>2;
1270 kind = type&3;
1272 if (kind != 2 && security_netlink_recv(skb, CAP_NET_ADMIN))
1273 return -EPERM;
1275 if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
1276 struct sock *rtnl;
1277 rtnl_dumpit_func dumpit;
1279 dumpit = rtnl_get_dumpit(family, type);
1280 if (dumpit == NULL)
1281 return -EOPNOTSUPP;
1283 __rtnl_unlock();
1284 rtnl = net->rtnl;
1285 err = netlink_dump_start(rtnl, skb, nlh, dumpit, NULL);
1286 rtnl_lock();
1287 return err;
1290 memset(rta_buf, 0, (rtattr_max * sizeof(struct rtattr *)));
1292 min_len = rtm_min[sz_idx];
1293 if (nlh->nlmsg_len < min_len)
1294 return -EINVAL;
1296 if (nlh->nlmsg_len > min_len) {
1297 int attrlen = nlh->nlmsg_len - NLMSG_ALIGN(min_len);
1298 struct rtattr *attr = (void*)nlh + NLMSG_ALIGN(min_len);
1300 while (RTA_OK(attr, attrlen)) {
1301 unsigned flavor = attr->rta_type;
1302 if (flavor) {
1303 if (flavor > rta_max[sz_idx])
1304 return -EINVAL;
1305 rta_buf[flavor-1] = attr;
1307 attr = RTA_NEXT(attr, attrlen);
1311 doit = rtnl_get_doit(family, type);
1312 if (doit == NULL)
1313 return -EOPNOTSUPP;
1315 return doit(skb, nlh, (void *)&rta_buf[0]);
1318 static void rtnetlink_rcv(struct sk_buff *skb)
1320 rtnl_lock();
1321 netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
1322 rtnl_unlock();
1325 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
1327 struct net_device *dev = ptr;
1329 switch (event) {
1330 case NETDEV_UNREGISTER:
1331 rtmsg_ifinfo(RTM_DELLINK, dev, ~0U);
1332 break;
1333 case NETDEV_REGISTER:
1334 rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U);
1335 break;
1336 case NETDEV_UP:
1337 case NETDEV_DOWN:
1338 rtmsg_ifinfo(RTM_NEWLINK, dev, IFF_UP|IFF_RUNNING);
1339 break;
1340 case NETDEV_CHANGE:
1341 case NETDEV_GOING_DOWN:
1342 break;
1343 default:
1344 rtmsg_ifinfo(RTM_NEWLINK, dev, 0);
1345 break;
1347 return NOTIFY_DONE;
1350 static struct notifier_block rtnetlink_dev_notifier = {
1351 .notifier_call = rtnetlink_event,
1355 static int rtnetlink_net_init(struct net *net)
1357 struct sock *sk;
1358 sk = netlink_kernel_create(net, NETLINK_ROUTE, RTNLGRP_MAX,
1359 rtnetlink_rcv, &rtnl_mutex, THIS_MODULE);
1360 if (!sk)
1361 return -ENOMEM;
1362 net->rtnl = sk;
1363 return 0;
1366 static void rtnetlink_net_exit(struct net *net)
1368 netlink_kernel_release(net->rtnl);
1369 net->rtnl = NULL;
1372 static struct pernet_operations rtnetlink_net_ops = {
1373 .init = rtnetlink_net_init,
1374 .exit = rtnetlink_net_exit,
1377 void __init rtnetlink_init(void)
1379 int i;
1381 rtattr_max = 0;
1382 for (i = 0; i < ARRAY_SIZE(rta_max); i++)
1383 if (rta_max[i] > rtattr_max)
1384 rtattr_max = rta_max[i];
1385 rta_buf = kmalloc(rtattr_max * sizeof(struct rtattr *), GFP_KERNEL);
1386 if (!rta_buf)
1387 panic("rtnetlink_init: cannot allocate rta_buf\n");
1389 if (register_pernet_subsys(&rtnetlink_net_ops))
1390 panic("rtnetlink_init: cannot initialize rtnetlink\n");
1392 netlink_set_nonroot(NETLINK_ROUTE, NL_NONROOT_RECV);
1393 register_netdevice_notifier(&rtnetlink_dev_notifier);
1395 rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink, rtnl_dump_ifinfo);
1396 rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL);
1397 rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL);
1398 rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL);
1400 rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all);
1401 rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all);
1404 EXPORT_SYMBOL(__rta_fill);
1405 EXPORT_SYMBOL(rtnetlink_put_metrics);
1406 EXPORT_SYMBOL(rtnl_lock);
1407 EXPORT_SYMBOL(rtnl_trylock);
1408 EXPORT_SYMBOL(rtnl_unlock);
1409 EXPORT_SYMBOL(rtnl_is_locked);
1410 EXPORT_SYMBOL(rtnl_unicast);
1411 EXPORT_SYMBOL(rtnl_notify);
1412 EXPORT_SYMBOL(rtnl_set_sk_err);
1413 EXPORT_SYMBOL(rtnl_create_link);
1414 EXPORT_SYMBOL(ifla_policy);