2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
6 * Routing netlink socket interface: protocol independent part.
8 * Authors: Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
10 * This program is free software; you can redistribute it and/or
11 * modify it under the terms of the GNU General Public License
12 * as published by the Free Software Foundation; either version
13 * 2 of the License, or (at your option) any later version.
16 * Vitaly E. Lavrov RTA_OK arithmetics was wrong.
19 #include <linux/errno.h>
20 #include <linux/module.h>
21 #include <linux/types.h>
22 #include <linux/socket.h>
23 #include <linux/kernel.h>
24 #include <linux/timer.h>
25 #include <linux/string.h>
26 #include <linux/sockios.h>
27 #include <linux/net.h>
28 #include <linux/fcntl.h>
30 #include <linux/slab.h>
31 #include <linux/interrupt.h>
32 #include <linux/capability.h>
33 #include <linux/skbuff.h>
34 #include <linux/init.h>
35 #include <linux/security.h>
36 #include <linux/mutex.h>
37 #include <linux/if_addr.h>
38 #include <linux/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>
47 #include <net/protocol.h>
49 #include <net/route.h>
52 #include <net/pkt_sched.h>
53 #include <net/fib_rules.h>
54 #include <net/rtnetlink.h>
59 rtnl_dumpit_func dumpit
;
62 static DEFINE_MUTEX(rtnl_mutex
);
63 static struct sock
*rtnl
;
67 mutex_lock(&rtnl_mutex
);
70 void __rtnl_unlock(void)
72 mutex_unlock(&rtnl_mutex
);
75 void rtnl_unlock(void)
77 mutex_unlock(&rtnl_mutex
);
81 int rtnl_trylock(void)
83 return mutex_trylock(&rtnl_mutex
);
86 int rtattr_parse(struct rtattr
*tb
[], int maxattr
, struct rtattr
*rta
, int len
)
88 memset(tb
, 0, sizeof(struct rtattr
*)*maxattr
);
90 while (RTA_OK(rta
, len
)) {
91 unsigned flavor
= rta
->rta_type
;
92 if (flavor
&& flavor
<= maxattr
)
94 rta
= RTA_NEXT(rta
, len
);
99 int __rtattr_parse_nested_compat(struct rtattr
*tb
[], int maxattr
,
100 struct rtattr
*rta
, int len
)
102 if (RTA_PAYLOAD(rta
) < len
)
104 if (RTA_PAYLOAD(rta
) >= RTA_ALIGN(len
) + sizeof(struct rtattr
)) {
105 rta
= RTA_DATA(rta
) + RTA_ALIGN(len
);
106 return rtattr_parse_nested(tb
, maxattr
, rta
);
108 memset(tb
, 0, sizeof(struct rtattr
*) * maxattr
);
112 static struct rtnl_link
*rtnl_msg_handlers
[NPROTO
];
114 static inline int rtm_msgindex(int msgtype
)
116 int msgindex
= msgtype
- RTM_BASE
;
119 * msgindex < 0 implies someone tried to register a netlink
120 * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
121 * the message type has not been added to linux/rtnetlink.h
123 BUG_ON(msgindex
< 0 || msgindex
>= RTM_NR_MSGTYPES
);
128 static rtnl_doit_func
rtnl_get_doit(int protocol
, int msgindex
)
130 struct rtnl_link
*tab
;
132 tab
= rtnl_msg_handlers
[protocol
];
133 if (tab
== NULL
|| tab
[msgindex
].doit
== NULL
)
134 tab
= rtnl_msg_handlers
[PF_UNSPEC
];
136 return tab
? tab
[msgindex
].doit
: NULL
;
139 static rtnl_dumpit_func
rtnl_get_dumpit(int protocol
, int msgindex
)
141 struct rtnl_link
*tab
;
143 tab
= rtnl_msg_handlers
[protocol
];
144 if (tab
== NULL
|| tab
[msgindex
].dumpit
== NULL
)
145 tab
= rtnl_msg_handlers
[PF_UNSPEC
];
147 return tab
? tab
[msgindex
].dumpit
: NULL
;
151 * __rtnl_register - Register a rtnetlink message type
152 * @protocol: Protocol family or PF_UNSPEC
153 * @msgtype: rtnetlink message type
154 * @doit: Function pointer called for each request message
155 * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
157 * Registers the specified function pointers (at least one of them has
158 * to be non-NULL) to be called whenever a request message for the
159 * specified protocol family and message type is received.
161 * The special protocol family PF_UNSPEC may be used to define fallback
162 * function pointers for the case when no entry for the specific protocol
165 * Returns 0 on success or a negative error code.
167 int __rtnl_register(int protocol
, int msgtype
,
168 rtnl_doit_func doit
, rtnl_dumpit_func dumpit
)
170 struct rtnl_link
*tab
;
173 BUG_ON(protocol
< 0 || protocol
>= NPROTO
);
174 msgindex
= rtm_msgindex(msgtype
);
176 tab
= rtnl_msg_handlers
[protocol
];
178 tab
= kcalloc(RTM_NR_MSGTYPES
, sizeof(*tab
), GFP_KERNEL
);
182 rtnl_msg_handlers
[protocol
] = tab
;
186 tab
[msgindex
].doit
= doit
;
189 tab
[msgindex
].dumpit
= dumpit
;
194 EXPORT_SYMBOL_GPL(__rtnl_register
);
197 * rtnl_register - Register a rtnetlink message type
199 * Identical to __rtnl_register() but panics on failure. This is useful
200 * as failure of this function is very unlikely, it can only happen due
201 * to lack of memory when allocating the chain to store all message
202 * handlers for a protocol. Meant for use in init functions where lack
203 * of memory implies no sense in continueing.
205 void rtnl_register(int protocol
, int msgtype
,
206 rtnl_doit_func doit
, rtnl_dumpit_func dumpit
)
208 if (__rtnl_register(protocol
, msgtype
, doit
, dumpit
) < 0)
209 panic("Unable to register rtnetlink message handler, "
210 "protocol = %d, message type = %d\n",
214 EXPORT_SYMBOL_GPL(rtnl_register
);
217 * rtnl_unregister - Unregister a rtnetlink message type
218 * @protocol: Protocol family or PF_UNSPEC
219 * @msgtype: rtnetlink message type
221 * Returns 0 on success or a negative error code.
223 int rtnl_unregister(int protocol
, int msgtype
)
227 BUG_ON(protocol
< 0 || protocol
>= NPROTO
);
228 msgindex
= rtm_msgindex(msgtype
);
230 if (rtnl_msg_handlers
[protocol
] == NULL
)
233 rtnl_msg_handlers
[protocol
][msgindex
].doit
= NULL
;
234 rtnl_msg_handlers
[protocol
][msgindex
].dumpit
= NULL
;
239 EXPORT_SYMBOL_GPL(rtnl_unregister
);
242 * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
243 * @protocol : Protocol family or PF_UNSPEC
245 * Identical to calling rtnl_unregster() for all registered message types
246 * of a certain protocol family.
248 void rtnl_unregister_all(int protocol
)
250 BUG_ON(protocol
< 0 || protocol
>= NPROTO
);
252 kfree(rtnl_msg_handlers
[protocol
]);
253 rtnl_msg_handlers
[protocol
] = NULL
;
256 EXPORT_SYMBOL_GPL(rtnl_unregister_all
);
258 static LIST_HEAD(link_ops
);
261 * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
262 * @ops: struct rtnl_link_ops * to register
264 * The caller must hold the rtnl_mutex. This function should be used
265 * by drivers that create devices during module initialization. It
266 * must be called before registering the devices.
268 * Returns 0 on success or a negative error code.
270 int __rtnl_link_register(struct rtnl_link_ops
*ops
)
273 ops
->dellink
= unregister_netdevice
;
275 list_add_tail(&ops
->list
, &link_ops
);
279 EXPORT_SYMBOL_GPL(__rtnl_link_register
);
282 * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
283 * @ops: struct rtnl_link_ops * to register
285 * Returns 0 on success or a negative error code.
287 int rtnl_link_register(struct rtnl_link_ops
*ops
)
292 err
= __rtnl_link_register(ops
);
297 EXPORT_SYMBOL_GPL(rtnl_link_register
);
300 * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
301 * @ops: struct rtnl_link_ops * to unregister
303 * The caller must hold the rtnl_mutex.
305 void __rtnl_link_unregister(struct rtnl_link_ops
*ops
)
307 struct net_device
*dev
, *n
;
311 for_each_netdev_safe(net
, dev
, n
) {
312 if (dev
->rtnl_link_ops
== ops
)
316 list_del(&ops
->list
);
319 EXPORT_SYMBOL_GPL(__rtnl_link_unregister
);
322 * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
323 * @ops: struct rtnl_link_ops * to unregister
325 void rtnl_link_unregister(struct rtnl_link_ops
*ops
)
328 __rtnl_link_unregister(ops
);
332 EXPORT_SYMBOL_GPL(rtnl_link_unregister
);
334 static const struct rtnl_link_ops
*rtnl_link_ops_get(const char *kind
)
336 const struct rtnl_link_ops
*ops
;
338 list_for_each_entry(ops
, &link_ops
, list
) {
339 if (!strcmp(ops
->kind
, kind
))
345 static size_t rtnl_link_get_size(const struct net_device
*dev
)
347 const struct rtnl_link_ops
*ops
= dev
->rtnl_link_ops
;
353 size
= nlmsg_total_size(sizeof(struct nlattr
)) + /* IFLA_LINKINFO */
354 nlmsg_total_size(strlen(ops
->kind
) + 1); /* IFLA_INFO_KIND */
357 /* IFLA_INFO_DATA + nested data */
358 size
+= nlmsg_total_size(sizeof(struct nlattr
)) +
361 if (ops
->get_xstats_size
)
362 size
+= ops
->get_xstats_size(dev
); /* IFLA_INFO_XSTATS */
367 static int rtnl_link_fill(struct sk_buff
*skb
, const struct net_device
*dev
)
369 const struct rtnl_link_ops
*ops
= dev
->rtnl_link_ops
;
370 struct nlattr
*linkinfo
, *data
;
373 linkinfo
= nla_nest_start(skb
, IFLA_LINKINFO
);
374 if (linkinfo
== NULL
)
377 if (nla_put_string(skb
, IFLA_INFO_KIND
, ops
->kind
) < 0)
378 goto err_cancel_link
;
379 if (ops
->fill_xstats
) {
380 err
= ops
->fill_xstats(skb
, dev
);
382 goto err_cancel_link
;
384 if (ops
->fill_info
) {
385 data
= nla_nest_start(skb
, IFLA_INFO_DATA
);
387 goto err_cancel_link
;
388 err
= ops
->fill_info(skb
, dev
);
390 goto err_cancel_data
;
391 nla_nest_end(skb
, data
);
394 nla_nest_end(skb
, linkinfo
);
398 nla_nest_cancel(skb
, data
);
400 nla_nest_cancel(skb
, linkinfo
);
405 static const int rtm_min
[RTM_NR_FAMILIES
] =
407 [RTM_FAM(RTM_NEWLINK
)] = NLMSG_LENGTH(sizeof(struct ifinfomsg
)),
408 [RTM_FAM(RTM_NEWADDR
)] = NLMSG_LENGTH(sizeof(struct ifaddrmsg
)),
409 [RTM_FAM(RTM_NEWROUTE
)] = NLMSG_LENGTH(sizeof(struct rtmsg
)),
410 [RTM_FAM(RTM_NEWRULE
)] = NLMSG_LENGTH(sizeof(struct fib_rule_hdr
)),
411 [RTM_FAM(RTM_NEWQDISC
)] = NLMSG_LENGTH(sizeof(struct tcmsg
)),
412 [RTM_FAM(RTM_NEWTCLASS
)] = NLMSG_LENGTH(sizeof(struct tcmsg
)),
413 [RTM_FAM(RTM_NEWTFILTER
)] = NLMSG_LENGTH(sizeof(struct tcmsg
)),
414 [RTM_FAM(RTM_NEWACTION
)] = NLMSG_LENGTH(sizeof(struct tcamsg
)),
415 [RTM_FAM(RTM_GETMULTICAST
)] = NLMSG_LENGTH(sizeof(struct rtgenmsg
)),
416 [RTM_FAM(RTM_GETANYCAST
)] = NLMSG_LENGTH(sizeof(struct rtgenmsg
)),
419 static const int rta_max
[RTM_NR_FAMILIES
] =
421 [RTM_FAM(RTM_NEWLINK
)] = IFLA_MAX
,
422 [RTM_FAM(RTM_NEWADDR
)] = IFA_MAX
,
423 [RTM_FAM(RTM_NEWROUTE
)] = RTA_MAX
,
424 [RTM_FAM(RTM_NEWRULE
)] = FRA_MAX
,
425 [RTM_FAM(RTM_NEWQDISC
)] = TCA_MAX
,
426 [RTM_FAM(RTM_NEWTCLASS
)] = TCA_MAX
,
427 [RTM_FAM(RTM_NEWTFILTER
)] = TCA_MAX
,
428 [RTM_FAM(RTM_NEWACTION
)] = TCAA_MAX
,
431 void __rta_fill(struct sk_buff
*skb
, int attrtype
, int attrlen
, const void *data
)
434 int size
= RTA_LENGTH(attrlen
);
436 rta
= (struct rtattr
*)skb_put(skb
, RTA_ALIGN(size
));
437 rta
->rta_type
= attrtype
;
439 memcpy(RTA_DATA(rta
), data
, attrlen
);
440 memset(RTA_DATA(rta
) + attrlen
, 0, RTA_ALIGN(size
) - size
);
443 size_t rtattr_strlcpy(char *dest
, const struct rtattr
*rta
, size_t size
)
445 size_t ret
= RTA_PAYLOAD(rta
);
446 char *src
= RTA_DATA(rta
);
448 if (ret
> 0 && src
[ret
- 1] == '\0')
451 size_t len
= (ret
>= size
) ? size
- 1 : ret
;
452 memset(dest
, 0, size
);
453 memcpy(dest
, src
, len
);
458 int rtnetlink_send(struct sk_buff
*skb
, u32 pid
, unsigned group
, int echo
)
462 NETLINK_CB(skb
).dst_group
= group
;
464 atomic_inc(&skb
->users
);
465 netlink_broadcast(rtnl
, skb
, pid
, group
, GFP_KERNEL
);
467 err
= netlink_unicast(rtnl
, skb
, pid
, MSG_DONTWAIT
);
471 int rtnl_unicast(struct sk_buff
*skb
, u32 pid
)
473 return nlmsg_unicast(rtnl
, skb
, pid
);
476 int rtnl_notify(struct sk_buff
*skb
, u32 pid
, u32 group
,
477 struct nlmsghdr
*nlh
, gfp_t flags
)
482 report
= nlmsg_report(nlh
);
484 return nlmsg_notify(rtnl
, skb
, pid
, group
, report
, flags
);
487 void rtnl_set_sk_err(u32 group
, int error
)
489 netlink_set_err(rtnl
, 0, group
, error
);
492 int rtnetlink_put_metrics(struct sk_buff
*skb
, u32
*metrics
)
497 mx
= nla_nest_start(skb
, RTA_METRICS
);
501 for (i
= 0; i
< RTAX_MAX
; i
++) {
504 NLA_PUT_U32(skb
, i
+1, metrics
[i
]);
509 nla_nest_cancel(skb
, mx
);
513 return nla_nest_end(skb
, mx
);
516 return nla_nest_cancel(skb
, mx
);
519 int rtnl_put_cacheinfo(struct sk_buff
*skb
, struct dst_entry
*dst
, u32 id
,
520 u32 ts
, u32 tsage
, long expires
, u32 error
)
522 struct rta_cacheinfo ci
= {
523 .rta_lastuse
= jiffies_to_clock_t(jiffies
- dst
->lastuse
),
524 .rta_used
= dst
->__use
,
525 .rta_clntref
= atomic_read(&(dst
->__refcnt
)),
533 ci
.rta_expires
= jiffies_to_clock_t(expires
);
535 return nla_put(skb
, RTA_CACHEINFO
, sizeof(ci
), &ci
);
538 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo
);
540 static void set_operstate(struct net_device
*dev
, unsigned char transition
)
542 unsigned char operstate
= dev
->operstate
;
546 if ((operstate
== IF_OPER_DORMANT
||
547 operstate
== IF_OPER_UNKNOWN
) &&
549 operstate
= IF_OPER_UP
;
552 case IF_OPER_DORMANT
:
553 if (operstate
== IF_OPER_UP
||
554 operstate
== IF_OPER_UNKNOWN
)
555 operstate
= IF_OPER_DORMANT
;
559 if (dev
->operstate
!= operstate
) {
560 write_lock_bh(&dev_base_lock
);
561 dev
->operstate
= operstate
;
562 write_unlock_bh(&dev_base_lock
);
563 netdev_state_change(dev
);
567 static void copy_rtnl_link_stats(struct rtnl_link_stats
*a
,
568 struct net_device_stats
*b
)
570 a
->rx_packets
= b
->rx_packets
;
571 a
->tx_packets
= b
->tx_packets
;
572 a
->rx_bytes
= b
->rx_bytes
;
573 a
->tx_bytes
= b
->tx_bytes
;
574 a
->rx_errors
= b
->rx_errors
;
575 a
->tx_errors
= b
->tx_errors
;
576 a
->rx_dropped
= b
->rx_dropped
;
577 a
->tx_dropped
= b
->tx_dropped
;
579 a
->multicast
= b
->multicast
;
580 a
->collisions
= b
->collisions
;
582 a
->rx_length_errors
= b
->rx_length_errors
;
583 a
->rx_over_errors
= b
->rx_over_errors
;
584 a
->rx_crc_errors
= b
->rx_crc_errors
;
585 a
->rx_frame_errors
= b
->rx_frame_errors
;
586 a
->rx_fifo_errors
= b
->rx_fifo_errors
;
587 a
->rx_missed_errors
= b
->rx_missed_errors
;
589 a
->tx_aborted_errors
= b
->tx_aborted_errors
;
590 a
->tx_carrier_errors
= b
->tx_carrier_errors
;
591 a
->tx_fifo_errors
= b
->tx_fifo_errors
;
592 a
->tx_heartbeat_errors
= b
->tx_heartbeat_errors
;
593 a
->tx_window_errors
= b
->tx_window_errors
;
595 a
->rx_compressed
= b
->rx_compressed
;
596 a
->tx_compressed
= b
->tx_compressed
;
599 static inline size_t if_nlmsg_size(const struct net_device
*dev
)
601 return NLMSG_ALIGN(sizeof(struct ifinfomsg
))
602 + nla_total_size(IFNAMSIZ
) /* IFLA_IFNAME */
603 + nla_total_size(IFNAMSIZ
) /* IFLA_QDISC */
604 + nla_total_size(sizeof(struct rtnl_link_ifmap
))
605 + nla_total_size(sizeof(struct rtnl_link_stats
))
606 + nla_total_size(MAX_ADDR_LEN
) /* IFLA_ADDRESS */
607 + nla_total_size(MAX_ADDR_LEN
) /* IFLA_BROADCAST */
608 + nla_total_size(4) /* IFLA_TXQLEN */
609 + nla_total_size(4) /* IFLA_WEIGHT */
610 + nla_total_size(4) /* IFLA_MTU */
611 + nla_total_size(4) /* IFLA_LINK */
612 + nla_total_size(4) /* IFLA_MASTER */
613 + nla_total_size(1) /* IFLA_OPERSTATE */
614 + nla_total_size(1) /* IFLA_LINKMODE */
615 + rtnl_link_get_size(dev
); /* IFLA_LINKINFO */
618 static int rtnl_fill_ifinfo(struct sk_buff
*skb
, struct net_device
*dev
,
619 int type
, u32 pid
, u32 seq
, u32 change
,
622 struct ifinfomsg
*ifm
;
623 struct nlmsghdr
*nlh
;
625 nlh
= nlmsg_put(skb
, pid
, seq
, type
, sizeof(*ifm
), flags
);
629 ifm
= nlmsg_data(nlh
);
630 ifm
->ifi_family
= AF_UNSPEC
;
632 ifm
->ifi_type
= dev
->type
;
633 ifm
->ifi_index
= dev
->ifindex
;
634 ifm
->ifi_flags
= dev_get_flags(dev
);
635 ifm
->ifi_change
= change
;
637 NLA_PUT_STRING(skb
, IFLA_IFNAME
, dev
->name
);
638 NLA_PUT_U32(skb
, IFLA_TXQLEN
, dev
->tx_queue_len
);
639 NLA_PUT_U8(skb
, IFLA_OPERSTATE
,
640 netif_running(dev
) ? dev
->operstate
: IF_OPER_DOWN
);
641 NLA_PUT_U8(skb
, IFLA_LINKMODE
, dev
->link_mode
);
642 NLA_PUT_U32(skb
, IFLA_MTU
, dev
->mtu
);
644 if (dev
->ifindex
!= dev
->iflink
)
645 NLA_PUT_U32(skb
, IFLA_LINK
, dev
->iflink
);
648 NLA_PUT_U32(skb
, IFLA_MASTER
, dev
->master
->ifindex
);
650 if (dev
->qdisc_sleeping
)
651 NLA_PUT_STRING(skb
, IFLA_QDISC
, dev
->qdisc_sleeping
->ops
->id
);
654 struct rtnl_link_ifmap map
= {
655 .mem_start
= dev
->mem_start
,
656 .mem_end
= dev
->mem_end
,
657 .base_addr
= dev
->base_addr
,
660 .port
= dev
->if_port
,
662 NLA_PUT(skb
, IFLA_MAP
, sizeof(map
), &map
);
666 NLA_PUT(skb
, IFLA_ADDRESS
, dev
->addr_len
, dev
->dev_addr
);
667 NLA_PUT(skb
, IFLA_BROADCAST
, dev
->addr_len
, dev
->broadcast
);
670 if (dev
->get_stats
) {
671 struct net_device_stats
*stats
= dev
->get_stats(dev
);
675 attr
= nla_reserve(skb
, IFLA_STATS
,
676 sizeof(struct rtnl_link_stats
));
678 goto nla_put_failure
;
680 copy_rtnl_link_stats(nla_data(attr
), stats
);
684 if (dev
->rtnl_link_ops
) {
685 if (rtnl_link_fill(skb
, dev
) < 0)
686 goto nla_put_failure
;
689 return nlmsg_end(skb
, nlh
);
692 nlmsg_cancel(skb
, nlh
);
696 static int rtnl_dump_ifinfo(struct sk_buff
*skb
, struct netlink_callback
*cb
)
698 struct net
*net
= skb
->sk
->sk_net
;
700 int s_idx
= cb
->args
[0];
701 struct net_device
*dev
;
704 for_each_netdev(net
, dev
) {
707 if (rtnl_fill_ifinfo(skb
, dev
, RTM_NEWLINK
,
708 NETLINK_CB(cb
->skb
).pid
,
709 cb
->nlh
->nlmsg_seq
, 0, NLM_F_MULTI
) <= 0)
719 const struct nla_policy ifla_policy
[IFLA_MAX
+1] = {
720 [IFLA_IFNAME
] = { .type
= NLA_STRING
, .len
= IFNAMSIZ
-1 },
721 [IFLA_ADDRESS
] = { .type
= NLA_BINARY
, .len
= MAX_ADDR_LEN
},
722 [IFLA_BROADCAST
] = { .type
= NLA_BINARY
, .len
= MAX_ADDR_LEN
},
723 [IFLA_MAP
] = { .len
= sizeof(struct rtnl_link_ifmap
) },
724 [IFLA_MTU
] = { .type
= NLA_U32
},
725 [IFLA_TXQLEN
] = { .type
= NLA_U32
},
726 [IFLA_WEIGHT
] = { .type
= NLA_U32
},
727 [IFLA_OPERSTATE
] = { .type
= NLA_U8
},
728 [IFLA_LINKMODE
] = { .type
= NLA_U8
},
729 [IFLA_NET_NS_PID
] = { .type
= NLA_U32
},
732 static const struct nla_policy ifla_info_policy
[IFLA_INFO_MAX
+1] = {
733 [IFLA_INFO_KIND
] = { .type
= NLA_STRING
},
734 [IFLA_INFO_DATA
] = { .type
= NLA_NESTED
},
737 static struct net
*get_net_ns_by_pid(pid_t pid
)
739 struct task_struct
*tsk
;
742 /* Lookup the network namespace */
743 net
= ERR_PTR(-ESRCH
);
745 tsk
= find_task_by_pid(pid
);
749 net
= get_net(tsk
->nsproxy
->net_ns
);
756 static int do_setlink(struct net_device
*dev
, struct ifinfomsg
*ifm
,
757 struct nlattr
**tb
, char *ifname
, int modified
)
759 int send_addr_notify
= 0;
762 if (tb
[IFLA_NET_NS_PID
]) {
764 net
= get_net_ns_by_pid(nla_get_u32(tb
[IFLA_NET_NS_PID
]));
769 err
= dev_change_net_namespace(dev
, net
, ifname
);
777 struct rtnl_link_ifmap
*u_map
;
780 if (!dev
->set_config
) {
785 if (!netif_device_present(dev
)) {
790 u_map
= nla_data(tb
[IFLA_MAP
]);
791 k_map
.mem_start
= (unsigned long) u_map
->mem_start
;
792 k_map
.mem_end
= (unsigned long) u_map
->mem_end
;
793 k_map
.base_addr
= (unsigned short) u_map
->base_addr
;
794 k_map
.irq
= (unsigned char) u_map
->irq
;
795 k_map
.dma
= (unsigned char) u_map
->dma
;
796 k_map
.port
= (unsigned char) u_map
->port
;
798 err
= dev
->set_config(dev
, &k_map
);
805 if (tb
[IFLA_ADDRESS
]) {
809 if (!dev
->set_mac_address
) {
814 if (!netif_device_present(dev
)) {
819 len
= sizeof(sa_family_t
) + dev
->addr_len
;
820 sa
= kmalloc(len
, GFP_KERNEL
);
825 sa
->sa_family
= dev
->type
;
826 memcpy(sa
->sa_data
, nla_data(tb
[IFLA_ADDRESS
]),
828 err
= dev
->set_mac_address(dev
, sa
);
832 send_addr_notify
= 1;
837 err
= dev_set_mtu(dev
, nla_get_u32(tb
[IFLA_MTU
]));
844 * Interface selected by interface index but interface
845 * name provided implies that a name change has been
848 if (ifm
->ifi_index
> 0 && ifname
[0]) {
849 err
= dev_change_name(dev
, ifname
);
855 if (tb
[IFLA_BROADCAST
]) {
856 nla_memcpy(dev
->broadcast
, tb
[IFLA_BROADCAST
], dev
->addr_len
);
857 send_addr_notify
= 1;
860 if (ifm
->ifi_flags
|| ifm
->ifi_change
) {
861 unsigned int flags
= ifm
->ifi_flags
;
863 /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
865 flags
= (flags
& ifm
->ifi_change
) |
866 (dev
->flags
& ~ifm
->ifi_change
);
867 dev_change_flags(dev
, flags
);
871 dev
->tx_queue_len
= nla_get_u32(tb
[IFLA_TXQLEN
]);
873 if (tb
[IFLA_OPERSTATE
])
874 set_operstate(dev
, nla_get_u8(tb
[IFLA_OPERSTATE
]));
876 if (tb
[IFLA_LINKMODE
]) {
877 write_lock_bh(&dev_base_lock
);
878 dev
->link_mode
= nla_get_u8(tb
[IFLA_LINKMODE
]);
879 write_unlock_bh(&dev_base_lock
);
885 if (err
< 0 && modified
&& net_ratelimit())
886 printk(KERN_WARNING
"A link change request failed with "
887 "some changes comitted already. Interface %s may "
888 "have been left with an inconsistent configuration, "
889 "please check.\n", dev
->name
);
891 if (send_addr_notify
)
892 call_netdevice_notifiers(NETDEV_CHANGEADDR
, dev
);
896 static int rtnl_setlink(struct sk_buff
*skb
, struct nlmsghdr
*nlh
, void *arg
)
898 struct net
*net
= skb
->sk
->sk_net
;
899 struct ifinfomsg
*ifm
;
900 struct net_device
*dev
;
902 struct nlattr
*tb
[IFLA_MAX
+1];
903 char ifname
[IFNAMSIZ
];
905 err
= nlmsg_parse(nlh
, sizeof(*ifm
), tb
, IFLA_MAX
, ifla_policy
);
910 nla_strlcpy(ifname
, tb
[IFLA_IFNAME
], IFNAMSIZ
);
915 ifm
= nlmsg_data(nlh
);
916 if (ifm
->ifi_index
> 0)
917 dev
= dev_get_by_index(net
, ifm
->ifi_index
);
918 else if (tb
[IFLA_IFNAME
])
919 dev
= dev_get_by_name(net
, ifname
);
928 if (tb
[IFLA_ADDRESS
] &&
929 nla_len(tb
[IFLA_ADDRESS
]) < dev
->addr_len
)
932 if (tb
[IFLA_BROADCAST
] &&
933 nla_len(tb
[IFLA_BROADCAST
]) < dev
->addr_len
)
936 err
= do_setlink(dev
, ifm
, tb
, ifname
, 0);
943 static int rtnl_dellink(struct sk_buff
*skb
, struct nlmsghdr
*nlh
, void *arg
)
945 struct net
*net
= skb
->sk
->sk_net
;
946 const struct rtnl_link_ops
*ops
;
947 struct net_device
*dev
;
948 struct ifinfomsg
*ifm
;
949 char ifname
[IFNAMSIZ
];
950 struct nlattr
*tb
[IFLA_MAX
+1];
953 err
= nlmsg_parse(nlh
, sizeof(*ifm
), tb
, IFLA_MAX
, ifla_policy
);
958 nla_strlcpy(ifname
, tb
[IFLA_IFNAME
], IFNAMSIZ
);
960 ifm
= nlmsg_data(nlh
);
961 if (ifm
->ifi_index
> 0)
962 dev
= __dev_get_by_index(net
, ifm
->ifi_index
);
963 else if (tb
[IFLA_IFNAME
])
964 dev
= __dev_get_by_name(net
, ifname
);
971 ops
= dev
->rtnl_link_ops
;
979 struct net_device
*rtnl_create_link(struct net
*net
, char *ifname
,
980 const struct rtnl_link_ops
*ops
, struct nlattr
*tb
[])
983 struct net_device
*dev
;
986 dev
= alloc_netdev(ops
->priv_size
, ifname
, ops
->setup
);
990 if (strchr(dev
->name
, '%')) {
991 err
= dev_alloc_name(dev
, dev
->name
);
997 dev
->rtnl_link_ops
= ops
;
1000 dev
->mtu
= nla_get_u32(tb
[IFLA_MTU
]);
1001 if (tb
[IFLA_ADDRESS
])
1002 memcpy(dev
->dev_addr
, nla_data(tb
[IFLA_ADDRESS
]),
1003 nla_len(tb
[IFLA_ADDRESS
]));
1004 if (tb
[IFLA_BROADCAST
])
1005 memcpy(dev
->broadcast
, nla_data(tb
[IFLA_BROADCAST
]),
1006 nla_len(tb
[IFLA_BROADCAST
]));
1007 if (tb
[IFLA_TXQLEN
])
1008 dev
->tx_queue_len
= nla_get_u32(tb
[IFLA_TXQLEN
]);
1009 if (tb
[IFLA_OPERSTATE
])
1010 set_operstate(dev
, nla_get_u8(tb
[IFLA_OPERSTATE
]));
1011 if (tb
[IFLA_LINKMODE
])
1012 dev
->link_mode
= nla_get_u8(tb
[IFLA_LINKMODE
]);
1019 return ERR_PTR(err
);
1022 static int rtnl_newlink(struct sk_buff
*skb
, struct nlmsghdr
*nlh
, void *arg
)
1024 struct net
*net
= skb
->sk
->sk_net
;
1025 const struct rtnl_link_ops
*ops
;
1026 struct net_device
*dev
;
1027 struct ifinfomsg
*ifm
;
1028 char kind
[MODULE_NAME_LEN
];
1029 char ifname
[IFNAMSIZ
];
1030 struct nlattr
*tb
[IFLA_MAX
+1];
1031 struct nlattr
*linkinfo
[IFLA_INFO_MAX
+1];
1037 err
= nlmsg_parse(nlh
, sizeof(*ifm
), tb
, IFLA_MAX
, ifla_policy
);
1041 if (tb
[IFLA_IFNAME
])
1042 nla_strlcpy(ifname
, tb
[IFLA_IFNAME
], IFNAMSIZ
);
1046 ifm
= nlmsg_data(nlh
);
1047 if (ifm
->ifi_index
> 0)
1048 dev
= __dev_get_by_index(net
, ifm
->ifi_index
);
1050 dev
= __dev_get_by_name(net
, ifname
);
1054 if (tb
[IFLA_LINKINFO
]) {
1055 err
= nla_parse_nested(linkinfo
, IFLA_INFO_MAX
,
1056 tb
[IFLA_LINKINFO
], ifla_info_policy
);
1060 memset(linkinfo
, 0, sizeof(linkinfo
));
1062 if (linkinfo
[IFLA_INFO_KIND
]) {
1063 nla_strlcpy(kind
, linkinfo
[IFLA_INFO_KIND
], sizeof(kind
));
1064 ops
= rtnl_link_ops_get(kind
);
1071 struct nlattr
*attr
[ops
? ops
->maxtype
+ 1 : 0], **data
= NULL
;
1074 if (ops
->maxtype
&& linkinfo
[IFLA_INFO_DATA
]) {
1075 err
= nla_parse_nested(attr
, ops
->maxtype
,
1076 linkinfo
[IFLA_INFO_DATA
],
1082 if (ops
->validate
) {
1083 err
= ops
->validate(tb
, data
);
1092 if (nlh
->nlmsg_flags
& NLM_F_EXCL
)
1094 if (nlh
->nlmsg_flags
& NLM_F_REPLACE
)
1097 if (linkinfo
[IFLA_INFO_DATA
]) {
1098 if (!ops
|| ops
!= dev
->rtnl_link_ops
||
1102 err
= ops
->changelink(dev
, tb
, data
);
1108 return do_setlink(dev
, ifm
, tb
, ifname
, modified
);
1111 if (!(nlh
->nlmsg_flags
& NLM_F_CREATE
))
1114 if (ifm
->ifi_index
|| ifm
->ifi_flags
|| ifm
->ifi_change
)
1116 if (tb
[IFLA_MAP
] || tb
[IFLA_MASTER
] || tb
[IFLA_PROTINFO
])
1123 request_module("rtnl-link-%s", kind
);
1125 ops
= rtnl_link_ops_get(kind
);
1134 snprintf(ifname
, IFNAMSIZ
, "%s%%d", ops
->kind
);
1136 dev
= rtnl_create_link(net
, ifname
, ops
, tb
);
1140 else if (ops
->newlink
)
1141 err
= ops
->newlink(dev
, tb
, data
);
1143 err
= register_netdevice(dev
);
1145 if (err
< 0 && !IS_ERR(dev
))
1151 static int rtnl_getlink(struct sk_buff
*skb
, struct nlmsghdr
* nlh
, void *arg
)
1153 struct net
*net
= skb
->sk
->sk_net
;
1154 struct ifinfomsg
*ifm
;
1155 struct nlattr
*tb
[IFLA_MAX
+1];
1156 struct net_device
*dev
= NULL
;
1157 struct sk_buff
*nskb
;
1160 err
= nlmsg_parse(nlh
, sizeof(*ifm
), tb
, IFLA_MAX
, ifla_policy
);
1164 ifm
= nlmsg_data(nlh
);
1165 if (ifm
->ifi_index
> 0) {
1166 dev
= dev_get_by_index(net
, ifm
->ifi_index
);
1172 nskb
= nlmsg_new(if_nlmsg_size(dev
), GFP_KERNEL
);
1178 err
= rtnl_fill_ifinfo(nskb
, dev
, RTM_NEWLINK
, NETLINK_CB(skb
).pid
,
1179 nlh
->nlmsg_seq
, 0, 0);
1181 /* -EMSGSIZE implies BUG in if_nlmsg_size */
1182 WARN_ON(err
== -EMSGSIZE
);
1186 err
= rtnl_unicast(nskb
, NETLINK_CB(skb
).pid
);
1193 static int rtnl_dump_all(struct sk_buff
*skb
, struct netlink_callback
*cb
)
1196 int s_idx
= cb
->family
;
1200 for (idx
=1; idx
<NPROTO
; idx
++) {
1201 int type
= cb
->nlh
->nlmsg_type
-RTM_BASE
;
1202 if (idx
< s_idx
|| idx
== PF_PACKET
)
1204 if (rtnl_msg_handlers
[idx
] == NULL
||
1205 rtnl_msg_handlers
[idx
][type
].dumpit
== NULL
)
1208 memset(&cb
->args
[0], 0, sizeof(cb
->args
));
1209 if (rtnl_msg_handlers
[idx
][type
].dumpit(skb
, cb
))
1217 void rtmsg_ifinfo(int type
, struct net_device
*dev
, unsigned change
)
1219 struct sk_buff
*skb
;
1222 skb
= nlmsg_new(if_nlmsg_size(dev
), GFP_KERNEL
);
1226 err
= rtnl_fill_ifinfo(skb
, dev
, type
, 0, 0, change
, 0);
1228 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
1229 WARN_ON(err
== -EMSGSIZE
);
1233 err
= rtnl_notify(skb
, 0, RTNLGRP_LINK
, NULL
, GFP_KERNEL
);
1236 rtnl_set_sk_err(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 rtnl_doit_func doit
;
1254 type
= nlh
->nlmsg_type
;
1260 /* All the messages must have at least 1 byte length */
1261 if (nlh
->nlmsg_len
< NLMSG_LENGTH(sizeof(struct rtgenmsg
)))
1264 family
= ((struct rtgenmsg
*)NLMSG_DATA(nlh
))->rtgen_family
;
1265 if (family
>= NPROTO
)
1266 return -EAFNOSUPPORT
;
1271 if (kind
!= 2 && security_netlink_recv(skb
, CAP_NET_ADMIN
))
1274 if (kind
== 2 && nlh
->nlmsg_flags
&NLM_F_DUMP
) {
1275 rtnl_dumpit_func dumpit
;
1277 dumpit
= rtnl_get_dumpit(family
, type
);
1282 err
= netlink_dump_start(rtnl
, skb
, nlh
, dumpit
, NULL
);
1287 memset(rta_buf
, 0, (rtattr_max
* sizeof(struct rtattr
*)));
1289 min_len
= rtm_min
[sz_idx
];
1290 if (nlh
->nlmsg_len
< min_len
)
1293 if (nlh
->nlmsg_len
> min_len
) {
1294 int attrlen
= nlh
->nlmsg_len
- NLMSG_ALIGN(min_len
);
1295 struct rtattr
*attr
= (void*)nlh
+ NLMSG_ALIGN(min_len
);
1297 while (RTA_OK(attr
, attrlen
)) {
1298 unsigned flavor
= attr
->rta_type
;
1300 if (flavor
> rta_max
[sz_idx
])
1302 rta_buf
[flavor
-1] = attr
;
1304 attr
= RTA_NEXT(attr
, attrlen
);
1308 doit
= rtnl_get_doit(family
, type
);
1312 return doit(skb
, nlh
, (void *)&rta_buf
[0]);
1315 static void rtnetlink_rcv(struct sk_buff
*skb
)
1318 netlink_rcv_skb(skb
, &rtnetlink_rcv_msg
);
1322 static int rtnetlink_event(struct notifier_block
*this, unsigned long event
, void *ptr
)
1324 struct net_device
*dev
= ptr
;
1326 if (dev
->nd_net
!= &init_net
)
1330 case NETDEV_UNREGISTER
:
1331 rtmsg_ifinfo(RTM_DELLINK
, dev
, ~0U);
1333 case NETDEV_REGISTER
:
1334 rtmsg_ifinfo(RTM_NEWLINK
, dev
, ~0U);
1338 rtmsg_ifinfo(RTM_NEWLINK
, dev
, IFF_UP
|IFF_RUNNING
);
1341 case NETDEV_GOING_DOWN
:
1344 rtmsg_ifinfo(RTM_NEWLINK
, dev
, 0);
1350 static struct notifier_block rtnetlink_dev_notifier
= {
1351 .notifier_call
= rtnetlink_event
,
1354 void __init
rtnetlink_init(void)
1359 for (i
= 0; i
< ARRAY_SIZE(rta_max
); i
++)
1360 if (rta_max
[i
] > rtattr_max
)
1361 rtattr_max
= rta_max
[i
];
1362 rta_buf
= kmalloc(rtattr_max
* sizeof(struct rtattr
*), GFP_KERNEL
);
1364 panic("rtnetlink_init: cannot allocate rta_buf\n");
1366 rtnl
= netlink_kernel_create(&init_net
, NETLINK_ROUTE
, RTNLGRP_MAX
,
1367 rtnetlink_rcv
, &rtnl_mutex
, THIS_MODULE
);
1369 panic("rtnetlink_init: cannot initialize rtnetlink\n");
1370 netlink_set_nonroot(NETLINK_ROUTE
, NL_NONROOT_RECV
);
1371 register_netdevice_notifier(&rtnetlink_dev_notifier
);
1373 rtnl_register(PF_UNSPEC
, RTM_GETLINK
, rtnl_getlink
, rtnl_dump_ifinfo
);
1374 rtnl_register(PF_UNSPEC
, RTM_SETLINK
, rtnl_setlink
, NULL
);
1375 rtnl_register(PF_UNSPEC
, RTM_NEWLINK
, rtnl_newlink
, NULL
);
1376 rtnl_register(PF_UNSPEC
, RTM_DELLINK
, rtnl_dellink
, NULL
);
1378 rtnl_register(PF_UNSPEC
, RTM_GETADDR
, NULL
, rtnl_dump_all
);
1379 rtnl_register(PF_UNSPEC
, RTM_GETROUTE
, NULL
, rtnl_dump_all
);
1382 EXPORT_SYMBOL(__rta_fill
);
1383 EXPORT_SYMBOL(rtattr_strlcpy
);
1384 EXPORT_SYMBOL(rtattr_parse
);
1385 EXPORT_SYMBOL(__rtattr_parse_nested_compat
);
1386 EXPORT_SYMBOL(rtnetlink_put_metrics
);
1387 EXPORT_SYMBOL(rtnl_lock
);
1388 EXPORT_SYMBOL(rtnl_trylock
);
1389 EXPORT_SYMBOL(rtnl_unlock
);
1390 EXPORT_SYMBOL(rtnl_unicast
);
1391 EXPORT_SYMBOL(rtnl_notify
);
1392 EXPORT_SYMBOL(rtnl_set_sk_err
);
1393 EXPORT_SYMBOL(rtnl_create_link
);
1394 EXPORT_SYMBOL(ifla_policy
);