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/pci.h>
40 #include <asm/uaccess.h>
41 #include <asm/system.h>
43 #include <linux/inet.h>
44 #include <linux/netdevice.h>
46 #include <net/protocol.h>
48 #include <net/route.h>
51 #include <net/pkt_sched.h>
52 #include <net/fib_rules.h>
53 #include <net/rtnetlink.h>
54 #include <net/net_namespace.h>
58 rtnl_dumpit_func dumpit
;
59 rtnl_calcit_func calcit
;
62 static DEFINE_MUTEX(rtnl_mutex
);
63 static u16 min_ifinfo_dump_size
;
67 mutex_lock(&rtnl_mutex
);
69 EXPORT_SYMBOL(rtnl_lock
);
71 void __rtnl_unlock(void)
73 mutex_unlock(&rtnl_mutex
);
76 void rtnl_unlock(void)
78 /* This fellow will unlock it for us. */
81 EXPORT_SYMBOL(rtnl_unlock
);
83 int rtnl_trylock(void)
85 return mutex_trylock(&rtnl_mutex
);
87 EXPORT_SYMBOL(rtnl_trylock
);
89 int rtnl_is_locked(void)
91 return mutex_is_locked(&rtnl_mutex
);
93 EXPORT_SYMBOL(rtnl_is_locked
);
95 #ifdef CONFIG_PROVE_LOCKING
96 int lockdep_rtnl_is_held(void)
98 return lockdep_is_held(&rtnl_mutex
);
100 EXPORT_SYMBOL(lockdep_rtnl_is_held
);
101 #endif /* #ifdef CONFIG_PROVE_LOCKING */
103 static struct rtnl_link
*rtnl_msg_handlers
[RTNL_FAMILY_MAX
+ 1];
105 static inline int rtm_msgindex(int msgtype
)
107 int msgindex
= msgtype
- RTM_BASE
;
110 * msgindex < 0 implies someone tried to register a netlink
111 * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
112 * the message type has not been added to linux/rtnetlink.h
114 BUG_ON(msgindex
< 0 || msgindex
>= RTM_NR_MSGTYPES
);
119 static rtnl_doit_func
rtnl_get_doit(int protocol
, int msgindex
)
121 struct rtnl_link
*tab
;
123 if (protocol
<= RTNL_FAMILY_MAX
)
124 tab
= rtnl_msg_handlers
[protocol
];
128 if (tab
== NULL
|| tab
[msgindex
].doit
== NULL
)
129 tab
= rtnl_msg_handlers
[PF_UNSPEC
];
131 return tab
? tab
[msgindex
].doit
: NULL
;
134 static rtnl_dumpit_func
rtnl_get_dumpit(int protocol
, int msgindex
)
136 struct rtnl_link
*tab
;
138 if (protocol
<= RTNL_FAMILY_MAX
)
139 tab
= rtnl_msg_handlers
[protocol
];
143 if (tab
== NULL
|| tab
[msgindex
].dumpit
== NULL
)
144 tab
= rtnl_msg_handlers
[PF_UNSPEC
];
146 return tab
? tab
[msgindex
].dumpit
: NULL
;
149 static rtnl_calcit_func
rtnl_get_calcit(int protocol
, int msgindex
)
151 struct rtnl_link
*tab
;
153 if (protocol
<= RTNL_FAMILY_MAX
)
154 tab
= rtnl_msg_handlers
[protocol
];
158 if (tab
== NULL
|| tab
[msgindex
].calcit
== NULL
)
159 tab
= rtnl_msg_handlers
[PF_UNSPEC
];
161 return tab
? tab
[msgindex
].calcit
: NULL
;
165 * __rtnl_register - Register a rtnetlink message type
166 * @protocol: Protocol family or PF_UNSPEC
167 * @msgtype: rtnetlink message type
168 * @doit: Function pointer called for each request message
169 * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
170 * @calcit: Function pointer to calc size of dump message
172 * Registers the specified function pointers (at least one of them has
173 * to be non-NULL) to be called whenever a request message for the
174 * specified protocol family and message type is received.
176 * The special protocol family PF_UNSPEC may be used to define fallback
177 * function pointers for the case when no entry for the specific protocol
180 * Returns 0 on success or a negative error code.
182 int __rtnl_register(int protocol
, int msgtype
,
183 rtnl_doit_func doit
, rtnl_dumpit_func dumpit
,
184 rtnl_calcit_func calcit
)
186 struct rtnl_link
*tab
;
189 BUG_ON(protocol
< 0 || protocol
> RTNL_FAMILY_MAX
);
190 msgindex
= rtm_msgindex(msgtype
);
192 tab
= rtnl_msg_handlers
[protocol
];
194 tab
= kcalloc(RTM_NR_MSGTYPES
, sizeof(*tab
), GFP_KERNEL
);
198 rtnl_msg_handlers
[protocol
] = tab
;
202 tab
[msgindex
].doit
= doit
;
205 tab
[msgindex
].dumpit
= dumpit
;
208 tab
[msgindex
].calcit
= calcit
;
212 EXPORT_SYMBOL_GPL(__rtnl_register
);
215 * rtnl_register - Register a rtnetlink message type
217 * Identical to __rtnl_register() but panics on failure. This is useful
218 * as failure of this function is very unlikely, it can only happen due
219 * to lack of memory when allocating the chain to store all message
220 * handlers for a protocol. Meant for use in init functions where lack
221 * of memory implies no sense in continuing.
223 void rtnl_register(int protocol
, int msgtype
,
224 rtnl_doit_func doit
, rtnl_dumpit_func dumpit
,
225 rtnl_calcit_func calcit
)
227 if (__rtnl_register(protocol
, msgtype
, doit
, dumpit
, calcit
) < 0)
228 panic("Unable to register rtnetlink message handler, "
229 "protocol = %d, message type = %d\n",
232 EXPORT_SYMBOL_GPL(rtnl_register
);
235 * rtnl_unregister - Unregister a rtnetlink message type
236 * @protocol: Protocol family or PF_UNSPEC
237 * @msgtype: rtnetlink message type
239 * Returns 0 on success or a negative error code.
241 int rtnl_unregister(int protocol
, int msgtype
)
245 BUG_ON(protocol
< 0 || protocol
> RTNL_FAMILY_MAX
);
246 msgindex
= rtm_msgindex(msgtype
);
248 if (rtnl_msg_handlers
[protocol
] == NULL
)
251 rtnl_msg_handlers
[protocol
][msgindex
].doit
= NULL
;
252 rtnl_msg_handlers
[protocol
][msgindex
].dumpit
= NULL
;
256 EXPORT_SYMBOL_GPL(rtnl_unregister
);
259 * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
260 * @protocol : Protocol family or PF_UNSPEC
262 * Identical to calling rtnl_unregster() for all registered message types
263 * of a certain protocol family.
265 void rtnl_unregister_all(int protocol
)
267 BUG_ON(protocol
< 0 || protocol
> RTNL_FAMILY_MAX
);
269 kfree(rtnl_msg_handlers
[protocol
]);
270 rtnl_msg_handlers
[protocol
] = NULL
;
272 EXPORT_SYMBOL_GPL(rtnl_unregister_all
);
274 static LIST_HEAD(link_ops
);
277 * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
278 * @ops: struct rtnl_link_ops * to register
280 * The caller must hold the rtnl_mutex. This function should be used
281 * by drivers that create devices during module initialization. It
282 * must be called before registering the devices.
284 * Returns 0 on success or a negative error code.
286 int __rtnl_link_register(struct rtnl_link_ops
*ops
)
289 ops
->dellink
= unregister_netdevice_queue
;
291 list_add_tail(&ops
->list
, &link_ops
);
294 EXPORT_SYMBOL_GPL(__rtnl_link_register
);
297 * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
298 * @ops: struct rtnl_link_ops * to register
300 * Returns 0 on success or a negative error code.
302 int rtnl_link_register(struct rtnl_link_ops
*ops
)
307 err
= __rtnl_link_register(ops
);
311 EXPORT_SYMBOL_GPL(rtnl_link_register
);
313 static void __rtnl_kill_links(struct net
*net
, struct rtnl_link_ops
*ops
)
315 struct net_device
*dev
;
316 LIST_HEAD(list_kill
);
318 for_each_netdev(net
, dev
) {
319 if (dev
->rtnl_link_ops
== ops
)
320 ops
->dellink(dev
, &list_kill
);
322 unregister_netdevice_many(&list_kill
);
326 * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
327 * @ops: struct rtnl_link_ops * to unregister
329 * The caller must hold the rtnl_mutex.
331 void __rtnl_link_unregister(struct rtnl_link_ops
*ops
)
336 __rtnl_kill_links(net
, ops
);
338 list_del(&ops
->list
);
340 EXPORT_SYMBOL_GPL(__rtnl_link_unregister
);
343 * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
344 * @ops: struct rtnl_link_ops * to unregister
346 void rtnl_link_unregister(struct rtnl_link_ops
*ops
)
349 __rtnl_link_unregister(ops
);
352 EXPORT_SYMBOL_GPL(rtnl_link_unregister
);
354 static const struct rtnl_link_ops
*rtnl_link_ops_get(const char *kind
)
356 const struct rtnl_link_ops
*ops
;
358 list_for_each_entry(ops
, &link_ops
, list
) {
359 if (!strcmp(ops
->kind
, kind
))
365 static size_t rtnl_link_get_size(const struct net_device
*dev
)
367 const struct rtnl_link_ops
*ops
= dev
->rtnl_link_ops
;
373 size
= nla_total_size(sizeof(struct nlattr
)) + /* IFLA_LINKINFO */
374 nla_total_size(strlen(ops
->kind
) + 1); /* IFLA_INFO_KIND */
377 /* IFLA_INFO_DATA + nested data */
378 size
+= nla_total_size(sizeof(struct nlattr
)) +
381 if (ops
->get_xstats_size
)
382 /* IFLA_INFO_XSTATS */
383 size
+= nla_total_size(ops
->get_xstats_size(dev
));
388 static LIST_HEAD(rtnl_af_ops
);
390 static const struct rtnl_af_ops
*rtnl_af_lookup(const int family
)
392 const struct rtnl_af_ops
*ops
;
394 list_for_each_entry(ops
, &rtnl_af_ops
, list
) {
395 if (ops
->family
== family
)
403 * __rtnl_af_register - Register rtnl_af_ops with rtnetlink.
404 * @ops: struct rtnl_af_ops * to register
406 * The caller must hold the rtnl_mutex.
408 * Returns 0 on success or a negative error code.
410 int __rtnl_af_register(struct rtnl_af_ops
*ops
)
412 list_add_tail(&ops
->list
, &rtnl_af_ops
);
415 EXPORT_SYMBOL_GPL(__rtnl_af_register
);
418 * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
419 * @ops: struct rtnl_af_ops * to register
421 * Returns 0 on success or a negative error code.
423 int rtnl_af_register(struct rtnl_af_ops
*ops
)
428 err
= __rtnl_af_register(ops
);
432 EXPORT_SYMBOL_GPL(rtnl_af_register
);
435 * __rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
436 * @ops: struct rtnl_af_ops * to unregister
438 * The caller must hold the rtnl_mutex.
440 void __rtnl_af_unregister(struct rtnl_af_ops
*ops
)
442 list_del(&ops
->list
);
444 EXPORT_SYMBOL_GPL(__rtnl_af_unregister
);
447 * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
448 * @ops: struct rtnl_af_ops * to unregister
450 void rtnl_af_unregister(struct rtnl_af_ops
*ops
)
453 __rtnl_af_unregister(ops
);
456 EXPORT_SYMBOL_GPL(rtnl_af_unregister
);
458 static size_t rtnl_link_get_af_size(const struct net_device
*dev
)
460 struct rtnl_af_ops
*af_ops
;
464 size
= nla_total_size(sizeof(struct nlattr
));
466 list_for_each_entry(af_ops
, &rtnl_af_ops
, list
) {
467 if (af_ops
->get_link_af_size
) {
468 /* AF_* + nested data */
469 size
+= nla_total_size(sizeof(struct nlattr
)) +
470 af_ops
->get_link_af_size(dev
);
477 static int rtnl_link_fill(struct sk_buff
*skb
, const struct net_device
*dev
)
479 const struct rtnl_link_ops
*ops
= dev
->rtnl_link_ops
;
480 struct nlattr
*linkinfo
, *data
;
483 linkinfo
= nla_nest_start(skb
, IFLA_LINKINFO
);
484 if (linkinfo
== NULL
)
487 if (nla_put_string(skb
, IFLA_INFO_KIND
, ops
->kind
) < 0)
488 goto err_cancel_link
;
489 if (ops
->fill_xstats
) {
490 err
= ops
->fill_xstats(skb
, dev
);
492 goto err_cancel_link
;
494 if (ops
->fill_info
) {
495 data
= nla_nest_start(skb
, IFLA_INFO_DATA
);
497 goto err_cancel_link
;
498 err
= ops
->fill_info(skb
, dev
);
500 goto err_cancel_data
;
501 nla_nest_end(skb
, data
);
504 nla_nest_end(skb
, linkinfo
);
508 nla_nest_cancel(skb
, data
);
510 nla_nest_cancel(skb
, linkinfo
);
515 static const int rtm_min
[RTM_NR_FAMILIES
] =
517 [RTM_FAM(RTM_NEWLINK
)] = NLMSG_LENGTH(sizeof(struct ifinfomsg
)),
518 [RTM_FAM(RTM_NEWADDR
)] = NLMSG_LENGTH(sizeof(struct ifaddrmsg
)),
519 [RTM_FAM(RTM_NEWROUTE
)] = NLMSG_LENGTH(sizeof(struct rtmsg
)),
520 [RTM_FAM(RTM_NEWRULE
)] = NLMSG_LENGTH(sizeof(struct fib_rule_hdr
)),
521 [RTM_FAM(RTM_NEWQDISC
)] = NLMSG_LENGTH(sizeof(struct tcmsg
)),
522 [RTM_FAM(RTM_NEWTCLASS
)] = NLMSG_LENGTH(sizeof(struct tcmsg
)),
523 [RTM_FAM(RTM_NEWTFILTER
)] = NLMSG_LENGTH(sizeof(struct tcmsg
)),
524 [RTM_FAM(RTM_NEWACTION
)] = NLMSG_LENGTH(sizeof(struct tcamsg
)),
525 [RTM_FAM(RTM_GETMULTICAST
)] = NLMSG_LENGTH(sizeof(struct rtgenmsg
)),
526 [RTM_FAM(RTM_GETANYCAST
)] = NLMSG_LENGTH(sizeof(struct rtgenmsg
)),
529 static const int rta_max
[RTM_NR_FAMILIES
] =
531 [RTM_FAM(RTM_NEWLINK
)] = IFLA_MAX
,
532 [RTM_FAM(RTM_NEWADDR
)] = IFA_MAX
,
533 [RTM_FAM(RTM_NEWROUTE
)] = RTA_MAX
,
534 [RTM_FAM(RTM_NEWRULE
)] = FRA_MAX
,
535 [RTM_FAM(RTM_NEWQDISC
)] = TCA_MAX
,
536 [RTM_FAM(RTM_NEWTCLASS
)] = TCA_MAX
,
537 [RTM_FAM(RTM_NEWTFILTER
)] = TCA_MAX
,
538 [RTM_FAM(RTM_NEWACTION
)] = TCAA_MAX
,
541 void __rta_fill(struct sk_buff
*skb
, int attrtype
, int attrlen
, const void *data
)
544 int size
= RTA_LENGTH(attrlen
);
546 rta
= (struct rtattr
*)skb_put(skb
, RTA_ALIGN(size
));
547 rta
->rta_type
= attrtype
;
549 memcpy(RTA_DATA(rta
), data
, attrlen
);
550 memset(RTA_DATA(rta
) + attrlen
, 0, RTA_ALIGN(size
) - size
);
552 EXPORT_SYMBOL(__rta_fill
);
554 int rtnetlink_send(struct sk_buff
*skb
, struct net
*net
, u32 pid
, unsigned group
, int echo
)
556 struct sock
*rtnl
= net
->rtnl
;
559 NETLINK_CB(skb
).dst_group
= group
;
561 atomic_inc(&skb
->users
);
562 netlink_broadcast(rtnl
, skb
, pid
, group
, GFP_KERNEL
);
564 err
= netlink_unicast(rtnl
, skb
, pid
, MSG_DONTWAIT
);
568 int rtnl_unicast(struct sk_buff
*skb
, struct net
*net
, u32 pid
)
570 struct sock
*rtnl
= net
->rtnl
;
572 return nlmsg_unicast(rtnl
, skb
, pid
);
574 EXPORT_SYMBOL(rtnl_unicast
);
576 void rtnl_notify(struct sk_buff
*skb
, struct net
*net
, u32 pid
, u32 group
,
577 struct nlmsghdr
*nlh
, gfp_t flags
)
579 struct sock
*rtnl
= net
->rtnl
;
583 report
= nlmsg_report(nlh
);
585 nlmsg_notify(rtnl
, skb
, pid
, group
, report
, flags
);
587 EXPORT_SYMBOL(rtnl_notify
);
589 void rtnl_set_sk_err(struct net
*net
, u32 group
, int error
)
591 struct sock
*rtnl
= net
->rtnl
;
593 netlink_set_err(rtnl
, 0, group
, error
);
595 EXPORT_SYMBOL(rtnl_set_sk_err
);
597 int rtnetlink_put_metrics(struct sk_buff
*skb
, u32
*metrics
)
602 mx
= nla_nest_start(skb
, RTA_METRICS
);
606 for (i
= 0; i
< RTAX_MAX
; i
++) {
609 NLA_PUT_U32(skb
, i
+1, metrics
[i
]);
614 nla_nest_cancel(skb
, mx
);
618 return nla_nest_end(skb
, mx
);
621 nla_nest_cancel(skb
, mx
);
624 EXPORT_SYMBOL(rtnetlink_put_metrics
);
626 int rtnl_put_cacheinfo(struct sk_buff
*skb
, struct dst_entry
*dst
, u32 id
,
627 u32 ts
, u32 tsage
, long expires
, u32 error
)
629 struct rta_cacheinfo ci
= {
630 .rta_lastuse
= jiffies_to_clock_t(jiffies
- dst
->lastuse
),
631 .rta_used
= dst
->__use
,
632 .rta_clntref
= atomic_read(&(dst
->__refcnt
)),
640 ci
.rta_expires
= jiffies_to_clock_t(expires
);
642 return nla_put(skb
, RTA_CACHEINFO
, sizeof(ci
), &ci
);
644 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo
);
646 static void set_operstate(struct net_device
*dev
, unsigned char transition
)
648 unsigned char operstate
= dev
->operstate
;
650 switch (transition
) {
652 if ((operstate
== IF_OPER_DORMANT
||
653 operstate
== IF_OPER_UNKNOWN
) &&
655 operstate
= IF_OPER_UP
;
658 case IF_OPER_DORMANT
:
659 if (operstate
== IF_OPER_UP
||
660 operstate
== IF_OPER_UNKNOWN
)
661 operstate
= IF_OPER_DORMANT
;
665 if (dev
->operstate
!= operstate
) {
666 write_lock_bh(&dev_base_lock
);
667 dev
->operstate
= operstate
;
668 write_unlock_bh(&dev_base_lock
);
669 netdev_state_change(dev
);
673 static unsigned int rtnl_dev_combine_flags(const struct net_device
*dev
,
674 const struct ifinfomsg
*ifm
)
676 unsigned int flags
= ifm
->ifi_flags
;
678 /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
680 flags
= (flags
& ifm
->ifi_change
) |
681 (dev
->flags
& ~ifm
->ifi_change
);
686 static void copy_rtnl_link_stats(struct rtnl_link_stats
*a
,
687 const struct rtnl_link_stats64
*b
)
689 a
->rx_packets
= b
->rx_packets
;
690 a
->tx_packets
= b
->tx_packets
;
691 a
->rx_bytes
= b
->rx_bytes
;
692 a
->tx_bytes
= b
->tx_bytes
;
693 a
->rx_errors
= b
->rx_errors
;
694 a
->tx_errors
= b
->tx_errors
;
695 a
->rx_dropped
= b
->rx_dropped
;
696 a
->tx_dropped
= b
->tx_dropped
;
698 a
->multicast
= b
->multicast
;
699 a
->collisions
= b
->collisions
;
701 a
->rx_length_errors
= b
->rx_length_errors
;
702 a
->rx_over_errors
= b
->rx_over_errors
;
703 a
->rx_crc_errors
= b
->rx_crc_errors
;
704 a
->rx_frame_errors
= b
->rx_frame_errors
;
705 a
->rx_fifo_errors
= b
->rx_fifo_errors
;
706 a
->rx_missed_errors
= b
->rx_missed_errors
;
708 a
->tx_aborted_errors
= b
->tx_aborted_errors
;
709 a
->tx_carrier_errors
= b
->tx_carrier_errors
;
710 a
->tx_fifo_errors
= b
->tx_fifo_errors
;
711 a
->tx_heartbeat_errors
= b
->tx_heartbeat_errors
;
712 a
->tx_window_errors
= b
->tx_window_errors
;
714 a
->rx_compressed
= b
->rx_compressed
;
715 a
->tx_compressed
= b
->tx_compressed
;
718 static void copy_rtnl_link_stats64(void *v
, const struct rtnl_link_stats64
*b
)
720 memcpy(v
, b
, sizeof(*b
));
724 static inline int rtnl_vfinfo_size(const struct net_device
*dev
)
726 if (dev
->dev
.parent
&& dev_is_pci(dev
->dev
.parent
)) {
728 int num_vfs
= dev_num_vf(dev
->dev
.parent
);
729 size_t size
= nla_total_size(sizeof(struct nlattr
));
730 size
+= nla_total_size(num_vfs
* sizeof(struct nlattr
));
732 (nla_total_size(sizeof(struct ifla_vf_mac
)) +
733 nla_total_size(sizeof(struct ifla_vf_vlan
)) +
734 nla_total_size(sizeof(struct ifla_vf_tx_rate
)));
740 static size_t rtnl_port_size(const struct net_device
*dev
)
742 size_t port_size
= nla_total_size(4) /* PORT_VF */
743 + nla_total_size(PORT_PROFILE_MAX
) /* PORT_PROFILE */
744 + nla_total_size(sizeof(struct ifla_port_vsi
))
746 + nla_total_size(PORT_UUID_MAX
) /* PORT_INSTANCE_UUID */
747 + nla_total_size(PORT_UUID_MAX
) /* PORT_HOST_UUID */
748 + nla_total_size(1) /* PROT_VDP_REQUEST */
749 + nla_total_size(2); /* PORT_VDP_RESPONSE */
750 size_t vf_ports_size
= nla_total_size(sizeof(struct nlattr
));
751 size_t vf_port_size
= nla_total_size(sizeof(struct nlattr
))
753 size_t port_self_size
= nla_total_size(sizeof(struct nlattr
))
756 if (!dev
->netdev_ops
->ndo_get_vf_port
|| !dev
->dev
.parent
)
758 if (dev_num_vf(dev
->dev
.parent
))
759 return port_self_size
+ vf_ports_size
+
760 vf_port_size
* dev_num_vf(dev
->dev
.parent
);
762 return port_self_size
;
765 static noinline
size_t if_nlmsg_size(const struct net_device
*dev
)
767 return NLMSG_ALIGN(sizeof(struct ifinfomsg
))
768 + nla_total_size(IFNAMSIZ
) /* IFLA_IFNAME */
769 + nla_total_size(IFALIASZ
) /* IFLA_IFALIAS */
770 + nla_total_size(IFNAMSIZ
) /* IFLA_QDISC */
771 + nla_total_size(sizeof(struct rtnl_link_ifmap
))
772 + nla_total_size(sizeof(struct rtnl_link_stats
))
773 + nla_total_size(sizeof(struct rtnl_link_stats64
))
774 + nla_total_size(MAX_ADDR_LEN
) /* IFLA_ADDRESS */
775 + nla_total_size(MAX_ADDR_LEN
) /* IFLA_BROADCAST */
776 + nla_total_size(4) /* IFLA_TXQLEN */
777 + nla_total_size(4) /* IFLA_WEIGHT */
778 + nla_total_size(4) /* IFLA_MTU */
779 + nla_total_size(4) /* IFLA_LINK */
780 + nla_total_size(4) /* IFLA_MASTER */
781 + nla_total_size(1) /* IFLA_OPERSTATE */
782 + nla_total_size(1) /* IFLA_LINKMODE */
783 + nla_total_size(4) /* IFLA_NUM_VF */
784 + rtnl_vfinfo_size(dev
) /* IFLA_VFINFO_LIST */
785 + rtnl_port_size(dev
) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
786 + rtnl_link_get_size(dev
) /* IFLA_LINKINFO */
787 + rtnl_link_get_af_size(dev
); /* IFLA_AF_SPEC */
790 static int rtnl_vf_ports_fill(struct sk_buff
*skb
, struct net_device
*dev
)
792 struct nlattr
*vf_ports
;
793 struct nlattr
*vf_port
;
797 vf_ports
= nla_nest_start(skb
, IFLA_VF_PORTS
);
801 for (vf
= 0; vf
< dev_num_vf(dev
->dev
.parent
); vf
++) {
802 vf_port
= nla_nest_start(skb
, IFLA_VF_PORT
);
804 goto nla_put_failure
;
805 NLA_PUT_U32(skb
, IFLA_PORT_VF
, vf
);
806 err
= dev
->netdev_ops
->ndo_get_vf_port(dev
, vf
, skb
);
807 if (err
== -EMSGSIZE
)
808 goto nla_put_failure
;
810 nla_nest_cancel(skb
, vf_port
);
813 nla_nest_end(skb
, vf_port
);
816 nla_nest_end(skb
, vf_ports
);
821 nla_nest_cancel(skb
, vf_ports
);
825 static int rtnl_port_self_fill(struct sk_buff
*skb
, struct net_device
*dev
)
827 struct nlattr
*port_self
;
830 port_self
= nla_nest_start(skb
, IFLA_PORT_SELF
);
834 err
= dev
->netdev_ops
->ndo_get_vf_port(dev
, PORT_SELF_VF
, skb
);
836 nla_nest_cancel(skb
, port_self
);
837 return (err
== -EMSGSIZE
) ? err
: 0;
840 nla_nest_end(skb
, port_self
);
845 static int rtnl_port_fill(struct sk_buff
*skb
, struct net_device
*dev
)
849 if (!dev
->netdev_ops
->ndo_get_vf_port
|| !dev
->dev
.parent
)
852 err
= rtnl_port_self_fill(skb
, dev
);
856 if (dev_num_vf(dev
->dev
.parent
)) {
857 err
= rtnl_vf_ports_fill(skb
, dev
);
865 static int rtnl_fill_ifinfo(struct sk_buff
*skb
, struct net_device
*dev
,
866 int type
, u32 pid
, u32 seq
, u32 change
,
869 struct ifinfomsg
*ifm
;
870 struct nlmsghdr
*nlh
;
871 struct rtnl_link_stats64 temp
;
872 const struct rtnl_link_stats64
*stats
;
873 struct nlattr
*attr
, *af_spec
;
874 struct rtnl_af_ops
*af_ops
;
877 nlh
= nlmsg_put(skb
, pid
, seq
, type
, sizeof(*ifm
), flags
);
881 ifm
= nlmsg_data(nlh
);
882 ifm
->ifi_family
= AF_UNSPEC
;
884 ifm
->ifi_type
= dev
->type
;
885 ifm
->ifi_index
= dev
->ifindex
;
886 ifm
->ifi_flags
= dev_get_flags(dev
);
887 ifm
->ifi_change
= change
;
889 NLA_PUT_STRING(skb
, IFLA_IFNAME
, dev
->name
);
890 NLA_PUT_U32(skb
, IFLA_TXQLEN
, dev
->tx_queue_len
);
891 NLA_PUT_U8(skb
, IFLA_OPERSTATE
,
892 netif_running(dev
) ? dev
->operstate
: IF_OPER_DOWN
);
893 NLA_PUT_U8(skb
, IFLA_LINKMODE
, dev
->link_mode
);
894 NLA_PUT_U32(skb
, IFLA_MTU
, dev
->mtu
);
895 NLA_PUT_U32(skb
, IFLA_GROUP
, dev
->group
);
897 if (dev
->ifindex
!= dev
->iflink
)
898 NLA_PUT_U32(skb
, IFLA_LINK
, dev
->iflink
);
901 NLA_PUT_U32(skb
, IFLA_MASTER
, dev
->master
->ifindex
);
904 NLA_PUT_STRING(skb
, IFLA_QDISC
, dev
->qdisc
->ops
->id
);
907 NLA_PUT_STRING(skb
, IFLA_IFALIAS
, dev
->ifalias
);
910 struct rtnl_link_ifmap map
= {
911 .mem_start
= dev
->mem_start
,
912 .mem_end
= dev
->mem_end
,
913 .base_addr
= dev
->base_addr
,
916 .port
= dev
->if_port
,
918 NLA_PUT(skb
, IFLA_MAP
, sizeof(map
), &map
);
922 NLA_PUT(skb
, IFLA_ADDRESS
, dev
->addr_len
, dev
->dev_addr
);
923 NLA_PUT(skb
, IFLA_BROADCAST
, dev
->addr_len
, dev
->broadcast
);
926 attr
= nla_reserve(skb
, IFLA_STATS
,
927 sizeof(struct rtnl_link_stats
));
929 goto nla_put_failure
;
931 stats
= dev_get_stats(dev
, &temp
);
932 copy_rtnl_link_stats(nla_data(attr
), stats
);
934 attr
= nla_reserve(skb
, IFLA_STATS64
,
935 sizeof(struct rtnl_link_stats64
));
937 goto nla_put_failure
;
938 copy_rtnl_link_stats64(nla_data(attr
), stats
);
941 NLA_PUT_U32(skb
, IFLA_NUM_VF
, dev_num_vf(dev
->dev
.parent
));
943 if (dev
->netdev_ops
->ndo_get_vf_config
&& dev
->dev
.parent
) {
946 struct nlattr
*vfinfo
, *vf
;
947 int num_vfs
= dev_num_vf(dev
->dev
.parent
);
949 vfinfo
= nla_nest_start(skb
, IFLA_VFINFO_LIST
);
951 goto nla_put_failure
;
952 for (i
= 0; i
< num_vfs
; i
++) {
953 struct ifla_vf_info ivi
;
954 struct ifla_vf_mac vf_mac
;
955 struct ifla_vf_vlan vf_vlan
;
956 struct ifla_vf_tx_rate vf_tx_rate
;
957 if (dev
->netdev_ops
->ndo_get_vf_config(dev
, i
, &ivi
))
959 vf_mac
.vf
= vf_vlan
.vf
= vf_tx_rate
.vf
= ivi
.vf
;
960 memcpy(vf_mac
.mac
, ivi
.mac
, sizeof(ivi
.mac
));
961 vf_vlan
.vlan
= ivi
.vlan
;
962 vf_vlan
.qos
= ivi
.qos
;
963 vf_tx_rate
.rate
= ivi
.tx_rate
;
964 vf
= nla_nest_start(skb
, IFLA_VF_INFO
);
966 nla_nest_cancel(skb
, vfinfo
);
967 goto nla_put_failure
;
969 NLA_PUT(skb
, IFLA_VF_MAC
, sizeof(vf_mac
), &vf_mac
);
970 NLA_PUT(skb
, IFLA_VF_VLAN
, sizeof(vf_vlan
), &vf_vlan
);
971 NLA_PUT(skb
, IFLA_VF_TX_RATE
, sizeof(vf_tx_rate
), &vf_tx_rate
);
972 nla_nest_end(skb
, vf
);
974 nla_nest_end(skb
, vfinfo
);
977 if (rtnl_port_fill(skb
, dev
))
978 goto nla_put_failure
;
980 if (dev
->rtnl_link_ops
) {
981 if (rtnl_link_fill(skb
, dev
) < 0)
982 goto nla_put_failure
;
985 if (!(af_spec
= nla_nest_start(skb
, IFLA_AF_SPEC
)))
986 goto nla_put_failure
;
988 list_for_each_entry(af_ops
, &rtnl_af_ops
, list
) {
989 if (af_ops
->fill_link_af
) {
993 if (!(af
= nla_nest_start(skb
, af_ops
->family
)))
994 goto nla_put_failure
;
996 err
= af_ops
->fill_link_af(skb
, dev
);
999 * Caller may return ENODATA to indicate that there
1000 * was no data to be dumped. This is not an error, it
1001 * means we should trim the attribute header and
1004 if (err
== -ENODATA
)
1005 nla_nest_cancel(skb
, af
);
1007 goto nla_put_failure
;
1009 nla_nest_end(skb
, af
);
1013 nla_nest_end(skb
, af_spec
);
1015 return nlmsg_end(skb
, nlh
);
1018 nlmsg_cancel(skb
, nlh
);
1022 static int rtnl_dump_ifinfo(struct sk_buff
*skb
, struct netlink_callback
*cb
)
1024 struct net
*net
= sock_net(skb
->sk
);
1027 struct net_device
*dev
;
1028 struct hlist_head
*head
;
1029 struct hlist_node
*node
;
1032 s_idx
= cb
->args
[1];
1035 cb
->seq
= net
->dev_base_seq
;
1037 for (h
= s_h
; h
< NETDEV_HASHENTRIES
; h
++, s_idx
= 0) {
1039 head
= &net
->dev_index_head
[h
];
1040 hlist_for_each_entry_rcu(dev
, node
, head
, index_hlist
) {
1043 if (rtnl_fill_ifinfo(skb
, dev
, RTM_NEWLINK
,
1044 NETLINK_CB(cb
->skb
).pid
,
1045 cb
->nlh
->nlmsg_seq
, 0,
1049 nl_dump_check_consistent(cb
, nlmsg_hdr(skb
));
1062 const struct nla_policy ifla_policy
[IFLA_MAX
+1] = {
1063 [IFLA_IFNAME
] = { .type
= NLA_STRING
, .len
= IFNAMSIZ
-1 },
1064 [IFLA_ADDRESS
] = { .type
= NLA_BINARY
, .len
= MAX_ADDR_LEN
},
1065 [IFLA_BROADCAST
] = { .type
= NLA_BINARY
, .len
= MAX_ADDR_LEN
},
1066 [IFLA_MAP
] = { .len
= sizeof(struct rtnl_link_ifmap
) },
1067 [IFLA_MTU
] = { .type
= NLA_U32
},
1068 [IFLA_LINK
] = { .type
= NLA_U32
},
1069 [IFLA_MASTER
] = { .type
= NLA_U32
},
1070 [IFLA_TXQLEN
] = { .type
= NLA_U32
},
1071 [IFLA_WEIGHT
] = { .type
= NLA_U32
},
1072 [IFLA_OPERSTATE
] = { .type
= NLA_U8
},
1073 [IFLA_LINKMODE
] = { .type
= NLA_U8
},
1074 [IFLA_LINKINFO
] = { .type
= NLA_NESTED
},
1075 [IFLA_NET_NS_PID
] = { .type
= NLA_U32
},
1076 [IFLA_NET_NS_FD
] = { .type
= NLA_U32
},
1077 [IFLA_IFALIAS
] = { .type
= NLA_STRING
, .len
= IFALIASZ
-1 },
1078 [IFLA_VFINFO_LIST
] = {. type
= NLA_NESTED
},
1079 [IFLA_VF_PORTS
] = { .type
= NLA_NESTED
},
1080 [IFLA_PORT_SELF
] = { .type
= NLA_NESTED
},
1081 [IFLA_AF_SPEC
] = { .type
= NLA_NESTED
},
1083 EXPORT_SYMBOL(ifla_policy
);
1085 static const struct nla_policy ifla_info_policy
[IFLA_INFO_MAX
+1] = {
1086 [IFLA_INFO_KIND
] = { .type
= NLA_STRING
},
1087 [IFLA_INFO_DATA
] = { .type
= NLA_NESTED
},
1090 static const struct nla_policy ifla_vfinfo_policy
[IFLA_VF_INFO_MAX
+1] = {
1091 [IFLA_VF_INFO
] = { .type
= NLA_NESTED
},
1094 static const struct nla_policy ifla_vf_policy
[IFLA_VF_MAX
+1] = {
1095 [IFLA_VF_MAC
] = { .type
= NLA_BINARY
,
1096 .len
= sizeof(struct ifla_vf_mac
) },
1097 [IFLA_VF_VLAN
] = { .type
= NLA_BINARY
,
1098 .len
= sizeof(struct ifla_vf_vlan
) },
1099 [IFLA_VF_TX_RATE
] = { .type
= NLA_BINARY
,
1100 .len
= sizeof(struct ifla_vf_tx_rate
) },
1103 static const struct nla_policy ifla_port_policy
[IFLA_PORT_MAX
+1] = {
1104 [IFLA_PORT_VF
] = { .type
= NLA_U32
},
1105 [IFLA_PORT_PROFILE
] = { .type
= NLA_STRING
,
1106 .len
= PORT_PROFILE_MAX
},
1107 [IFLA_PORT_VSI_TYPE
] = { .type
= NLA_BINARY
,
1108 .len
= sizeof(struct ifla_port_vsi
)},
1109 [IFLA_PORT_INSTANCE_UUID
] = { .type
= NLA_BINARY
,
1110 .len
= PORT_UUID_MAX
},
1111 [IFLA_PORT_HOST_UUID
] = { .type
= NLA_STRING
,
1112 .len
= PORT_UUID_MAX
},
1113 [IFLA_PORT_REQUEST
] = { .type
= NLA_U8
, },
1114 [IFLA_PORT_RESPONSE
] = { .type
= NLA_U16
, },
1117 struct net
*rtnl_link_get_net(struct net
*src_net
, struct nlattr
*tb
[])
1120 /* Examine the link attributes and figure out which
1121 * network namespace we are talking about.
1123 if (tb
[IFLA_NET_NS_PID
])
1124 net
= get_net_ns_by_pid(nla_get_u32(tb
[IFLA_NET_NS_PID
]));
1125 else if (tb
[IFLA_NET_NS_FD
])
1126 net
= get_net_ns_by_fd(nla_get_u32(tb
[IFLA_NET_NS_FD
]));
1128 net
= get_net(src_net
);
1131 EXPORT_SYMBOL(rtnl_link_get_net
);
1133 static int validate_linkmsg(struct net_device
*dev
, struct nlattr
*tb
[])
1136 if (tb
[IFLA_ADDRESS
] &&
1137 nla_len(tb
[IFLA_ADDRESS
]) < dev
->addr_len
)
1140 if (tb
[IFLA_BROADCAST
] &&
1141 nla_len(tb
[IFLA_BROADCAST
]) < dev
->addr_len
)
1145 if (tb
[IFLA_AF_SPEC
]) {
1149 nla_for_each_nested(af
, tb
[IFLA_AF_SPEC
], rem
) {
1150 const struct rtnl_af_ops
*af_ops
;
1152 if (!(af_ops
= rtnl_af_lookup(nla_type(af
))))
1153 return -EAFNOSUPPORT
;
1155 if (!af_ops
->set_link_af
)
1158 if (af_ops
->validate_link_af
) {
1159 err
= af_ops
->validate_link_af(dev
, af
);
1169 static int do_setvfinfo(struct net_device
*dev
, struct nlattr
*attr
)
1171 int rem
, err
= -EINVAL
;
1173 const struct net_device_ops
*ops
= dev
->netdev_ops
;
1175 nla_for_each_nested(vf
, attr
, rem
) {
1176 switch (nla_type(vf
)) {
1178 struct ifla_vf_mac
*ivm
;
1181 if (ops
->ndo_set_vf_mac
)
1182 err
= ops
->ndo_set_vf_mac(dev
, ivm
->vf
,
1186 case IFLA_VF_VLAN
: {
1187 struct ifla_vf_vlan
*ivv
;
1190 if (ops
->ndo_set_vf_vlan
)
1191 err
= ops
->ndo_set_vf_vlan(dev
, ivv
->vf
,
1196 case IFLA_VF_TX_RATE
: {
1197 struct ifla_vf_tx_rate
*ivt
;
1200 if (ops
->ndo_set_vf_tx_rate
)
1201 err
= ops
->ndo_set_vf_tx_rate(dev
, ivt
->vf
,
1215 static int do_set_master(struct net_device
*dev
, int ifindex
)
1217 struct net_device
*master_dev
;
1218 const struct net_device_ops
*ops
;
1222 if (dev
->master
->ifindex
== ifindex
)
1224 ops
= dev
->master
->netdev_ops
;
1225 if (ops
->ndo_del_slave
) {
1226 err
= ops
->ndo_del_slave(dev
->master
, dev
);
1235 master_dev
= __dev_get_by_index(dev_net(dev
), ifindex
);
1238 ops
= master_dev
->netdev_ops
;
1239 if (ops
->ndo_add_slave
) {
1240 err
= ops
->ndo_add_slave(master_dev
, dev
);
1250 static int do_setlink(struct net_device
*dev
, struct ifinfomsg
*ifm
,
1251 struct nlattr
**tb
, char *ifname
, int modified
)
1253 const struct net_device_ops
*ops
= dev
->netdev_ops
;
1254 int send_addr_notify
= 0;
1257 if (tb
[IFLA_NET_NS_PID
] || tb
[IFLA_NET_NS_FD
]) {
1258 struct net
*net
= rtnl_link_get_net(dev_net(dev
), tb
);
1263 err
= dev_change_net_namespace(dev
, net
, ifname
);
1271 struct rtnl_link_ifmap
*u_map
;
1274 if (!ops
->ndo_set_config
) {
1279 if (!netif_device_present(dev
)) {
1284 u_map
= nla_data(tb
[IFLA_MAP
]);
1285 k_map
.mem_start
= (unsigned long) u_map
->mem_start
;
1286 k_map
.mem_end
= (unsigned long) u_map
->mem_end
;
1287 k_map
.base_addr
= (unsigned short) u_map
->base_addr
;
1288 k_map
.irq
= (unsigned char) u_map
->irq
;
1289 k_map
.dma
= (unsigned char) u_map
->dma
;
1290 k_map
.port
= (unsigned char) u_map
->port
;
1292 err
= ops
->ndo_set_config(dev
, &k_map
);
1299 if (tb
[IFLA_ADDRESS
]) {
1300 struct sockaddr
*sa
;
1303 if (!ops
->ndo_set_mac_address
) {
1308 if (!netif_device_present(dev
)) {
1313 len
= sizeof(sa_family_t
) + dev
->addr_len
;
1314 sa
= kmalloc(len
, GFP_KERNEL
);
1319 sa
->sa_family
= dev
->type
;
1320 memcpy(sa
->sa_data
, nla_data(tb
[IFLA_ADDRESS
]),
1322 err
= ops
->ndo_set_mac_address(dev
, sa
);
1326 send_addr_notify
= 1;
1331 err
= dev_set_mtu(dev
, nla_get_u32(tb
[IFLA_MTU
]));
1337 if (tb
[IFLA_GROUP
]) {
1338 dev_set_group(dev
, nla_get_u32(tb
[IFLA_GROUP
]));
1343 * Interface selected by interface index but interface
1344 * name provided implies that a name change has been
1347 if (ifm
->ifi_index
> 0 && ifname
[0]) {
1348 err
= dev_change_name(dev
, ifname
);
1354 if (tb
[IFLA_IFALIAS
]) {
1355 err
= dev_set_alias(dev
, nla_data(tb
[IFLA_IFALIAS
]),
1356 nla_len(tb
[IFLA_IFALIAS
]));
1362 if (tb
[IFLA_BROADCAST
]) {
1363 nla_memcpy(dev
->broadcast
, tb
[IFLA_BROADCAST
], dev
->addr_len
);
1364 send_addr_notify
= 1;
1367 if (ifm
->ifi_flags
|| ifm
->ifi_change
) {
1368 err
= dev_change_flags(dev
, rtnl_dev_combine_flags(dev
, ifm
));
1373 if (tb
[IFLA_MASTER
]) {
1374 err
= do_set_master(dev
, nla_get_u32(tb
[IFLA_MASTER
]));
1380 if (tb
[IFLA_TXQLEN
])
1381 dev
->tx_queue_len
= nla_get_u32(tb
[IFLA_TXQLEN
]);
1383 if (tb
[IFLA_OPERSTATE
])
1384 set_operstate(dev
, nla_get_u8(tb
[IFLA_OPERSTATE
]));
1386 if (tb
[IFLA_LINKMODE
]) {
1387 write_lock_bh(&dev_base_lock
);
1388 dev
->link_mode
= nla_get_u8(tb
[IFLA_LINKMODE
]);
1389 write_unlock_bh(&dev_base_lock
);
1392 if (tb
[IFLA_VFINFO_LIST
]) {
1393 struct nlattr
*attr
;
1395 nla_for_each_nested(attr
, tb
[IFLA_VFINFO_LIST
], rem
) {
1396 if (nla_type(attr
) != IFLA_VF_INFO
) {
1400 err
= do_setvfinfo(dev
, attr
);
1408 if (tb
[IFLA_VF_PORTS
]) {
1409 struct nlattr
*port
[IFLA_PORT_MAX
+1];
1410 struct nlattr
*attr
;
1415 if (!ops
->ndo_set_vf_port
)
1418 nla_for_each_nested(attr
, tb
[IFLA_VF_PORTS
], rem
) {
1419 if (nla_type(attr
) != IFLA_VF_PORT
)
1421 err
= nla_parse_nested(port
, IFLA_PORT_MAX
,
1422 attr
, ifla_port_policy
);
1425 if (!port
[IFLA_PORT_VF
]) {
1429 vf
= nla_get_u32(port
[IFLA_PORT_VF
]);
1430 err
= ops
->ndo_set_vf_port(dev
, vf
, port
);
1438 if (tb
[IFLA_PORT_SELF
]) {
1439 struct nlattr
*port
[IFLA_PORT_MAX
+1];
1441 err
= nla_parse_nested(port
, IFLA_PORT_MAX
,
1442 tb
[IFLA_PORT_SELF
], ifla_port_policy
);
1447 if (ops
->ndo_set_vf_port
)
1448 err
= ops
->ndo_set_vf_port(dev
, PORT_SELF_VF
, port
);
1454 if (tb
[IFLA_AF_SPEC
]) {
1458 nla_for_each_nested(af
, tb
[IFLA_AF_SPEC
], rem
) {
1459 const struct rtnl_af_ops
*af_ops
;
1461 if (!(af_ops
= rtnl_af_lookup(nla_type(af
))))
1464 err
= af_ops
->set_link_af(dev
, af
);
1474 if (err
< 0 && modified
&& net_ratelimit())
1475 printk(KERN_WARNING
"A link change request failed with "
1476 "some changes committed already. Interface %s may "
1477 "have been left with an inconsistent configuration, "
1478 "please check.\n", dev
->name
);
1480 if (send_addr_notify
)
1481 call_netdevice_notifiers(NETDEV_CHANGEADDR
, dev
);
1485 static int rtnl_setlink(struct sk_buff
*skb
, struct nlmsghdr
*nlh
, void *arg
)
1487 struct net
*net
= sock_net(skb
->sk
);
1488 struct ifinfomsg
*ifm
;
1489 struct net_device
*dev
;
1491 struct nlattr
*tb
[IFLA_MAX
+1];
1492 char ifname
[IFNAMSIZ
];
1494 err
= nlmsg_parse(nlh
, sizeof(*ifm
), tb
, IFLA_MAX
, ifla_policy
);
1498 if (tb
[IFLA_IFNAME
])
1499 nla_strlcpy(ifname
, tb
[IFLA_IFNAME
], IFNAMSIZ
);
1504 ifm
= nlmsg_data(nlh
);
1505 if (ifm
->ifi_index
> 0)
1506 dev
= __dev_get_by_index(net
, ifm
->ifi_index
);
1507 else if (tb
[IFLA_IFNAME
])
1508 dev
= __dev_get_by_name(net
, ifname
);
1517 err
= validate_linkmsg(dev
, tb
);
1521 err
= do_setlink(dev
, ifm
, tb
, ifname
, 0);
1526 static int rtnl_dellink(struct sk_buff
*skb
, struct nlmsghdr
*nlh
, void *arg
)
1528 struct net
*net
= sock_net(skb
->sk
);
1529 const struct rtnl_link_ops
*ops
;
1530 struct net_device
*dev
;
1531 struct ifinfomsg
*ifm
;
1532 char ifname
[IFNAMSIZ
];
1533 struct nlattr
*tb
[IFLA_MAX
+1];
1535 LIST_HEAD(list_kill
);
1537 err
= nlmsg_parse(nlh
, sizeof(*ifm
), tb
, IFLA_MAX
, ifla_policy
);
1541 if (tb
[IFLA_IFNAME
])
1542 nla_strlcpy(ifname
, tb
[IFLA_IFNAME
], IFNAMSIZ
);
1544 ifm
= nlmsg_data(nlh
);
1545 if (ifm
->ifi_index
> 0)
1546 dev
= __dev_get_by_index(net
, ifm
->ifi_index
);
1547 else if (tb
[IFLA_IFNAME
])
1548 dev
= __dev_get_by_name(net
, ifname
);
1555 ops
= dev
->rtnl_link_ops
;
1559 ops
->dellink(dev
, &list_kill
);
1560 unregister_netdevice_many(&list_kill
);
1561 list_del(&list_kill
);
1565 int rtnl_configure_link(struct net_device
*dev
, const struct ifinfomsg
*ifm
)
1567 unsigned int old_flags
;
1570 old_flags
= dev
->flags
;
1571 if (ifm
&& (ifm
->ifi_flags
|| ifm
->ifi_change
)) {
1572 err
= __dev_change_flags(dev
, rtnl_dev_combine_flags(dev
, ifm
));
1577 dev
->rtnl_link_state
= RTNL_LINK_INITIALIZED
;
1578 rtmsg_ifinfo(RTM_NEWLINK
, dev
, ~0U);
1580 __dev_notify_flags(dev
, old_flags
);
1583 EXPORT_SYMBOL(rtnl_configure_link
);
1585 struct net_device
*rtnl_create_link(struct net
*src_net
, struct net
*net
,
1586 char *ifname
, const struct rtnl_link_ops
*ops
, struct nlattr
*tb
[])
1589 struct net_device
*dev
;
1590 unsigned int num_queues
= 1;
1591 unsigned int real_num_queues
= 1;
1593 if (ops
->get_tx_queues
) {
1594 err
= ops
->get_tx_queues(src_net
, tb
, &num_queues
,
1600 dev
= alloc_netdev_mq(ops
->priv_size
, ifname
, ops
->setup
, num_queues
);
1604 dev_net_set(dev
, net
);
1605 dev
->rtnl_link_ops
= ops
;
1606 dev
->rtnl_link_state
= RTNL_LINK_INITIALIZING
;
1607 dev
->real_num_tx_queues
= real_num_queues
;
1610 dev
->mtu
= nla_get_u32(tb
[IFLA_MTU
]);
1611 if (tb
[IFLA_ADDRESS
])
1612 memcpy(dev
->dev_addr
, nla_data(tb
[IFLA_ADDRESS
]),
1613 nla_len(tb
[IFLA_ADDRESS
]));
1614 if (tb
[IFLA_BROADCAST
])
1615 memcpy(dev
->broadcast
, nla_data(tb
[IFLA_BROADCAST
]),
1616 nla_len(tb
[IFLA_BROADCAST
]));
1617 if (tb
[IFLA_TXQLEN
])
1618 dev
->tx_queue_len
= nla_get_u32(tb
[IFLA_TXQLEN
]);
1619 if (tb
[IFLA_OPERSTATE
])
1620 set_operstate(dev
, nla_get_u8(tb
[IFLA_OPERSTATE
]));
1621 if (tb
[IFLA_LINKMODE
])
1622 dev
->link_mode
= nla_get_u8(tb
[IFLA_LINKMODE
]);
1624 dev_set_group(dev
, nla_get_u32(tb
[IFLA_GROUP
]));
1629 return ERR_PTR(err
);
1631 EXPORT_SYMBOL(rtnl_create_link
);
1633 static int rtnl_group_changelink(struct net
*net
, int group
,
1634 struct ifinfomsg
*ifm
,
1637 struct net_device
*dev
;
1640 for_each_netdev(net
, dev
) {
1641 if (dev
->group
== group
) {
1642 err
= do_setlink(dev
, ifm
, tb
, NULL
, 0);
1651 static int rtnl_newlink(struct sk_buff
*skb
, struct nlmsghdr
*nlh
, void *arg
)
1653 struct net
*net
= sock_net(skb
->sk
);
1654 const struct rtnl_link_ops
*ops
;
1655 struct net_device
*dev
;
1656 struct ifinfomsg
*ifm
;
1657 char kind
[MODULE_NAME_LEN
];
1658 char ifname
[IFNAMSIZ
];
1659 struct nlattr
*tb
[IFLA_MAX
+1];
1660 struct nlattr
*linkinfo
[IFLA_INFO_MAX
+1];
1663 #ifdef CONFIG_MODULES
1666 err
= nlmsg_parse(nlh
, sizeof(*ifm
), tb
, IFLA_MAX
, ifla_policy
);
1670 if (tb
[IFLA_IFNAME
])
1671 nla_strlcpy(ifname
, tb
[IFLA_IFNAME
], IFNAMSIZ
);
1675 ifm
= nlmsg_data(nlh
);
1676 if (ifm
->ifi_index
> 0)
1677 dev
= __dev_get_by_index(net
, ifm
->ifi_index
);
1680 dev
= __dev_get_by_name(net
, ifname
);
1685 err
= validate_linkmsg(dev
, tb
);
1689 if (tb
[IFLA_LINKINFO
]) {
1690 err
= nla_parse_nested(linkinfo
, IFLA_INFO_MAX
,
1691 tb
[IFLA_LINKINFO
], ifla_info_policy
);
1695 memset(linkinfo
, 0, sizeof(linkinfo
));
1697 if (linkinfo
[IFLA_INFO_KIND
]) {
1698 nla_strlcpy(kind
, linkinfo
[IFLA_INFO_KIND
], sizeof(kind
));
1699 ops
= rtnl_link_ops_get(kind
);
1706 struct nlattr
*attr
[ops
? ops
->maxtype
+ 1 : 0], **data
= NULL
;
1707 struct net
*dest_net
;
1710 if (ops
->maxtype
&& linkinfo
[IFLA_INFO_DATA
]) {
1711 err
= nla_parse_nested(attr
, ops
->maxtype
,
1712 linkinfo
[IFLA_INFO_DATA
],
1718 if (ops
->validate
) {
1719 err
= ops
->validate(tb
, data
);
1728 if (nlh
->nlmsg_flags
& NLM_F_EXCL
)
1730 if (nlh
->nlmsg_flags
& NLM_F_REPLACE
)
1733 if (linkinfo
[IFLA_INFO_DATA
]) {
1734 if (!ops
|| ops
!= dev
->rtnl_link_ops
||
1738 err
= ops
->changelink(dev
, tb
, data
);
1744 return do_setlink(dev
, ifm
, tb
, ifname
, modified
);
1747 if (!(nlh
->nlmsg_flags
& NLM_F_CREATE
)) {
1748 if (ifm
->ifi_index
== 0 && tb
[IFLA_GROUP
])
1749 return rtnl_group_changelink(net
,
1750 nla_get_u32(tb
[IFLA_GROUP
]),
1757 if (tb
[IFLA_MAP
] || tb
[IFLA_MASTER
] || tb
[IFLA_PROTINFO
])
1761 #ifdef CONFIG_MODULES
1764 request_module("rtnl-link-%s", kind
);
1766 ops
= rtnl_link_ops_get(kind
);
1775 snprintf(ifname
, IFNAMSIZ
, "%s%%d", ops
->kind
);
1777 dest_net
= rtnl_link_get_net(net
, tb
);
1778 if (IS_ERR(dest_net
))
1779 return PTR_ERR(dest_net
);
1781 dev
= rtnl_create_link(net
, dest_net
, ifname
, ops
, tb
);
1785 else if (ops
->newlink
)
1786 err
= ops
->newlink(net
, dev
, tb
, data
);
1788 err
= register_netdevice(dev
);
1790 if (err
< 0 && !IS_ERR(dev
))
1795 err
= rtnl_configure_link(dev
, ifm
);
1797 unregister_netdevice(dev
);
1804 static int rtnl_getlink(struct sk_buff
*skb
, struct nlmsghdr
* nlh
, void *arg
)
1806 struct net
*net
= sock_net(skb
->sk
);
1807 struct ifinfomsg
*ifm
;
1808 char ifname
[IFNAMSIZ
];
1809 struct nlattr
*tb
[IFLA_MAX
+1];
1810 struct net_device
*dev
= NULL
;
1811 struct sk_buff
*nskb
;
1814 err
= nlmsg_parse(nlh
, sizeof(*ifm
), tb
, IFLA_MAX
, ifla_policy
);
1818 if (tb
[IFLA_IFNAME
])
1819 nla_strlcpy(ifname
, tb
[IFLA_IFNAME
], IFNAMSIZ
);
1821 ifm
= nlmsg_data(nlh
);
1822 if (ifm
->ifi_index
> 0)
1823 dev
= __dev_get_by_index(net
, ifm
->ifi_index
);
1824 else if (tb
[IFLA_IFNAME
])
1825 dev
= __dev_get_by_name(net
, ifname
);
1832 nskb
= nlmsg_new(if_nlmsg_size(dev
), GFP_KERNEL
);
1836 err
= rtnl_fill_ifinfo(nskb
, dev
, RTM_NEWLINK
, NETLINK_CB(skb
).pid
,
1837 nlh
->nlmsg_seq
, 0, 0);
1839 /* -EMSGSIZE implies BUG in if_nlmsg_size */
1840 WARN_ON(err
== -EMSGSIZE
);
1843 err
= rtnl_unicast(nskb
, net
, NETLINK_CB(skb
).pid
);
1848 static u16
rtnl_calcit(struct sk_buff
*skb
)
1850 return min_ifinfo_dump_size
;
1853 static int rtnl_dump_all(struct sk_buff
*skb
, struct netlink_callback
*cb
)
1856 int s_idx
= cb
->family
;
1860 for (idx
= 1; idx
<= RTNL_FAMILY_MAX
; idx
++) {
1861 int type
= cb
->nlh
->nlmsg_type
-RTM_BASE
;
1862 if (idx
< s_idx
|| idx
== PF_PACKET
)
1864 if (rtnl_msg_handlers
[idx
] == NULL
||
1865 rtnl_msg_handlers
[idx
][type
].dumpit
== NULL
)
1868 memset(&cb
->args
[0], 0, sizeof(cb
->args
));
1869 if (rtnl_msg_handlers
[idx
][type
].dumpit(skb
, cb
))
1877 void rtmsg_ifinfo(int type
, struct net_device
*dev
, unsigned change
)
1879 struct net
*net
= dev_net(dev
);
1880 struct sk_buff
*skb
;
1882 size_t if_info_size
;
1884 skb
= nlmsg_new((if_info_size
= if_nlmsg_size(dev
)), GFP_KERNEL
);
1888 min_ifinfo_dump_size
= max_t(u16
, if_info_size
, min_ifinfo_dump_size
);
1890 err
= rtnl_fill_ifinfo(skb
, dev
, type
, 0, 0, change
, 0);
1892 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
1893 WARN_ON(err
== -EMSGSIZE
);
1897 rtnl_notify(skb
, net
, 0, RTNLGRP_LINK
, NULL
, GFP_KERNEL
);
1901 rtnl_set_sk_err(net
, RTNLGRP_LINK
, err
);
1904 /* Protected by RTNL sempahore. */
1905 static struct rtattr
**rta_buf
;
1906 static int rtattr_max
;
1908 /* Process one rtnetlink message. */
1910 static int rtnetlink_rcv_msg(struct sk_buff
*skb
, struct nlmsghdr
*nlh
)
1912 struct net
*net
= sock_net(skb
->sk
);
1913 rtnl_doit_func doit
;
1920 type
= nlh
->nlmsg_type
;
1926 /* All the messages must have at least 1 byte length */
1927 if (nlh
->nlmsg_len
< NLMSG_LENGTH(sizeof(struct rtgenmsg
)))
1930 family
= ((struct rtgenmsg
*)NLMSG_DATA(nlh
))->rtgen_family
;
1934 if (kind
!= 2 && security_netlink_recv(skb
, CAP_NET_ADMIN
))
1937 if (kind
== 2 && nlh
->nlmsg_flags
&NLM_F_DUMP
) {
1939 rtnl_dumpit_func dumpit
;
1940 rtnl_calcit_func calcit
;
1941 u16 min_dump_alloc
= 0;
1943 dumpit
= rtnl_get_dumpit(family
, type
);
1946 calcit
= rtnl_get_calcit(family
, type
);
1948 min_dump_alloc
= calcit(skb
);
1952 err
= netlink_dump_start(rtnl
, skb
, nlh
, dumpit
,
1953 NULL
, min_dump_alloc
);
1958 memset(rta_buf
, 0, (rtattr_max
* sizeof(struct rtattr
*)));
1960 min_len
= rtm_min
[sz_idx
];
1961 if (nlh
->nlmsg_len
< min_len
)
1964 if (nlh
->nlmsg_len
> min_len
) {
1965 int attrlen
= nlh
->nlmsg_len
- NLMSG_ALIGN(min_len
);
1966 struct rtattr
*attr
= (void *)nlh
+ NLMSG_ALIGN(min_len
);
1968 while (RTA_OK(attr
, attrlen
)) {
1969 unsigned flavor
= attr
->rta_type
;
1971 if (flavor
> rta_max
[sz_idx
])
1973 rta_buf
[flavor
-1] = attr
;
1975 attr
= RTA_NEXT(attr
, attrlen
);
1979 doit
= rtnl_get_doit(family
, type
);
1983 return doit(skb
, nlh
, (void *)&rta_buf
[0]);
1986 static void rtnetlink_rcv(struct sk_buff
*skb
)
1989 netlink_rcv_skb(skb
, &rtnetlink_rcv_msg
);
1993 static int rtnetlink_event(struct notifier_block
*this, unsigned long event
, void *ptr
)
1995 struct net_device
*dev
= ptr
;
2001 case NETDEV_POST_INIT
:
2002 case NETDEV_REGISTER
:
2004 case NETDEV_PRE_TYPE_CHANGE
:
2005 case NETDEV_GOING_DOWN
:
2006 case NETDEV_UNREGISTER
:
2007 case NETDEV_UNREGISTER_BATCH
:
2008 case NETDEV_RELEASE
:
2012 rtmsg_ifinfo(RTM_NEWLINK
, dev
, 0);
2018 static struct notifier_block rtnetlink_dev_notifier
= {
2019 .notifier_call
= rtnetlink_event
,
2023 static int __net_init
rtnetlink_net_init(struct net
*net
)
2026 sk
= netlink_kernel_create(net
, NETLINK_ROUTE
, RTNLGRP_MAX
,
2027 rtnetlink_rcv
, &rtnl_mutex
, THIS_MODULE
);
2034 static void __net_exit
rtnetlink_net_exit(struct net
*net
)
2036 netlink_kernel_release(net
->rtnl
);
2040 static struct pernet_operations rtnetlink_net_ops
= {
2041 .init
= rtnetlink_net_init
,
2042 .exit
= rtnetlink_net_exit
,
2045 void __init
rtnetlink_init(void)
2050 for (i
= 0; i
< ARRAY_SIZE(rta_max
); i
++)
2051 if (rta_max
[i
] > rtattr_max
)
2052 rtattr_max
= rta_max
[i
];
2053 rta_buf
= kmalloc(rtattr_max
* sizeof(struct rtattr
*), GFP_KERNEL
);
2055 panic("rtnetlink_init: cannot allocate rta_buf\n");
2057 if (register_pernet_subsys(&rtnetlink_net_ops
))
2058 panic("rtnetlink_init: cannot initialize rtnetlink\n");
2060 netlink_set_nonroot(NETLINK_ROUTE
, NL_NONROOT_RECV
);
2061 register_netdevice_notifier(&rtnetlink_dev_notifier
);
2063 rtnl_register(PF_UNSPEC
, RTM_GETLINK
, rtnl_getlink
,
2064 rtnl_dump_ifinfo
, rtnl_calcit
);
2065 rtnl_register(PF_UNSPEC
, RTM_SETLINK
, rtnl_setlink
, NULL
, NULL
);
2066 rtnl_register(PF_UNSPEC
, RTM_NEWLINK
, rtnl_newlink
, NULL
, NULL
);
2067 rtnl_register(PF_UNSPEC
, RTM_DELLINK
, rtnl_dellink
, NULL
, NULL
);
2069 rtnl_register(PF_UNSPEC
, RTM_GETADDR
, NULL
, rtnl_dump_all
, NULL
);
2070 rtnl_register(PF_UNSPEC
, RTM_GETROUTE
, NULL
, rtnl_dump_all
, NULL
);