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/config.h>
20 #include <linux/errno.h>
21 #include <linux/module.h>
22 #include <linux/types.h>
23 #include <linux/socket.h>
24 #include <linux/kernel.h>
25 #include <linux/sched.h>
26 #include <linux/timer.h>
27 #include <linux/string.h>
28 #include <linux/sockios.h>
29 #include <linux/net.h>
30 #include <linux/fcntl.h>
32 #include <linux/slab.h>
33 #include <linux/interrupt.h>
34 #include <linux/capability.h>
35 #include <linux/skbuff.h>
36 #include <linux/init.h>
37 #include <linux/security.h>
38 #include <linux/mutex.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/netlink.h>
55 static DEFINE_MUTEX(rtnl_mutex
);
59 mutex_lock(&rtnl_mutex
);
62 void __rtnl_unlock(void)
64 mutex_unlock(&rtnl_mutex
);
67 void rtnl_unlock(void)
69 mutex_unlock(&rtnl_mutex
);
70 if (rtnl
&& rtnl
->sk_receive_queue
.qlen
)
71 rtnl
->sk_data_ready(rtnl
, 0);
75 int rtnl_trylock(void)
77 return mutex_trylock(&rtnl_mutex
);
80 int rtattr_parse(struct rtattr
*tb
[], int maxattr
, struct rtattr
*rta
, int len
)
82 memset(tb
, 0, sizeof(struct rtattr
*)*maxattr
);
84 while (RTA_OK(rta
, len
)) {
85 unsigned flavor
= rta
->rta_type
;
86 if (flavor
&& flavor
<= maxattr
)
88 rta
= RTA_NEXT(rta
, len
);
95 struct rtnetlink_link
* rtnetlink_links
[NPROTO
];
97 static const int rtm_min
[RTM_NR_FAMILIES
] =
99 [RTM_FAM(RTM_NEWLINK
)] = NLMSG_LENGTH(sizeof(struct ifinfomsg
)),
100 [RTM_FAM(RTM_NEWADDR
)] = NLMSG_LENGTH(sizeof(struct ifaddrmsg
)),
101 [RTM_FAM(RTM_NEWROUTE
)] = NLMSG_LENGTH(sizeof(struct rtmsg
)),
102 [RTM_FAM(RTM_NEWNEIGH
)] = NLMSG_LENGTH(sizeof(struct ndmsg
)),
103 [RTM_FAM(RTM_NEWRULE
)] = NLMSG_LENGTH(sizeof(struct rtmsg
)),
104 [RTM_FAM(RTM_NEWQDISC
)] = NLMSG_LENGTH(sizeof(struct tcmsg
)),
105 [RTM_FAM(RTM_NEWTCLASS
)] = NLMSG_LENGTH(sizeof(struct tcmsg
)),
106 [RTM_FAM(RTM_NEWTFILTER
)] = NLMSG_LENGTH(sizeof(struct tcmsg
)),
107 [RTM_FAM(RTM_NEWACTION
)] = NLMSG_LENGTH(sizeof(struct tcamsg
)),
108 [RTM_FAM(RTM_NEWPREFIX
)] = NLMSG_LENGTH(sizeof(struct rtgenmsg
)),
109 [RTM_FAM(RTM_GETMULTICAST
)] = NLMSG_LENGTH(sizeof(struct rtgenmsg
)),
110 [RTM_FAM(RTM_GETANYCAST
)] = NLMSG_LENGTH(sizeof(struct rtgenmsg
)),
111 [RTM_FAM(RTM_NEWNEIGHTBL
)] = NLMSG_LENGTH(sizeof(struct ndtmsg
)),
114 static const int rta_max
[RTM_NR_FAMILIES
] =
116 [RTM_FAM(RTM_NEWLINK
)] = IFLA_MAX
,
117 [RTM_FAM(RTM_NEWADDR
)] = IFA_MAX
,
118 [RTM_FAM(RTM_NEWROUTE
)] = RTA_MAX
,
119 [RTM_FAM(RTM_NEWNEIGH
)] = NDA_MAX
,
120 [RTM_FAM(RTM_NEWRULE
)] = RTA_MAX
,
121 [RTM_FAM(RTM_NEWQDISC
)] = TCA_MAX
,
122 [RTM_FAM(RTM_NEWTCLASS
)] = TCA_MAX
,
123 [RTM_FAM(RTM_NEWTFILTER
)] = TCA_MAX
,
124 [RTM_FAM(RTM_NEWACTION
)] = TCAA_MAX
,
125 [RTM_FAM(RTM_NEWNEIGHTBL
)] = NDTA_MAX
,
128 void __rta_fill(struct sk_buff
*skb
, int attrtype
, int attrlen
, const void *data
)
131 int size
= RTA_LENGTH(attrlen
);
133 rta
= (struct rtattr
*)skb_put(skb
, RTA_ALIGN(size
));
134 rta
->rta_type
= attrtype
;
136 memcpy(RTA_DATA(rta
), data
, attrlen
);
137 memset(RTA_DATA(rta
) + attrlen
, 0, RTA_ALIGN(size
) - size
);
140 size_t rtattr_strlcpy(char *dest
, const struct rtattr
*rta
, size_t size
)
142 size_t ret
= RTA_PAYLOAD(rta
);
143 char *src
= RTA_DATA(rta
);
145 if (ret
> 0 && src
[ret
- 1] == '\0')
148 size_t len
= (ret
>= size
) ? size
- 1 : ret
;
149 memset(dest
, 0, size
);
150 memcpy(dest
, src
, len
);
155 int rtnetlink_send(struct sk_buff
*skb
, u32 pid
, unsigned group
, int echo
)
159 NETLINK_CB(skb
).dst_group
= group
;
161 atomic_inc(&skb
->users
);
162 netlink_broadcast(rtnl
, skb
, pid
, group
, GFP_KERNEL
);
164 err
= netlink_unicast(rtnl
, skb
, pid
, MSG_DONTWAIT
);
168 int rtnetlink_put_metrics(struct sk_buff
*skb
, u32
*metrics
)
170 struct rtattr
*mx
= (struct rtattr
*)skb
->tail
;
173 RTA_PUT(skb
, RTA_METRICS
, 0, NULL
);
174 for (i
=0; i
<RTAX_MAX
; i
++) {
176 RTA_PUT(skb
, i
+1, sizeof(u32
), metrics
+i
);
178 mx
->rta_len
= skb
->tail
- (u8
*)mx
;
179 if (mx
->rta_len
== RTA_LENGTH(0))
180 skb_trim(skb
, (u8
*)mx
- skb
->data
);
184 skb_trim(skb
, (u8
*)mx
- skb
->data
);
189 static void set_operstate(struct net_device
*dev
, unsigned char transition
)
191 unsigned char operstate
= dev
->operstate
;
195 if ((operstate
== IF_OPER_DORMANT
||
196 operstate
== IF_OPER_UNKNOWN
) &&
198 operstate
= IF_OPER_UP
;
201 case IF_OPER_DORMANT
:
202 if (operstate
== IF_OPER_UP
||
203 operstate
== IF_OPER_UNKNOWN
)
204 operstate
= IF_OPER_DORMANT
;
208 if (dev
->operstate
!= operstate
) {
209 write_lock_bh(&dev_base_lock
);
210 dev
->operstate
= operstate
;
211 write_unlock_bh(&dev_base_lock
);
212 netdev_state_change(dev
);
216 static int rtnetlink_fill_ifinfo(struct sk_buff
*skb
, struct net_device
*dev
,
217 int type
, u32 pid
, u32 seq
, u32 change
,
221 struct nlmsghdr
*nlh
;
222 unsigned char *b
= skb
->tail
;
224 nlh
= NLMSG_NEW(skb
, pid
, seq
, type
, sizeof(*r
), flags
);
226 r
->ifi_family
= AF_UNSPEC
;
228 r
->ifi_type
= dev
->type
;
229 r
->ifi_index
= dev
->ifindex
;
230 r
->ifi_flags
= dev_get_flags(dev
);
231 r
->ifi_change
= change
;
233 RTA_PUT(skb
, IFLA_IFNAME
, strlen(dev
->name
)+1, dev
->name
);
236 u32 txqlen
= dev
->tx_queue_len
;
237 RTA_PUT(skb
, IFLA_TXQLEN
, sizeof(txqlen
), &txqlen
);
241 u32 weight
= dev
->weight
;
242 RTA_PUT(skb
, IFLA_WEIGHT
, sizeof(weight
), &weight
);
246 u8 operstate
= netif_running(dev
)?dev
->operstate
:IF_OPER_DOWN
;
247 u8 link_mode
= dev
->link_mode
;
248 RTA_PUT(skb
, IFLA_OPERSTATE
, sizeof(operstate
), &operstate
);
249 RTA_PUT(skb
, IFLA_LINKMODE
, sizeof(link_mode
), &link_mode
);
253 struct rtnl_link_ifmap map
= {
254 .mem_start
= dev
->mem_start
,
255 .mem_end
= dev
->mem_end
,
256 .base_addr
= dev
->base_addr
,
259 .port
= dev
->if_port
,
261 RTA_PUT(skb
, IFLA_MAP
, sizeof(map
), &map
);
265 RTA_PUT(skb
, IFLA_ADDRESS
, dev
->addr_len
, dev
->dev_addr
);
266 RTA_PUT(skb
, IFLA_BROADCAST
, dev
->addr_len
, dev
->broadcast
);
271 RTA_PUT(skb
, IFLA_MTU
, sizeof(mtu
), &mtu
);
274 if (dev
->ifindex
!= dev
->iflink
) {
275 u32 iflink
= dev
->iflink
;
276 RTA_PUT(skb
, IFLA_LINK
, sizeof(iflink
), &iflink
);
279 if (dev
->qdisc_sleeping
)
280 RTA_PUT(skb
, IFLA_QDISC
,
281 strlen(dev
->qdisc_sleeping
->ops
->id
) + 1,
282 dev
->qdisc_sleeping
->ops
->id
);
285 u32 master
= dev
->master
->ifindex
;
286 RTA_PUT(skb
, IFLA_MASTER
, sizeof(master
), &master
);
289 if (dev
->get_stats
) {
290 unsigned long *stats
= (unsigned long*)dev
->get_stats(dev
);
295 int n
= sizeof(struct rtnl_link_stats
)/4;
297 a
= __RTA_PUT(skb
, IFLA_STATS
, n
*4);
303 nlh
->nlmsg_len
= skb
->tail
- b
;
308 skb_trim(skb
, b
- skb
->data
);
312 static int rtnetlink_dump_ifinfo(struct sk_buff
*skb
, struct netlink_callback
*cb
)
315 int s_idx
= cb
->args
[0];
316 struct net_device
*dev
;
318 read_lock(&dev_base_lock
);
319 for (dev
=dev_base
, idx
=0; dev
; dev
= dev
->next
, idx
++) {
322 if (rtnetlink_fill_ifinfo(skb
, dev
, RTM_NEWLINK
,
323 NETLINK_CB(cb
->skb
).pid
,
324 cb
->nlh
->nlmsg_seq
, 0,
328 read_unlock(&dev_base_lock
);
334 static int do_setlink(struct sk_buff
*skb
, struct nlmsghdr
*nlh
, void *arg
)
336 struct ifinfomsg
*ifm
= NLMSG_DATA(nlh
);
337 struct rtattr
**ida
= arg
;
338 struct net_device
*dev
;
339 int err
, send_addr_notify
= 0;
341 if (ifm
->ifi_index
>= 0)
342 dev
= dev_get_by_index(ifm
->ifi_index
);
343 else if (ida
[IFLA_IFNAME
- 1]) {
344 char ifname
[IFNAMSIZ
];
346 if (rtattr_strlcpy(ifname
, ida
[IFLA_IFNAME
- 1],
347 IFNAMSIZ
) >= IFNAMSIZ
)
349 dev
= dev_get_by_name(ifname
);
359 dev_change_flags(dev
, ifm
->ifi_flags
);
361 if (ida
[IFLA_MAP
- 1]) {
362 struct rtnl_link_ifmap
*u_map
;
365 if (!dev
->set_config
) {
370 if (!netif_device_present(dev
)) {
375 if (ida
[IFLA_MAP
- 1]->rta_len
!= RTA_LENGTH(sizeof(*u_map
)))
378 u_map
= RTA_DATA(ida
[IFLA_MAP
- 1]);
380 k_map
.mem_start
= (unsigned long) u_map
->mem_start
;
381 k_map
.mem_end
= (unsigned long) u_map
->mem_end
;
382 k_map
.base_addr
= (unsigned short) u_map
->base_addr
;
383 k_map
.irq
= (unsigned char) u_map
->irq
;
384 k_map
.dma
= (unsigned char) u_map
->dma
;
385 k_map
.port
= (unsigned char) u_map
->port
;
387 err
= dev
->set_config(dev
, &k_map
);
393 if (ida
[IFLA_ADDRESS
- 1]) {
394 if (!dev
->set_mac_address
) {
398 if (!netif_device_present(dev
)) {
402 if (ida
[IFLA_ADDRESS
- 1]->rta_len
!= RTA_LENGTH(dev
->addr_len
))
405 err
= dev
->set_mac_address(dev
, RTA_DATA(ida
[IFLA_ADDRESS
- 1]));
408 send_addr_notify
= 1;
411 if (ida
[IFLA_BROADCAST
- 1]) {
412 if (ida
[IFLA_BROADCAST
- 1]->rta_len
!= RTA_LENGTH(dev
->addr_len
))
414 memcpy(dev
->broadcast
, RTA_DATA(ida
[IFLA_BROADCAST
- 1]),
416 send_addr_notify
= 1;
419 if (ida
[IFLA_MTU
- 1]) {
420 if (ida
[IFLA_MTU
- 1]->rta_len
!= RTA_LENGTH(sizeof(u32
)))
422 err
= dev_set_mtu(dev
, *((u32
*) RTA_DATA(ida
[IFLA_MTU
- 1])));
429 if (ida
[IFLA_TXQLEN
- 1]) {
430 if (ida
[IFLA_TXQLEN
- 1]->rta_len
!= RTA_LENGTH(sizeof(u32
)))
433 dev
->tx_queue_len
= *((u32
*) RTA_DATA(ida
[IFLA_TXQLEN
- 1]));
436 if (ida
[IFLA_WEIGHT
- 1]) {
437 if (ida
[IFLA_WEIGHT
- 1]->rta_len
!= RTA_LENGTH(sizeof(u32
)))
440 dev
->weight
= *((u32
*) RTA_DATA(ida
[IFLA_WEIGHT
- 1]));
443 if (ida
[IFLA_OPERSTATE
- 1]) {
444 if (ida
[IFLA_OPERSTATE
- 1]->rta_len
!= RTA_LENGTH(sizeof(u8
)))
447 set_operstate(dev
, *((u8
*) RTA_DATA(ida
[IFLA_OPERSTATE
- 1])));
450 if (ida
[IFLA_LINKMODE
- 1]) {
451 if (ida
[IFLA_LINKMODE
- 1]->rta_len
!= RTA_LENGTH(sizeof(u8
)))
454 write_lock_bh(&dev_base_lock
);
455 dev
->link_mode
= *((u8
*) RTA_DATA(ida
[IFLA_LINKMODE
- 1]));
456 write_unlock_bh(&dev_base_lock
);
459 if (ifm
->ifi_index
>= 0 && ida
[IFLA_IFNAME
- 1]) {
460 char ifname
[IFNAMSIZ
];
462 if (rtattr_strlcpy(ifname
, ida
[IFLA_IFNAME
- 1],
463 IFNAMSIZ
) >= IFNAMSIZ
)
465 err
= dev_change_name(dev
, ifname
);
473 if (send_addr_notify
)
474 call_netdevice_notifiers(NETDEV_CHANGEADDR
, dev
);
480 static int rtnetlink_dump_all(struct sk_buff
*skb
, struct netlink_callback
*cb
)
483 int s_idx
= cb
->family
;
487 for (idx
=1; idx
<NPROTO
; idx
++) {
488 int type
= cb
->nlh
->nlmsg_type
-RTM_BASE
;
489 if (idx
< s_idx
|| idx
== PF_PACKET
)
491 if (rtnetlink_links
[idx
] == NULL
||
492 rtnetlink_links
[idx
][type
].dumpit
== NULL
)
495 memset(&cb
->args
[0], 0, sizeof(cb
->args
));
496 if (rtnetlink_links
[idx
][type
].dumpit(skb
, cb
))
504 void rtmsg_ifinfo(int type
, struct net_device
*dev
, unsigned change
)
507 int size
= NLMSG_SPACE(sizeof(struct ifinfomsg
) +
508 sizeof(struct rtnl_link_ifmap
) +
509 sizeof(struct rtnl_link_stats
) + 128);
511 skb
= alloc_skb(size
, GFP_KERNEL
);
515 if (rtnetlink_fill_ifinfo(skb
, dev
, type
, 0, 0, change
, 0) < 0) {
519 NETLINK_CB(skb
).dst_group
= RTNLGRP_LINK
;
520 netlink_broadcast(rtnl
, skb
, 0, RTNLGRP_LINK
, GFP_KERNEL
);
523 /* Protected by RTNL sempahore. */
524 static struct rtattr
**rta_buf
;
525 static int rtattr_max
;
527 /* Process one rtnetlink message. */
529 static __inline__
int
530 rtnetlink_rcv_msg(struct sk_buff
*skb
, struct nlmsghdr
*nlh
, int *errp
)
532 struct rtnetlink_link
*link
;
533 struct rtnetlink_link
*link_tab
;
540 /* Only requests are handled by kernel now */
541 if (!(nlh
->nlmsg_flags
&NLM_F_REQUEST
))
544 type
= nlh
->nlmsg_type
;
546 /* A control message: ignore them */
550 /* Unknown message: reply with EINVAL */
556 /* All the messages must have at least 1 byte length */
557 if (nlh
->nlmsg_len
< NLMSG_LENGTH(sizeof(struct rtgenmsg
)))
560 family
= ((struct rtgenmsg
*)NLMSG_DATA(nlh
))->rtgen_family
;
561 if (family
>= NPROTO
) {
562 *errp
= -EAFNOSUPPORT
;
566 link_tab
= rtnetlink_links
[family
];
567 if (link_tab
== NULL
)
568 link_tab
= rtnetlink_links
[PF_UNSPEC
];
569 link
= &link_tab
[type
];
574 if (kind
!= 2 && security_netlink_recv(skb
)) {
579 if (kind
== 2 && nlh
->nlmsg_flags
&NLM_F_DUMP
) {
580 if (link
->dumpit
== NULL
)
581 link
= &(rtnetlink_links
[PF_UNSPEC
][type
]);
583 if (link
->dumpit
== NULL
)
586 if ((*errp
= netlink_dump_start(rtnl
, skb
, nlh
,
587 link
->dumpit
, NULL
)) != 0) {
591 netlink_queue_skip(nlh
, skb
);
595 memset(rta_buf
, 0, (rtattr_max
* sizeof(struct rtattr
*)));
597 min_len
= rtm_min
[sz_idx
];
598 if (nlh
->nlmsg_len
< min_len
)
601 if (nlh
->nlmsg_len
> min_len
) {
602 int attrlen
= nlh
->nlmsg_len
- NLMSG_ALIGN(min_len
);
603 struct rtattr
*attr
= (void*)nlh
+ NLMSG_ALIGN(min_len
);
605 while (RTA_OK(attr
, attrlen
)) {
606 unsigned flavor
= attr
->rta_type
;
608 if (flavor
> rta_max
[sz_idx
])
610 rta_buf
[flavor
-1] = attr
;
612 attr
= RTA_NEXT(attr
, attrlen
);
616 if (link
->doit
== NULL
)
617 link
= &(rtnetlink_links
[PF_UNSPEC
][type
]);
618 if (link
->doit
== NULL
)
620 err
= link
->doit(skb
, nlh
, (void *)&rta_buf
[0]);
630 static void rtnetlink_rcv(struct sock
*sk
, int len
)
632 unsigned int qlen
= 0;
635 mutex_lock(&rtnl_mutex
);
636 netlink_run_queue(sk
, &qlen
, &rtnetlink_rcv_msg
);
637 mutex_unlock(&rtnl_mutex
);
643 static struct rtnetlink_link link_rtnetlink_table
[RTM_NR_MSGTYPES
] =
645 [RTM_GETLINK
- RTM_BASE
] = { .dumpit
= rtnetlink_dump_ifinfo
},
646 [RTM_SETLINK
- RTM_BASE
] = { .doit
= do_setlink
},
647 [RTM_GETADDR
- RTM_BASE
] = { .dumpit
= rtnetlink_dump_all
},
648 [RTM_GETROUTE
- RTM_BASE
] = { .dumpit
= rtnetlink_dump_all
},
649 [RTM_NEWNEIGH
- RTM_BASE
] = { .doit
= neigh_add
},
650 [RTM_DELNEIGH
- RTM_BASE
] = { .doit
= neigh_delete
},
651 [RTM_GETNEIGH
- RTM_BASE
] = { .dumpit
= neigh_dump_info
},
652 [RTM_GETRULE
- RTM_BASE
] = { .dumpit
= rtnetlink_dump_all
},
653 [RTM_GETNEIGHTBL
- RTM_BASE
] = { .dumpit
= neightbl_dump_info
},
654 [RTM_SETNEIGHTBL
- RTM_BASE
] = { .doit
= neightbl_set
},
657 static int rtnetlink_event(struct notifier_block
*this, unsigned long event
, void *ptr
)
659 struct net_device
*dev
= ptr
;
661 case NETDEV_UNREGISTER
:
662 rtmsg_ifinfo(RTM_DELLINK
, dev
, ~0U);
664 case NETDEV_REGISTER
:
665 rtmsg_ifinfo(RTM_NEWLINK
, dev
, ~0U);
669 rtmsg_ifinfo(RTM_NEWLINK
, dev
, IFF_UP
|IFF_RUNNING
);
672 case NETDEV_GOING_DOWN
:
675 rtmsg_ifinfo(RTM_NEWLINK
, dev
, 0);
681 static struct notifier_block rtnetlink_dev_notifier
= {
682 .notifier_call
= rtnetlink_event
,
685 void __init
rtnetlink_init(void)
690 for (i
= 0; i
< ARRAY_SIZE(rta_max
); i
++)
691 if (rta_max
[i
] > rtattr_max
)
692 rtattr_max
= rta_max
[i
];
693 rta_buf
= kmalloc(rtattr_max
* sizeof(struct rtattr
*), GFP_KERNEL
);
695 panic("rtnetlink_init: cannot allocate rta_buf\n");
697 rtnl
= netlink_kernel_create(NETLINK_ROUTE
, RTNLGRP_MAX
, rtnetlink_rcv
,
700 panic("rtnetlink_init: cannot initialize rtnetlink\n");
701 netlink_set_nonroot(NETLINK_ROUTE
, NL_NONROOT_RECV
);
702 register_netdevice_notifier(&rtnetlink_dev_notifier
);
703 rtnetlink_links
[PF_UNSPEC
] = link_rtnetlink_table
;
704 rtnetlink_links
[PF_PACKET
] = link_rtnetlink_table
;
707 EXPORT_SYMBOL(__rta_fill
);
708 EXPORT_SYMBOL(rtattr_strlcpy
);
709 EXPORT_SYMBOL(rtattr_parse
);
710 EXPORT_SYMBOL(rtnetlink_links
);
711 EXPORT_SYMBOL(rtnetlink_put_metrics
);
713 EXPORT_SYMBOL(rtnl_lock
);
714 EXPORT_SYMBOL(rtnl_trylock
);
715 EXPORT_SYMBOL(rtnl_unlock
);