[NETNS]: Add netns refcnt debug to fib rules.
[linux-2.6/mini2440.git] / net / core / rtnetlink.c
blobbc39e417694a4dab95e9a463de3a1daef78558a0
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 static struct rtnl_link *rtnl_msg_handlers[NPROTO];
87 static inline int rtm_msgindex(int msgtype)
89 int msgindex = msgtype - RTM_BASE;
92 * msgindex < 0 implies someone tried to register a netlink
93 * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
94 * the message type has not been added to linux/rtnetlink.h
96 BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
98 return msgindex;
101 static rtnl_doit_func rtnl_get_doit(int protocol, int msgindex)
103 struct rtnl_link *tab;
105 tab = rtnl_msg_handlers[protocol];
106 if (tab == NULL || tab[msgindex].doit == NULL)
107 tab = rtnl_msg_handlers[PF_UNSPEC];
109 return tab ? tab[msgindex].doit : NULL;
112 static rtnl_dumpit_func rtnl_get_dumpit(int protocol, int msgindex)
114 struct rtnl_link *tab;
116 tab = rtnl_msg_handlers[protocol];
117 if (tab == NULL || tab[msgindex].dumpit == NULL)
118 tab = rtnl_msg_handlers[PF_UNSPEC];
120 return tab ? tab[msgindex].dumpit : NULL;
124 * __rtnl_register - Register a rtnetlink message type
125 * @protocol: Protocol family or PF_UNSPEC
126 * @msgtype: rtnetlink message type
127 * @doit: Function pointer called for each request message
128 * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
130 * Registers the specified function pointers (at least one of them has
131 * to be non-NULL) to be called whenever a request message for the
132 * specified protocol family and message type is received.
134 * The special protocol family PF_UNSPEC may be used to define fallback
135 * function pointers for the case when no entry for the specific protocol
136 * family exists.
138 * Returns 0 on success or a negative error code.
140 int __rtnl_register(int protocol, int msgtype,
141 rtnl_doit_func doit, rtnl_dumpit_func dumpit)
143 struct rtnl_link *tab;
144 int msgindex;
146 BUG_ON(protocol < 0 || protocol >= NPROTO);
147 msgindex = rtm_msgindex(msgtype);
149 tab = rtnl_msg_handlers[protocol];
150 if (tab == NULL) {
151 tab = kcalloc(RTM_NR_MSGTYPES, sizeof(*tab), GFP_KERNEL);
152 if (tab == NULL)
153 return -ENOBUFS;
155 rtnl_msg_handlers[protocol] = tab;
158 if (doit)
159 tab[msgindex].doit = doit;
161 if (dumpit)
162 tab[msgindex].dumpit = dumpit;
164 return 0;
167 EXPORT_SYMBOL_GPL(__rtnl_register);
170 * rtnl_register - Register a rtnetlink message type
172 * Identical to __rtnl_register() but panics on failure. This is useful
173 * as failure of this function is very unlikely, it can only happen due
174 * to lack of memory when allocating the chain to store all message
175 * handlers for a protocol. Meant for use in init functions where lack
176 * of memory implies no sense in continueing.
178 void rtnl_register(int protocol, int msgtype,
179 rtnl_doit_func doit, rtnl_dumpit_func dumpit)
181 if (__rtnl_register(protocol, msgtype, doit, dumpit) < 0)
182 panic("Unable to register rtnetlink message handler, "
183 "protocol = %d, message type = %d\n",
184 protocol, msgtype);
187 EXPORT_SYMBOL_GPL(rtnl_register);
190 * rtnl_unregister - Unregister a rtnetlink message type
191 * @protocol: Protocol family or PF_UNSPEC
192 * @msgtype: rtnetlink message type
194 * Returns 0 on success or a negative error code.
196 int rtnl_unregister(int protocol, int msgtype)
198 int msgindex;
200 BUG_ON(protocol < 0 || protocol >= NPROTO);
201 msgindex = rtm_msgindex(msgtype);
203 if (rtnl_msg_handlers[protocol] == NULL)
204 return -ENOENT;
206 rtnl_msg_handlers[protocol][msgindex].doit = NULL;
207 rtnl_msg_handlers[protocol][msgindex].dumpit = NULL;
209 return 0;
212 EXPORT_SYMBOL_GPL(rtnl_unregister);
215 * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
216 * @protocol : Protocol family or PF_UNSPEC
218 * Identical to calling rtnl_unregster() for all registered message types
219 * of a certain protocol family.
221 void rtnl_unregister_all(int protocol)
223 BUG_ON(protocol < 0 || protocol >= NPROTO);
225 kfree(rtnl_msg_handlers[protocol]);
226 rtnl_msg_handlers[protocol] = NULL;
229 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
231 static LIST_HEAD(link_ops);
234 * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
235 * @ops: struct rtnl_link_ops * to register
237 * The caller must hold the rtnl_mutex. This function should be used
238 * by drivers that create devices during module initialization. It
239 * must be called before registering the devices.
241 * Returns 0 on success or a negative error code.
243 int __rtnl_link_register(struct rtnl_link_ops *ops)
245 if (!ops->dellink)
246 ops->dellink = unregister_netdevice;
248 list_add_tail(&ops->list, &link_ops);
249 return 0;
252 EXPORT_SYMBOL_GPL(__rtnl_link_register);
255 * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
256 * @ops: struct rtnl_link_ops * to register
258 * Returns 0 on success or a negative error code.
260 int rtnl_link_register(struct rtnl_link_ops *ops)
262 int err;
264 rtnl_lock();
265 err = __rtnl_link_register(ops);
266 rtnl_unlock();
267 return err;
270 EXPORT_SYMBOL_GPL(rtnl_link_register);
272 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
274 struct net_device *dev;
275 restart:
276 for_each_netdev(net, dev) {
277 if (dev->rtnl_link_ops == ops) {
278 ops->dellink(dev);
279 goto restart;
284 void rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
286 rtnl_lock();
287 __rtnl_kill_links(net, ops);
288 rtnl_unlock();
290 EXPORT_SYMBOL_GPL(rtnl_kill_links);
293 * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
294 * @ops: struct rtnl_link_ops * to unregister
296 * The caller must hold the rtnl_mutex.
298 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
300 struct net *net;
302 for_each_net(net) {
303 __rtnl_kill_links(net, ops);
305 list_del(&ops->list);
308 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
311 * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
312 * @ops: struct rtnl_link_ops * to unregister
314 void rtnl_link_unregister(struct rtnl_link_ops *ops)
316 rtnl_lock();
317 __rtnl_link_unregister(ops);
318 rtnl_unlock();
321 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
323 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
325 const struct rtnl_link_ops *ops;
327 list_for_each_entry(ops, &link_ops, list) {
328 if (!strcmp(ops->kind, kind))
329 return ops;
331 return NULL;
334 static size_t rtnl_link_get_size(const struct net_device *dev)
336 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
337 size_t size;
339 if (!ops)
340 return 0;
342 size = nlmsg_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
343 nlmsg_total_size(strlen(ops->kind) + 1); /* IFLA_INFO_KIND */
345 if (ops->get_size)
346 /* IFLA_INFO_DATA + nested data */
347 size += nlmsg_total_size(sizeof(struct nlattr)) +
348 ops->get_size(dev);
350 if (ops->get_xstats_size)
351 size += ops->get_xstats_size(dev); /* IFLA_INFO_XSTATS */
353 return size;
356 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
358 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
359 struct nlattr *linkinfo, *data;
360 int err = -EMSGSIZE;
362 linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
363 if (linkinfo == NULL)
364 goto out;
366 if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
367 goto err_cancel_link;
368 if (ops->fill_xstats) {
369 err = ops->fill_xstats(skb, dev);
370 if (err < 0)
371 goto err_cancel_link;
373 if (ops->fill_info) {
374 data = nla_nest_start(skb, IFLA_INFO_DATA);
375 if (data == NULL)
376 goto err_cancel_link;
377 err = ops->fill_info(skb, dev);
378 if (err < 0)
379 goto err_cancel_data;
380 nla_nest_end(skb, data);
383 nla_nest_end(skb, linkinfo);
384 return 0;
386 err_cancel_data:
387 nla_nest_cancel(skb, data);
388 err_cancel_link:
389 nla_nest_cancel(skb, linkinfo);
390 out:
391 return err;
394 static const int rtm_min[RTM_NR_FAMILIES] =
396 [RTM_FAM(RTM_NEWLINK)] = NLMSG_LENGTH(sizeof(struct ifinfomsg)),
397 [RTM_FAM(RTM_NEWADDR)] = NLMSG_LENGTH(sizeof(struct ifaddrmsg)),
398 [RTM_FAM(RTM_NEWROUTE)] = NLMSG_LENGTH(sizeof(struct rtmsg)),
399 [RTM_FAM(RTM_NEWRULE)] = NLMSG_LENGTH(sizeof(struct fib_rule_hdr)),
400 [RTM_FAM(RTM_NEWQDISC)] = NLMSG_LENGTH(sizeof(struct tcmsg)),
401 [RTM_FAM(RTM_NEWTCLASS)] = NLMSG_LENGTH(sizeof(struct tcmsg)),
402 [RTM_FAM(RTM_NEWTFILTER)] = NLMSG_LENGTH(sizeof(struct tcmsg)),
403 [RTM_FAM(RTM_NEWACTION)] = NLMSG_LENGTH(sizeof(struct tcamsg)),
404 [RTM_FAM(RTM_GETMULTICAST)] = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
405 [RTM_FAM(RTM_GETANYCAST)] = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
408 static const int rta_max[RTM_NR_FAMILIES] =
410 [RTM_FAM(RTM_NEWLINK)] = IFLA_MAX,
411 [RTM_FAM(RTM_NEWADDR)] = IFA_MAX,
412 [RTM_FAM(RTM_NEWROUTE)] = RTA_MAX,
413 [RTM_FAM(RTM_NEWRULE)] = FRA_MAX,
414 [RTM_FAM(RTM_NEWQDISC)] = TCA_MAX,
415 [RTM_FAM(RTM_NEWTCLASS)] = TCA_MAX,
416 [RTM_FAM(RTM_NEWTFILTER)] = TCA_MAX,
417 [RTM_FAM(RTM_NEWACTION)] = TCAA_MAX,
420 void __rta_fill(struct sk_buff *skb, int attrtype, int attrlen, const void *data)
422 struct rtattr *rta;
423 int size = RTA_LENGTH(attrlen);
425 rta = (struct rtattr*)skb_put(skb, RTA_ALIGN(size));
426 rta->rta_type = attrtype;
427 rta->rta_len = size;
428 memcpy(RTA_DATA(rta), data, attrlen);
429 memset(RTA_DATA(rta) + attrlen, 0, RTA_ALIGN(size) - size);
432 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned group, int echo)
434 struct sock *rtnl = net->rtnl;
435 int err = 0;
437 NETLINK_CB(skb).dst_group = group;
438 if (echo)
439 atomic_inc(&skb->users);
440 netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
441 if (echo)
442 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
443 return err;
446 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
448 struct sock *rtnl = net->rtnl;
450 return nlmsg_unicast(rtnl, skb, pid);
453 int rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
454 struct nlmsghdr *nlh, gfp_t flags)
456 struct sock *rtnl = net->rtnl;
457 int report = 0;
459 if (nlh)
460 report = nlmsg_report(nlh);
462 return nlmsg_notify(rtnl, skb, pid, group, report, flags);
465 void rtnl_set_sk_err(struct net *net, u32 group, int error)
467 struct sock *rtnl = net->rtnl;
469 netlink_set_err(rtnl, 0, group, error);
472 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
474 struct nlattr *mx;
475 int i, valid = 0;
477 mx = nla_nest_start(skb, RTA_METRICS);
478 if (mx == NULL)
479 return -ENOBUFS;
481 for (i = 0; i < RTAX_MAX; i++) {
482 if (metrics[i]) {
483 valid++;
484 NLA_PUT_U32(skb, i+1, metrics[i]);
488 if (!valid) {
489 nla_nest_cancel(skb, mx);
490 return 0;
493 return nla_nest_end(skb, mx);
495 nla_put_failure:
496 return nla_nest_cancel(skb, mx);
499 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
500 u32 ts, u32 tsage, long expires, u32 error)
502 struct rta_cacheinfo ci = {
503 .rta_lastuse = jiffies_to_clock_t(jiffies - dst->lastuse),
504 .rta_used = dst->__use,
505 .rta_clntref = atomic_read(&(dst->__refcnt)),
506 .rta_error = error,
507 .rta_id = id,
508 .rta_ts = ts,
509 .rta_tsage = tsage,
512 if (expires)
513 ci.rta_expires = jiffies_to_clock_t(expires);
515 return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
518 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
520 static void set_operstate(struct net_device *dev, unsigned char transition)
522 unsigned char operstate = dev->operstate;
524 switch(transition) {
525 case IF_OPER_UP:
526 if ((operstate == IF_OPER_DORMANT ||
527 operstate == IF_OPER_UNKNOWN) &&
528 !netif_dormant(dev))
529 operstate = IF_OPER_UP;
530 break;
532 case IF_OPER_DORMANT:
533 if (operstate == IF_OPER_UP ||
534 operstate == IF_OPER_UNKNOWN)
535 operstate = IF_OPER_DORMANT;
536 break;
539 if (dev->operstate != operstate) {
540 write_lock_bh(&dev_base_lock);
541 dev->operstate = operstate;
542 write_unlock_bh(&dev_base_lock);
543 netdev_state_change(dev);
547 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
548 struct net_device_stats *b)
550 a->rx_packets = b->rx_packets;
551 a->tx_packets = b->tx_packets;
552 a->rx_bytes = b->rx_bytes;
553 a->tx_bytes = b->tx_bytes;
554 a->rx_errors = b->rx_errors;
555 a->tx_errors = b->tx_errors;
556 a->rx_dropped = b->rx_dropped;
557 a->tx_dropped = b->tx_dropped;
559 a->multicast = b->multicast;
560 a->collisions = b->collisions;
562 a->rx_length_errors = b->rx_length_errors;
563 a->rx_over_errors = b->rx_over_errors;
564 a->rx_crc_errors = b->rx_crc_errors;
565 a->rx_frame_errors = b->rx_frame_errors;
566 a->rx_fifo_errors = b->rx_fifo_errors;
567 a->rx_missed_errors = b->rx_missed_errors;
569 a->tx_aborted_errors = b->tx_aborted_errors;
570 a->tx_carrier_errors = b->tx_carrier_errors;
571 a->tx_fifo_errors = b->tx_fifo_errors;
572 a->tx_heartbeat_errors = b->tx_heartbeat_errors;
573 a->tx_window_errors = b->tx_window_errors;
575 a->rx_compressed = b->rx_compressed;
576 a->tx_compressed = b->tx_compressed;
579 static inline size_t if_nlmsg_size(const struct net_device *dev)
581 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
582 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
583 + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
584 + nla_total_size(sizeof(struct rtnl_link_ifmap))
585 + nla_total_size(sizeof(struct rtnl_link_stats))
586 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
587 + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
588 + nla_total_size(4) /* IFLA_TXQLEN */
589 + nla_total_size(4) /* IFLA_WEIGHT */
590 + nla_total_size(4) /* IFLA_MTU */
591 + nla_total_size(4) /* IFLA_LINK */
592 + nla_total_size(4) /* IFLA_MASTER */
593 + nla_total_size(1) /* IFLA_OPERSTATE */
594 + nla_total_size(1) /* IFLA_LINKMODE */
595 + rtnl_link_get_size(dev); /* IFLA_LINKINFO */
598 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
599 int type, u32 pid, u32 seq, u32 change,
600 unsigned int flags)
602 struct ifinfomsg *ifm;
603 struct nlmsghdr *nlh;
605 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
606 if (nlh == NULL)
607 return -EMSGSIZE;
609 ifm = nlmsg_data(nlh);
610 ifm->ifi_family = AF_UNSPEC;
611 ifm->__ifi_pad = 0;
612 ifm->ifi_type = dev->type;
613 ifm->ifi_index = dev->ifindex;
614 ifm->ifi_flags = dev_get_flags(dev);
615 ifm->ifi_change = change;
617 NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
618 NLA_PUT_U32(skb, IFLA_TXQLEN, dev->tx_queue_len);
619 NLA_PUT_U8(skb, IFLA_OPERSTATE,
620 netif_running(dev) ? dev->operstate : IF_OPER_DOWN);
621 NLA_PUT_U8(skb, IFLA_LINKMODE, dev->link_mode);
622 NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
624 if (dev->ifindex != dev->iflink)
625 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
627 if (dev->master)
628 NLA_PUT_U32(skb, IFLA_MASTER, dev->master->ifindex);
630 if (dev->qdisc_sleeping)
631 NLA_PUT_STRING(skb, IFLA_QDISC, dev->qdisc_sleeping->ops->id);
633 if (1) {
634 struct rtnl_link_ifmap map = {
635 .mem_start = dev->mem_start,
636 .mem_end = dev->mem_end,
637 .base_addr = dev->base_addr,
638 .irq = dev->irq,
639 .dma = dev->dma,
640 .port = dev->if_port,
642 NLA_PUT(skb, IFLA_MAP, sizeof(map), &map);
645 if (dev->addr_len) {
646 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
647 NLA_PUT(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast);
650 if (dev->get_stats) {
651 struct net_device_stats *stats = dev->get_stats(dev);
652 if (stats) {
653 struct nlattr *attr;
655 attr = nla_reserve(skb, IFLA_STATS,
656 sizeof(struct rtnl_link_stats));
657 if (attr == NULL)
658 goto nla_put_failure;
660 copy_rtnl_link_stats(nla_data(attr), stats);
664 if (dev->rtnl_link_ops) {
665 if (rtnl_link_fill(skb, dev) < 0)
666 goto nla_put_failure;
669 return nlmsg_end(skb, nlh);
671 nla_put_failure:
672 nlmsg_cancel(skb, nlh);
673 return -EMSGSIZE;
676 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
678 struct net *net = sock_net(skb->sk);
679 int idx;
680 int s_idx = cb->args[0];
681 struct net_device *dev;
683 idx = 0;
684 for_each_netdev(net, dev) {
685 if (idx < s_idx)
686 goto cont;
687 if (rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
688 NETLINK_CB(cb->skb).pid,
689 cb->nlh->nlmsg_seq, 0, NLM_F_MULTI) <= 0)
690 break;
691 cont:
692 idx++;
694 cb->args[0] = idx;
696 return skb->len;
699 const struct nla_policy ifla_policy[IFLA_MAX+1] = {
700 [IFLA_IFNAME] = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
701 [IFLA_ADDRESS] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
702 [IFLA_BROADCAST] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
703 [IFLA_MAP] = { .len = sizeof(struct rtnl_link_ifmap) },
704 [IFLA_MTU] = { .type = NLA_U32 },
705 [IFLA_LINK] = { .type = NLA_U32 },
706 [IFLA_TXQLEN] = { .type = NLA_U32 },
707 [IFLA_WEIGHT] = { .type = NLA_U32 },
708 [IFLA_OPERSTATE] = { .type = NLA_U8 },
709 [IFLA_LINKMODE] = { .type = NLA_U8 },
710 [IFLA_LINKINFO] = { .type = NLA_NESTED },
711 [IFLA_NET_NS_PID] = { .type = NLA_U32 },
714 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
715 [IFLA_INFO_KIND] = { .type = NLA_STRING },
716 [IFLA_INFO_DATA] = { .type = NLA_NESTED },
719 static struct net *get_net_ns_by_pid(pid_t pid)
721 struct task_struct *tsk;
722 struct net *net;
724 /* Lookup the network namespace */
725 net = ERR_PTR(-ESRCH);
726 rcu_read_lock();
727 tsk = find_task_by_vpid(pid);
728 if (tsk) {
729 struct nsproxy *nsproxy;
730 nsproxy = task_nsproxy(tsk);
731 if (nsproxy)
732 net = get_net(nsproxy->net_ns);
734 rcu_read_unlock();
735 return net;
738 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
740 if (dev) {
741 if (tb[IFLA_ADDRESS] &&
742 nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
743 return -EINVAL;
745 if (tb[IFLA_BROADCAST] &&
746 nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
747 return -EINVAL;
750 return 0;
753 static int do_setlink(struct net_device *dev, struct ifinfomsg *ifm,
754 struct nlattr **tb, char *ifname, int modified)
756 int send_addr_notify = 0;
757 int err;
759 if (tb[IFLA_NET_NS_PID]) {
760 struct net *net;
761 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
762 if (IS_ERR(net)) {
763 err = PTR_ERR(net);
764 goto errout;
766 err = dev_change_net_namespace(dev, net, ifname);
767 put_net(net);
768 if (err)
769 goto errout;
770 modified = 1;
773 if (tb[IFLA_MAP]) {
774 struct rtnl_link_ifmap *u_map;
775 struct ifmap k_map;
777 if (!dev->set_config) {
778 err = -EOPNOTSUPP;
779 goto errout;
782 if (!netif_device_present(dev)) {
783 err = -ENODEV;
784 goto errout;
787 u_map = nla_data(tb[IFLA_MAP]);
788 k_map.mem_start = (unsigned long) u_map->mem_start;
789 k_map.mem_end = (unsigned long) u_map->mem_end;
790 k_map.base_addr = (unsigned short) u_map->base_addr;
791 k_map.irq = (unsigned char) u_map->irq;
792 k_map.dma = (unsigned char) u_map->dma;
793 k_map.port = (unsigned char) u_map->port;
795 err = dev->set_config(dev, &k_map);
796 if (err < 0)
797 goto errout;
799 modified = 1;
802 if (tb[IFLA_ADDRESS]) {
803 struct sockaddr *sa;
804 int len;
806 if (!dev->set_mac_address) {
807 err = -EOPNOTSUPP;
808 goto errout;
811 if (!netif_device_present(dev)) {
812 err = -ENODEV;
813 goto errout;
816 len = sizeof(sa_family_t) + dev->addr_len;
817 sa = kmalloc(len, GFP_KERNEL);
818 if (!sa) {
819 err = -ENOMEM;
820 goto errout;
822 sa->sa_family = dev->type;
823 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
824 dev->addr_len);
825 err = dev->set_mac_address(dev, sa);
826 kfree(sa);
827 if (err)
828 goto errout;
829 send_addr_notify = 1;
830 modified = 1;
833 if (tb[IFLA_MTU]) {
834 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
835 if (err < 0)
836 goto errout;
837 modified = 1;
841 * Interface selected by interface index but interface
842 * name provided implies that a name change has been
843 * requested.
845 if (ifm->ifi_index > 0 && ifname[0]) {
846 err = dev_change_name(dev, ifname);
847 if (err < 0)
848 goto errout;
849 modified = 1;
852 if (tb[IFLA_BROADCAST]) {
853 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
854 send_addr_notify = 1;
857 if (ifm->ifi_flags || ifm->ifi_change) {
858 unsigned int flags = ifm->ifi_flags;
860 /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
861 if (ifm->ifi_change)
862 flags = (flags & ifm->ifi_change) |
863 (dev->flags & ~ifm->ifi_change);
864 dev_change_flags(dev, flags);
867 if (tb[IFLA_TXQLEN])
868 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
870 if (tb[IFLA_OPERSTATE])
871 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
873 if (tb[IFLA_LINKMODE]) {
874 write_lock_bh(&dev_base_lock);
875 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
876 write_unlock_bh(&dev_base_lock);
879 err = 0;
881 errout:
882 if (err < 0 && modified && net_ratelimit())
883 printk(KERN_WARNING "A link change request failed with "
884 "some changes comitted already. Interface %s may "
885 "have been left with an inconsistent configuration, "
886 "please check.\n", dev->name);
888 if (send_addr_notify)
889 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
890 return err;
893 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
895 struct net *net = sock_net(skb->sk);
896 struct ifinfomsg *ifm;
897 struct net_device *dev;
898 int err;
899 struct nlattr *tb[IFLA_MAX+1];
900 char ifname[IFNAMSIZ];
902 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
903 if (err < 0)
904 goto errout;
906 if (tb[IFLA_IFNAME])
907 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
908 else
909 ifname[0] = '\0';
911 err = -EINVAL;
912 ifm = nlmsg_data(nlh);
913 if (ifm->ifi_index > 0)
914 dev = dev_get_by_index(net, ifm->ifi_index);
915 else if (tb[IFLA_IFNAME])
916 dev = dev_get_by_name(net, ifname);
917 else
918 goto errout;
920 if (dev == NULL) {
921 err = -ENODEV;
922 goto errout;
925 if ((err = validate_linkmsg(dev, tb)) < 0)
926 goto errout_dev;
928 err = do_setlink(dev, ifm, tb, ifname, 0);
929 errout_dev:
930 dev_put(dev);
931 errout:
932 return err;
935 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
937 struct net *net = sock_net(skb->sk);
938 const struct rtnl_link_ops *ops;
939 struct net_device *dev;
940 struct ifinfomsg *ifm;
941 char ifname[IFNAMSIZ];
942 struct nlattr *tb[IFLA_MAX+1];
943 int err;
945 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
946 if (err < 0)
947 return err;
949 if (tb[IFLA_IFNAME])
950 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
952 ifm = nlmsg_data(nlh);
953 if (ifm->ifi_index > 0)
954 dev = __dev_get_by_index(net, ifm->ifi_index);
955 else if (tb[IFLA_IFNAME])
956 dev = __dev_get_by_name(net, ifname);
957 else
958 return -EINVAL;
960 if (!dev)
961 return -ENODEV;
963 ops = dev->rtnl_link_ops;
964 if (!ops)
965 return -EOPNOTSUPP;
967 ops->dellink(dev);
968 return 0;
971 struct net_device *rtnl_create_link(struct net *net, char *ifname,
972 const struct rtnl_link_ops *ops, struct nlattr *tb[])
974 int err;
975 struct net_device *dev;
977 err = -ENOMEM;
978 dev = alloc_netdev(ops->priv_size, ifname, ops->setup);
979 if (!dev)
980 goto err;
982 if (strchr(dev->name, '%')) {
983 err = dev_alloc_name(dev, dev->name);
984 if (err < 0)
985 goto err_free;
988 dev_net_set(dev, net);
989 dev->rtnl_link_ops = ops;
991 if (tb[IFLA_MTU])
992 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
993 if (tb[IFLA_ADDRESS])
994 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
995 nla_len(tb[IFLA_ADDRESS]));
996 if (tb[IFLA_BROADCAST])
997 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
998 nla_len(tb[IFLA_BROADCAST]));
999 if (tb[IFLA_TXQLEN])
1000 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1001 if (tb[IFLA_OPERSTATE])
1002 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1003 if (tb[IFLA_LINKMODE])
1004 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1006 return dev;
1008 err_free:
1009 free_netdev(dev);
1010 err:
1011 return ERR_PTR(err);
1014 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1016 struct net *net = sock_net(skb->sk);
1017 const struct rtnl_link_ops *ops;
1018 struct net_device *dev;
1019 struct ifinfomsg *ifm;
1020 char kind[MODULE_NAME_LEN];
1021 char ifname[IFNAMSIZ];
1022 struct nlattr *tb[IFLA_MAX+1];
1023 struct nlattr *linkinfo[IFLA_INFO_MAX+1];
1024 int err;
1026 #ifdef CONFIG_KMOD
1027 replay:
1028 #endif
1029 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1030 if (err < 0)
1031 return err;
1033 if (tb[IFLA_IFNAME])
1034 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1035 else
1036 ifname[0] = '\0';
1038 ifm = nlmsg_data(nlh);
1039 if (ifm->ifi_index > 0)
1040 dev = __dev_get_by_index(net, ifm->ifi_index);
1041 else if (ifname[0])
1042 dev = __dev_get_by_name(net, ifname);
1043 else
1044 dev = NULL;
1046 if ((err = validate_linkmsg(dev, tb)) < 0)
1047 return err;
1049 if (tb[IFLA_LINKINFO]) {
1050 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
1051 tb[IFLA_LINKINFO], ifla_info_policy);
1052 if (err < 0)
1053 return err;
1054 } else
1055 memset(linkinfo, 0, sizeof(linkinfo));
1057 if (linkinfo[IFLA_INFO_KIND]) {
1058 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
1059 ops = rtnl_link_ops_get(kind);
1060 } else {
1061 kind[0] = '\0';
1062 ops = NULL;
1065 if (1) {
1066 struct nlattr *attr[ops ? ops->maxtype + 1 : 0], **data = NULL;
1068 if (ops) {
1069 if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
1070 err = nla_parse_nested(attr, ops->maxtype,
1071 linkinfo[IFLA_INFO_DATA],
1072 ops->policy);
1073 if (err < 0)
1074 return err;
1075 data = attr;
1077 if (ops->validate) {
1078 err = ops->validate(tb, data);
1079 if (err < 0)
1080 return err;
1084 if (dev) {
1085 int modified = 0;
1087 if (nlh->nlmsg_flags & NLM_F_EXCL)
1088 return -EEXIST;
1089 if (nlh->nlmsg_flags & NLM_F_REPLACE)
1090 return -EOPNOTSUPP;
1092 if (linkinfo[IFLA_INFO_DATA]) {
1093 if (!ops || ops != dev->rtnl_link_ops ||
1094 !ops->changelink)
1095 return -EOPNOTSUPP;
1097 err = ops->changelink(dev, tb, data);
1098 if (err < 0)
1099 return err;
1100 modified = 1;
1103 return do_setlink(dev, ifm, tb, ifname, modified);
1106 if (!(nlh->nlmsg_flags & NLM_F_CREATE))
1107 return -ENODEV;
1109 if (ifm->ifi_index || ifm->ifi_flags || ifm->ifi_change)
1110 return -EOPNOTSUPP;
1111 if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
1112 return -EOPNOTSUPP;
1114 if (!ops) {
1115 #ifdef CONFIG_KMOD
1116 if (kind[0]) {
1117 __rtnl_unlock();
1118 request_module("rtnl-link-%s", kind);
1119 rtnl_lock();
1120 ops = rtnl_link_ops_get(kind);
1121 if (ops)
1122 goto replay;
1124 #endif
1125 return -EOPNOTSUPP;
1128 if (!ifname[0])
1129 snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
1131 dev = rtnl_create_link(net, ifname, ops, tb);
1133 if (IS_ERR(dev))
1134 err = PTR_ERR(dev);
1135 else if (ops->newlink)
1136 err = ops->newlink(dev, tb, data);
1137 else
1138 err = register_netdevice(dev);
1140 if (err < 0 && !IS_ERR(dev))
1141 free_netdev(dev);
1142 return err;
1146 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
1148 struct net *net = sock_net(skb->sk);
1149 struct ifinfomsg *ifm;
1150 struct nlattr *tb[IFLA_MAX+1];
1151 struct net_device *dev = NULL;
1152 struct sk_buff *nskb;
1153 int err;
1155 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1156 if (err < 0)
1157 return err;
1159 ifm = nlmsg_data(nlh);
1160 if (ifm->ifi_index > 0) {
1161 dev = dev_get_by_index(net, ifm->ifi_index);
1162 if (dev == NULL)
1163 return -ENODEV;
1164 } else
1165 return -EINVAL;
1167 nskb = nlmsg_new(if_nlmsg_size(dev), GFP_KERNEL);
1168 if (nskb == NULL) {
1169 err = -ENOBUFS;
1170 goto errout;
1173 err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).pid,
1174 nlh->nlmsg_seq, 0, 0);
1175 if (err < 0) {
1176 /* -EMSGSIZE implies BUG in if_nlmsg_size */
1177 WARN_ON(err == -EMSGSIZE);
1178 kfree_skb(nskb);
1179 goto errout;
1181 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).pid);
1182 errout:
1183 dev_put(dev);
1185 return err;
1188 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
1190 int idx;
1191 int s_idx = cb->family;
1193 if (s_idx == 0)
1194 s_idx = 1;
1195 for (idx=1; idx<NPROTO; idx++) {
1196 int type = cb->nlh->nlmsg_type-RTM_BASE;
1197 if (idx < s_idx || idx == PF_PACKET)
1198 continue;
1199 if (rtnl_msg_handlers[idx] == NULL ||
1200 rtnl_msg_handlers[idx][type].dumpit == NULL)
1201 continue;
1202 if (idx > s_idx)
1203 memset(&cb->args[0], 0, sizeof(cb->args));
1204 if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
1205 break;
1207 cb->family = idx;
1209 return skb->len;
1212 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned change)
1214 struct net *net = dev_net(dev);
1215 struct sk_buff *skb;
1216 int err = -ENOBUFS;
1218 skb = nlmsg_new(if_nlmsg_size(dev), GFP_KERNEL);
1219 if (skb == NULL)
1220 goto errout;
1222 err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0);
1223 if (err < 0) {
1224 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
1225 WARN_ON(err == -EMSGSIZE);
1226 kfree_skb(skb);
1227 goto errout;
1229 err = rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_KERNEL);
1230 errout:
1231 if (err < 0)
1232 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
1235 /* Protected by RTNL sempahore. */
1236 static struct rtattr **rta_buf;
1237 static int rtattr_max;
1239 /* Process one rtnetlink message. */
1241 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
1243 struct net *net = sock_net(skb->sk);
1244 rtnl_doit_func doit;
1245 int sz_idx, kind;
1246 int min_len;
1247 int family;
1248 int type;
1249 int err;
1251 type = nlh->nlmsg_type;
1252 if (type > RTM_MAX)
1253 return -EOPNOTSUPP;
1255 type -= RTM_BASE;
1257 /* All the messages must have at least 1 byte length */
1258 if (nlh->nlmsg_len < NLMSG_LENGTH(sizeof(struct rtgenmsg)))
1259 return 0;
1261 family = ((struct rtgenmsg*)NLMSG_DATA(nlh))->rtgen_family;
1262 if (family >= NPROTO)
1263 return -EAFNOSUPPORT;
1265 sz_idx = type>>2;
1266 kind = type&3;
1268 if (kind != 2 && security_netlink_recv(skb, CAP_NET_ADMIN))
1269 return -EPERM;
1271 if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
1272 struct sock *rtnl;
1273 rtnl_dumpit_func dumpit;
1275 dumpit = rtnl_get_dumpit(family, type);
1276 if (dumpit == NULL)
1277 return -EOPNOTSUPP;
1279 __rtnl_unlock();
1280 rtnl = net->rtnl;
1281 err = netlink_dump_start(rtnl, skb, nlh, dumpit, NULL);
1282 rtnl_lock();
1283 return err;
1286 memset(rta_buf, 0, (rtattr_max * sizeof(struct rtattr *)));
1288 min_len = rtm_min[sz_idx];
1289 if (nlh->nlmsg_len < min_len)
1290 return -EINVAL;
1292 if (nlh->nlmsg_len > min_len) {
1293 int attrlen = nlh->nlmsg_len - NLMSG_ALIGN(min_len);
1294 struct rtattr *attr = (void*)nlh + NLMSG_ALIGN(min_len);
1296 while (RTA_OK(attr, attrlen)) {
1297 unsigned flavor = attr->rta_type;
1298 if (flavor) {
1299 if (flavor > rta_max[sz_idx])
1300 return -EINVAL;
1301 rta_buf[flavor-1] = attr;
1303 attr = RTA_NEXT(attr, attrlen);
1307 doit = rtnl_get_doit(family, type);
1308 if (doit == NULL)
1309 return -EOPNOTSUPP;
1311 return doit(skb, nlh, (void *)&rta_buf[0]);
1314 static void rtnetlink_rcv(struct sk_buff *skb)
1316 rtnl_lock();
1317 netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
1318 rtnl_unlock();
1321 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
1323 struct net_device *dev = ptr;
1325 switch (event) {
1326 case NETDEV_UNREGISTER:
1327 rtmsg_ifinfo(RTM_DELLINK, dev, ~0U);
1328 break;
1329 case NETDEV_REGISTER:
1330 rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U);
1331 break;
1332 case NETDEV_UP:
1333 case NETDEV_DOWN:
1334 rtmsg_ifinfo(RTM_NEWLINK, dev, IFF_UP|IFF_RUNNING);
1335 break;
1336 case NETDEV_CHANGE:
1337 case NETDEV_GOING_DOWN:
1338 break;
1339 default:
1340 rtmsg_ifinfo(RTM_NEWLINK, dev, 0);
1341 break;
1343 return NOTIFY_DONE;
1346 static struct notifier_block rtnetlink_dev_notifier = {
1347 .notifier_call = rtnetlink_event,
1351 static int rtnetlink_net_init(struct net *net)
1353 struct sock *sk;
1354 sk = netlink_kernel_create(net, NETLINK_ROUTE, RTNLGRP_MAX,
1355 rtnetlink_rcv, &rtnl_mutex, THIS_MODULE);
1356 if (!sk)
1357 return -ENOMEM;
1358 net->rtnl = sk;
1359 return 0;
1362 static void rtnetlink_net_exit(struct net *net)
1364 netlink_kernel_release(net->rtnl);
1365 net->rtnl = NULL;
1368 static struct pernet_operations rtnetlink_net_ops = {
1369 .init = rtnetlink_net_init,
1370 .exit = rtnetlink_net_exit,
1373 void __init rtnetlink_init(void)
1375 int i;
1377 rtattr_max = 0;
1378 for (i = 0; i < ARRAY_SIZE(rta_max); i++)
1379 if (rta_max[i] > rtattr_max)
1380 rtattr_max = rta_max[i];
1381 rta_buf = kmalloc(rtattr_max * sizeof(struct rtattr *), GFP_KERNEL);
1382 if (!rta_buf)
1383 panic("rtnetlink_init: cannot allocate rta_buf\n");
1385 if (register_pernet_subsys(&rtnetlink_net_ops))
1386 panic("rtnetlink_init: cannot initialize rtnetlink\n");
1388 netlink_set_nonroot(NETLINK_ROUTE, NL_NONROOT_RECV);
1389 register_netdevice_notifier(&rtnetlink_dev_notifier);
1391 rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink, rtnl_dump_ifinfo);
1392 rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL);
1393 rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL);
1394 rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL);
1396 rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all);
1397 rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all);
1400 EXPORT_SYMBOL(__rta_fill);
1401 EXPORT_SYMBOL(rtnetlink_put_metrics);
1402 EXPORT_SYMBOL(rtnl_lock);
1403 EXPORT_SYMBOL(rtnl_trylock);
1404 EXPORT_SYMBOL(rtnl_unlock);
1405 EXPORT_SYMBOL(rtnl_unicast);
1406 EXPORT_SYMBOL(rtnl_notify);
1407 EXPORT_SYMBOL(rtnl_set_sk_err);
1408 EXPORT_SYMBOL(rtnl_create_link);
1409 EXPORT_SYMBOL(ifla_policy);