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/if_bridge.h>
39 #include <linux/pci.h>
40 #include <linux/etherdevice.h>
42 #include <asm/uaccess.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>
55 #include <net/net_namespace.h>
59 rtnl_dumpit_func dumpit
;
60 rtnl_calcit_func calcit
;
63 static DEFINE_MUTEX(rtnl_mutex
);
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
[msgindex
].doit
;
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
[msgindex
].dumpit
;
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
[msgindex
].calcit
;
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
);
276 static const struct rtnl_link_ops
*rtnl_link_ops_get(const char *kind
)
278 const struct rtnl_link_ops
*ops
;
280 list_for_each_entry(ops
, &link_ops
, list
) {
281 if (!strcmp(ops
->kind
, kind
))
288 * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
289 * @ops: struct rtnl_link_ops * to register
291 * The caller must hold the rtnl_mutex. This function should be used
292 * by drivers that create devices during module initialization. It
293 * must be called before registering the devices.
295 * Returns 0 on success or a negative error code.
297 int __rtnl_link_register(struct rtnl_link_ops
*ops
)
299 if (rtnl_link_ops_get(ops
->kind
))
303 ops
->dellink
= unregister_netdevice_queue
;
305 list_add_tail(&ops
->list
, &link_ops
);
308 EXPORT_SYMBOL_GPL(__rtnl_link_register
);
311 * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
312 * @ops: struct rtnl_link_ops * to register
314 * Returns 0 on success or a negative error code.
316 int rtnl_link_register(struct rtnl_link_ops
*ops
)
321 err
= __rtnl_link_register(ops
);
325 EXPORT_SYMBOL_GPL(rtnl_link_register
);
327 static void __rtnl_kill_links(struct net
*net
, struct rtnl_link_ops
*ops
)
329 struct net_device
*dev
;
330 LIST_HEAD(list_kill
);
332 for_each_netdev(net
, dev
) {
333 if (dev
->rtnl_link_ops
== ops
)
334 ops
->dellink(dev
, &list_kill
);
336 unregister_netdevice_many(&list_kill
);
340 * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
341 * @ops: struct rtnl_link_ops * to unregister
343 * The caller must hold the rtnl_mutex.
345 void __rtnl_link_unregister(struct rtnl_link_ops
*ops
)
350 __rtnl_kill_links(net
, ops
);
352 list_del(&ops
->list
);
354 EXPORT_SYMBOL_GPL(__rtnl_link_unregister
);
356 /* Return with the rtnl_lock held when there are no network
357 * devices unregistering in any network namespace.
359 static void rtnl_lock_unregistering_all(void)
366 prepare_to_wait(&netdev_unregistering_wq
, &wait
,
367 TASK_UNINTERRUPTIBLE
);
368 unregistering
= false;
371 if (net
->dev_unreg_count
> 0) {
372 unregistering
= true;
381 finish_wait(&netdev_unregistering_wq
, &wait
);
385 * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
386 * @ops: struct rtnl_link_ops * to unregister
388 void rtnl_link_unregister(struct rtnl_link_ops
*ops
)
390 /* Close the race with cleanup_net() */
391 mutex_lock(&net_mutex
);
392 rtnl_lock_unregistering_all();
393 __rtnl_link_unregister(ops
);
395 mutex_unlock(&net_mutex
);
397 EXPORT_SYMBOL_GPL(rtnl_link_unregister
);
399 static size_t rtnl_link_get_slave_info_data_size(const struct net_device
*dev
)
401 struct net_device
*master_dev
;
402 const struct rtnl_link_ops
*ops
;
404 master_dev
= netdev_master_upper_dev_get((struct net_device
*) dev
);
407 ops
= master_dev
->rtnl_link_ops
;
408 if (!ops
|| !ops
->get_slave_size
)
410 /* IFLA_INFO_SLAVE_DATA + nested data */
411 return nla_total_size(sizeof(struct nlattr
)) +
412 ops
->get_slave_size(master_dev
, dev
);
415 static size_t rtnl_link_get_size(const struct net_device
*dev
)
417 const struct rtnl_link_ops
*ops
= dev
->rtnl_link_ops
;
423 size
= nla_total_size(sizeof(struct nlattr
)) + /* IFLA_LINKINFO */
424 nla_total_size(strlen(ops
->kind
) + 1); /* IFLA_INFO_KIND */
427 /* IFLA_INFO_DATA + nested data */
428 size
+= nla_total_size(sizeof(struct nlattr
)) +
431 if (ops
->get_xstats_size
)
432 /* IFLA_INFO_XSTATS */
433 size
+= nla_total_size(ops
->get_xstats_size(dev
));
435 size
+= rtnl_link_get_slave_info_data_size(dev
);
440 static LIST_HEAD(rtnl_af_ops
);
442 static const struct rtnl_af_ops
*rtnl_af_lookup(const int family
)
444 const struct rtnl_af_ops
*ops
;
446 list_for_each_entry(ops
, &rtnl_af_ops
, list
) {
447 if (ops
->family
== family
)
455 * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
456 * @ops: struct rtnl_af_ops * to register
458 * Returns 0 on success or a negative error code.
460 void rtnl_af_register(struct rtnl_af_ops
*ops
)
463 list_add_tail(&ops
->list
, &rtnl_af_ops
);
466 EXPORT_SYMBOL_GPL(rtnl_af_register
);
469 * __rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
470 * @ops: struct rtnl_af_ops * to unregister
472 * The caller must hold the rtnl_mutex.
474 void __rtnl_af_unregister(struct rtnl_af_ops
*ops
)
476 list_del(&ops
->list
);
478 EXPORT_SYMBOL_GPL(__rtnl_af_unregister
);
481 * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
482 * @ops: struct rtnl_af_ops * to unregister
484 void rtnl_af_unregister(struct rtnl_af_ops
*ops
)
487 __rtnl_af_unregister(ops
);
490 EXPORT_SYMBOL_GPL(rtnl_af_unregister
);
492 static size_t rtnl_link_get_af_size(const struct net_device
*dev
)
494 struct rtnl_af_ops
*af_ops
;
498 size
= nla_total_size(sizeof(struct nlattr
));
500 list_for_each_entry(af_ops
, &rtnl_af_ops
, list
) {
501 if (af_ops
->get_link_af_size
) {
502 /* AF_* + nested data */
503 size
+= nla_total_size(sizeof(struct nlattr
)) +
504 af_ops
->get_link_af_size(dev
);
511 static bool rtnl_have_link_slave_info(const struct net_device
*dev
)
513 struct net_device
*master_dev
;
515 master_dev
= netdev_master_upper_dev_get((struct net_device
*) dev
);
516 if (master_dev
&& master_dev
->rtnl_link_ops
)
521 static int rtnl_link_slave_info_fill(struct sk_buff
*skb
,
522 const struct net_device
*dev
)
524 struct net_device
*master_dev
;
525 const struct rtnl_link_ops
*ops
;
526 struct nlattr
*slave_data
;
529 master_dev
= netdev_master_upper_dev_get((struct net_device
*) dev
);
532 ops
= master_dev
->rtnl_link_ops
;
535 if (nla_put_string(skb
, IFLA_INFO_SLAVE_KIND
, ops
->kind
) < 0)
537 if (ops
->fill_slave_info
) {
538 slave_data
= nla_nest_start(skb
, IFLA_INFO_SLAVE_DATA
);
541 err
= ops
->fill_slave_info(skb
, master_dev
, dev
);
543 goto err_cancel_slave_data
;
544 nla_nest_end(skb
, slave_data
);
548 err_cancel_slave_data
:
549 nla_nest_cancel(skb
, slave_data
);
553 static int rtnl_link_info_fill(struct sk_buff
*skb
,
554 const struct net_device
*dev
)
556 const struct rtnl_link_ops
*ops
= dev
->rtnl_link_ops
;
562 if (nla_put_string(skb
, IFLA_INFO_KIND
, ops
->kind
) < 0)
564 if (ops
->fill_xstats
) {
565 err
= ops
->fill_xstats(skb
, dev
);
569 if (ops
->fill_info
) {
570 data
= nla_nest_start(skb
, IFLA_INFO_DATA
);
573 err
= ops
->fill_info(skb
, dev
);
575 goto err_cancel_data
;
576 nla_nest_end(skb
, data
);
581 nla_nest_cancel(skb
, data
);
585 static int rtnl_link_fill(struct sk_buff
*skb
, const struct net_device
*dev
)
587 struct nlattr
*linkinfo
;
590 linkinfo
= nla_nest_start(skb
, IFLA_LINKINFO
);
591 if (linkinfo
== NULL
)
594 err
= rtnl_link_info_fill(skb
, dev
);
596 goto err_cancel_link
;
598 err
= rtnl_link_slave_info_fill(skb
, dev
);
600 goto err_cancel_link
;
602 nla_nest_end(skb
, linkinfo
);
606 nla_nest_cancel(skb
, linkinfo
);
611 int rtnetlink_send(struct sk_buff
*skb
, struct net
*net
, u32 pid
, unsigned int group
, int echo
)
613 struct sock
*rtnl
= net
->rtnl
;
616 NETLINK_CB(skb
).dst_group
= group
;
618 atomic_inc(&skb
->users
);
619 netlink_broadcast(rtnl
, skb
, pid
, group
, GFP_KERNEL
);
621 err
= netlink_unicast(rtnl
, skb
, pid
, MSG_DONTWAIT
);
625 int rtnl_unicast(struct sk_buff
*skb
, struct net
*net
, u32 pid
)
627 struct sock
*rtnl
= net
->rtnl
;
629 return nlmsg_unicast(rtnl
, skb
, pid
);
631 EXPORT_SYMBOL(rtnl_unicast
);
633 void rtnl_notify(struct sk_buff
*skb
, struct net
*net
, u32 pid
, u32 group
,
634 struct nlmsghdr
*nlh
, gfp_t flags
)
636 struct sock
*rtnl
= net
->rtnl
;
640 report
= nlmsg_report(nlh
);
642 nlmsg_notify(rtnl
, skb
, pid
, group
, report
, flags
);
644 EXPORT_SYMBOL(rtnl_notify
);
646 void rtnl_set_sk_err(struct net
*net
, u32 group
, int error
)
648 struct sock
*rtnl
= net
->rtnl
;
650 netlink_set_err(rtnl
, 0, group
, error
);
652 EXPORT_SYMBOL(rtnl_set_sk_err
);
654 int rtnetlink_put_metrics(struct sk_buff
*skb
, u32
*metrics
)
659 mx
= nla_nest_start(skb
, RTA_METRICS
);
663 for (i
= 0; i
< RTAX_MAX
; i
++) {
666 if (nla_put_u32(skb
, i
+1, metrics
[i
]))
667 goto nla_put_failure
;
672 nla_nest_cancel(skb
, mx
);
676 return nla_nest_end(skb
, mx
);
679 nla_nest_cancel(skb
, mx
);
682 EXPORT_SYMBOL(rtnetlink_put_metrics
);
684 int rtnl_put_cacheinfo(struct sk_buff
*skb
, struct dst_entry
*dst
, u32 id
,
685 long expires
, u32 error
)
687 struct rta_cacheinfo ci
= {
688 .rta_lastuse
= jiffies_delta_to_clock_t(jiffies
- dst
->lastuse
),
689 .rta_used
= dst
->__use
,
690 .rta_clntref
= atomic_read(&(dst
->__refcnt
)),
698 clock
= jiffies_to_clock_t(abs(expires
));
699 clock
= min_t(unsigned long, clock
, INT_MAX
);
700 ci
.rta_expires
= (expires
> 0) ? clock
: -clock
;
702 return nla_put(skb
, RTA_CACHEINFO
, sizeof(ci
), &ci
);
704 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo
);
706 static void set_operstate(struct net_device
*dev
, unsigned char transition
)
708 unsigned char operstate
= dev
->operstate
;
710 switch (transition
) {
712 if ((operstate
== IF_OPER_DORMANT
||
713 operstate
== IF_OPER_UNKNOWN
) &&
715 operstate
= IF_OPER_UP
;
718 case IF_OPER_DORMANT
:
719 if (operstate
== IF_OPER_UP
||
720 operstate
== IF_OPER_UNKNOWN
)
721 operstate
= IF_OPER_DORMANT
;
725 if (dev
->operstate
!= operstate
) {
726 write_lock_bh(&dev_base_lock
);
727 dev
->operstate
= operstate
;
728 write_unlock_bh(&dev_base_lock
);
729 netdev_state_change(dev
);
733 static unsigned int rtnl_dev_get_flags(const struct net_device
*dev
)
735 return (dev
->flags
& ~(IFF_PROMISC
| IFF_ALLMULTI
)) |
736 (dev
->gflags
& (IFF_PROMISC
| IFF_ALLMULTI
));
739 static unsigned int rtnl_dev_combine_flags(const struct net_device
*dev
,
740 const struct ifinfomsg
*ifm
)
742 unsigned int flags
= ifm
->ifi_flags
;
744 /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
746 flags
= (flags
& ifm
->ifi_change
) |
747 (rtnl_dev_get_flags(dev
) & ~ifm
->ifi_change
);
752 static void copy_rtnl_link_stats(struct rtnl_link_stats
*a
,
753 const struct rtnl_link_stats64
*b
)
755 a
->rx_packets
= b
->rx_packets
;
756 a
->tx_packets
= b
->tx_packets
;
757 a
->rx_bytes
= b
->rx_bytes
;
758 a
->tx_bytes
= b
->tx_bytes
;
759 a
->rx_errors
= b
->rx_errors
;
760 a
->tx_errors
= b
->tx_errors
;
761 a
->rx_dropped
= b
->rx_dropped
;
762 a
->tx_dropped
= b
->tx_dropped
;
764 a
->multicast
= b
->multicast
;
765 a
->collisions
= b
->collisions
;
767 a
->rx_length_errors
= b
->rx_length_errors
;
768 a
->rx_over_errors
= b
->rx_over_errors
;
769 a
->rx_crc_errors
= b
->rx_crc_errors
;
770 a
->rx_frame_errors
= b
->rx_frame_errors
;
771 a
->rx_fifo_errors
= b
->rx_fifo_errors
;
772 a
->rx_missed_errors
= b
->rx_missed_errors
;
774 a
->tx_aborted_errors
= b
->tx_aborted_errors
;
775 a
->tx_carrier_errors
= b
->tx_carrier_errors
;
776 a
->tx_fifo_errors
= b
->tx_fifo_errors
;
777 a
->tx_heartbeat_errors
= b
->tx_heartbeat_errors
;
778 a
->tx_window_errors
= b
->tx_window_errors
;
780 a
->rx_compressed
= b
->rx_compressed
;
781 a
->tx_compressed
= b
->tx_compressed
;
784 static void copy_rtnl_link_stats64(void *v
, const struct rtnl_link_stats64
*b
)
786 memcpy(v
, b
, sizeof(*b
));
790 static inline int rtnl_vfinfo_size(const struct net_device
*dev
,
793 if (dev
->dev
.parent
&& dev_is_pci(dev
->dev
.parent
) &&
794 (ext_filter_mask
& RTEXT_FILTER_VF
)) {
795 int num_vfs
= dev_num_vf(dev
->dev
.parent
);
796 size_t size
= nla_total_size(sizeof(struct nlattr
));
797 size
+= nla_total_size(num_vfs
* sizeof(struct nlattr
));
799 (nla_total_size(sizeof(struct ifla_vf_mac
)) +
800 nla_total_size(sizeof(struct ifla_vf_vlan
)) +
801 nla_total_size(sizeof(struct ifla_vf_spoofchk
)) +
802 nla_total_size(sizeof(struct ifla_vf_rate
)));
808 static size_t rtnl_port_size(const struct net_device
*dev
,
811 size_t port_size
= nla_total_size(4) /* PORT_VF */
812 + nla_total_size(PORT_PROFILE_MAX
) /* PORT_PROFILE */
813 + nla_total_size(sizeof(struct ifla_port_vsi
))
815 + nla_total_size(PORT_UUID_MAX
) /* PORT_INSTANCE_UUID */
816 + nla_total_size(PORT_UUID_MAX
) /* PORT_HOST_UUID */
817 + nla_total_size(1) /* PROT_VDP_REQUEST */
818 + nla_total_size(2); /* PORT_VDP_RESPONSE */
819 size_t vf_ports_size
= nla_total_size(sizeof(struct nlattr
));
820 size_t vf_port_size
= nla_total_size(sizeof(struct nlattr
))
822 size_t port_self_size
= nla_total_size(sizeof(struct nlattr
))
825 if (!dev
->netdev_ops
->ndo_get_vf_port
|| !dev
->dev
.parent
||
826 !(ext_filter_mask
& RTEXT_FILTER_VF
))
828 if (dev_num_vf(dev
->dev
.parent
))
829 return port_self_size
+ vf_ports_size
+
830 vf_port_size
* dev_num_vf(dev
->dev
.parent
);
832 return port_self_size
;
835 static noinline
size_t if_nlmsg_size(const struct net_device
*dev
,
838 return NLMSG_ALIGN(sizeof(struct ifinfomsg
))
839 + nla_total_size(IFNAMSIZ
) /* IFLA_IFNAME */
840 + nla_total_size(IFALIASZ
) /* IFLA_IFALIAS */
841 + nla_total_size(IFNAMSIZ
) /* IFLA_QDISC */
842 + nla_total_size(sizeof(struct rtnl_link_ifmap
))
843 + nla_total_size(sizeof(struct rtnl_link_stats
))
844 + nla_total_size(sizeof(struct rtnl_link_stats64
))
845 + nla_total_size(MAX_ADDR_LEN
) /* IFLA_ADDRESS */
846 + nla_total_size(MAX_ADDR_LEN
) /* IFLA_BROADCAST */
847 + nla_total_size(4) /* IFLA_TXQLEN */
848 + nla_total_size(4) /* IFLA_WEIGHT */
849 + nla_total_size(4) /* IFLA_MTU */
850 + nla_total_size(4) /* IFLA_LINK */
851 + nla_total_size(4) /* IFLA_MASTER */
852 + nla_total_size(1) /* IFLA_CARRIER */
853 + nla_total_size(4) /* IFLA_PROMISCUITY */
854 + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
855 + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
856 + nla_total_size(1) /* IFLA_OPERSTATE */
857 + nla_total_size(1) /* IFLA_LINKMODE */
858 + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
859 + nla_total_size(ext_filter_mask
860 & RTEXT_FILTER_VF
? 4 : 0) /* IFLA_NUM_VF */
861 + rtnl_vfinfo_size(dev
, ext_filter_mask
) /* IFLA_VFINFO_LIST */
862 + rtnl_port_size(dev
, ext_filter_mask
) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
863 + rtnl_link_get_size(dev
) /* IFLA_LINKINFO */
864 + rtnl_link_get_af_size(dev
) /* IFLA_AF_SPEC */
865 + nla_total_size(MAX_PHYS_PORT_ID_LEN
); /* IFLA_PHYS_PORT_ID */
868 static int rtnl_vf_ports_fill(struct sk_buff
*skb
, struct net_device
*dev
)
870 struct nlattr
*vf_ports
;
871 struct nlattr
*vf_port
;
875 vf_ports
= nla_nest_start(skb
, IFLA_VF_PORTS
);
879 for (vf
= 0; vf
< dev_num_vf(dev
->dev
.parent
); vf
++) {
880 vf_port
= nla_nest_start(skb
, IFLA_VF_PORT
);
882 goto nla_put_failure
;
883 if (nla_put_u32(skb
, IFLA_PORT_VF
, vf
))
884 goto nla_put_failure
;
885 err
= dev
->netdev_ops
->ndo_get_vf_port(dev
, vf
, skb
);
886 if (err
== -EMSGSIZE
)
887 goto nla_put_failure
;
889 nla_nest_cancel(skb
, vf_port
);
892 nla_nest_end(skb
, vf_port
);
895 nla_nest_end(skb
, vf_ports
);
900 nla_nest_cancel(skb
, vf_ports
);
904 static int rtnl_port_self_fill(struct sk_buff
*skb
, struct net_device
*dev
)
906 struct nlattr
*port_self
;
909 port_self
= nla_nest_start(skb
, IFLA_PORT_SELF
);
913 err
= dev
->netdev_ops
->ndo_get_vf_port(dev
, PORT_SELF_VF
, skb
);
915 nla_nest_cancel(skb
, port_self
);
916 return (err
== -EMSGSIZE
) ? err
: 0;
919 nla_nest_end(skb
, port_self
);
924 static int rtnl_port_fill(struct sk_buff
*skb
, struct net_device
*dev
,
929 if (!dev
->netdev_ops
->ndo_get_vf_port
|| !dev
->dev
.parent
||
930 !(ext_filter_mask
& RTEXT_FILTER_VF
))
933 err
= rtnl_port_self_fill(skb
, dev
);
937 if (dev_num_vf(dev
->dev
.parent
)) {
938 err
= rtnl_vf_ports_fill(skb
, dev
);
946 static int rtnl_phys_port_id_fill(struct sk_buff
*skb
, struct net_device
*dev
)
949 struct netdev_phys_port_id ppid
;
951 err
= dev_get_phys_port_id(dev
, &ppid
);
953 if (err
== -EOPNOTSUPP
)
958 if (nla_put(skb
, IFLA_PHYS_PORT_ID
, ppid
.id_len
, ppid
.id
))
964 static int rtnl_fill_ifinfo(struct sk_buff
*skb
, struct net_device
*dev
,
965 int type
, u32 pid
, u32 seq
, u32 change
,
966 unsigned int flags
, u32 ext_filter_mask
)
968 struct ifinfomsg
*ifm
;
969 struct nlmsghdr
*nlh
;
970 struct rtnl_link_stats64 temp
;
971 const struct rtnl_link_stats64
*stats
;
972 struct nlattr
*attr
, *af_spec
;
973 struct rtnl_af_ops
*af_ops
;
974 struct net_device
*upper_dev
= netdev_master_upper_dev_get(dev
);
977 nlh
= nlmsg_put(skb
, pid
, seq
, type
, sizeof(*ifm
), flags
);
981 ifm
= nlmsg_data(nlh
);
982 ifm
->ifi_family
= AF_UNSPEC
;
984 ifm
->ifi_type
= dev
->type
;
985 ifm
->ifi_index
= dev
->ifindex
;
986 ifm
->ifi_flags
= dev_get_flags(dev
);
987 ifm
->ifi_change
= change
;
989 if (nla_put_string(skb
, IFLA_IFNAME
, dev
->name
) ||
990 nla_put_u32(skb
, IFLA_TXQLEN
, dev
->tx_queue_len
) ||
991 nla_put_u8(skb
, IFLA_OPERSTATE
,
992 netif_running(dev
) ? dev
->operstate
: IF_OPER_DOWN
) ||
993 nla_put_u8(skb
, IFLA_LINKMODE
, dev
->link_mode
) ||
994 nla_put_u32(skb
, IFLA_MTU
, dev
->mtu
) ||
995 nla_put_u32(skb
, IFLA_GROUP
, dev
->group
) ||
996 nla_put_u32(skb
, IFLA_PROMISCUITY
, dev
->promiscuity
) ||
997 nla_put_u32(skb
, IFLA_NUM_TX_QUEUES
, dev
->num_tx_queues
) ||
999 nla_put_u32(skb
, IFLA_NUM_RX_QUEUES
, dev
->num_rx_queues
) ||
1001 (dev
->ifindex
!= dev
->iflink
&&
1002 nla_put_u32(skb
, IFLA_LINK
, dev
->iflink
)) ||
1004 nla_put_u32(skb
, IFLA_MASTER
, upper_dev
->ifindex
)) ||
1005 nla_put_u8(skb
, IFLA_CARRIER
, netif_carrier_ok(dev
)) ||
1007 nla_put_string(skb
, IFLA_QDISC
, dev
->qdisc
->ops
->id
)) ||
1009 nla_put_string(skb
, IFLA_IFALIAS
, dev
->ifalias
)) ||
1010 nla_put_u32(skb
, IFLA_CARRIER_CHANGES
,
1011 atomic_read(&dev
->carrier_changes
)))
1012 goto nla_put_failure
;
1015 struct rtnl_link_ifmap map
= {
1016 .mem_start
= dev
->mem_start
,
1017 .mem_end
= dev
->mem_end
,
1018 .base_addr
= dev
->base_addr
,
1021 .port
= dev
->if_port
,
1023 if (nla_put(skb
, IFLA_MAP
, sizeof(map
), &map
))
1024 goto nla_put_failure
;
1027 if (dev
->addr_len
) {
1028 if (nla_put(skb
, IFLA_ADDRESS
, dev
->addr_len
, dev
->dev_addr
) ||
1029 nla_put(skb
, IFLA_BROADCAST
, dev
->addr_len
, dev
->broadcast
))
1030 goto nla_put_failure
;
1033 if (rtnl_phys_port_id_fill(skb
, dev
))
1034 goto nla_put_failure
;
1036 attr
= nla_reserve(skb
, IFLA_STATS
,
1037 sizeof(struct rtnl_link_stats
));
1039 goto nla_put_failure
;
1041 stats
= dev_get_stats(dev
, &temp
);
1042 copy_rtnl_link_stats(nla_data(attr
), stats
);
1044 attr
= nla_reserve(skb
, IFLA_STATS64
,
1045 sizeof(struct rtnl_link_stats64
));
1047 goto nla_put_failure
;
1048 copy_rtnl_link_stats64(nla_data(attr
), stats
);
1050 if (dev
->dev
.parent
&& (ext_filter_mask
& RTEXT_FILTER_VF
) &&
1051 nla_put_u32(skb
, IFLA_NUM_VF
, dev_num_vf(dev
->dev
.parent
)))
1052 goto nla_put_failure
;
1054 if (dev
->netdev_ops
->ndo_get_vf_config
&& dev
->dev
.parent
1055 && (ext_filter_mask
& RTEXT_FILTER_VF
)) {
1058 struct nlattr
*vfinfo
, *vf
;
1059 int num_vfs
= dev_num_vf(dev
->dev
.parent
);
1061 vfinfo
= nla_nest_start(skb
, IFLA_VFINFO_LIST
);
1063 goto nla_put_failure
;
1064 for (i
= 0; i
< num_vfs
; i
++) {
1065 struct ifla_vf_info ivi
;
1066 struct ifla_vf_mac vf_mac
;
1067 struct ifla_vf_vlan vf_vlan
;
1068 struct ifla_vf_rate vf_rate
;
1069 struct ifla_vf_tx_rate vf_tx_rate
;
1070 struct ifla_vf_spoofchk vf_spoofchk
;
1071 struct ifla_vf_link_state vf_linkstate
;
1074 * Not all SR-IOV capable drivers support the
1075 * spoofcheck query. Preset to -1 so the user
1076 * space tool can detect that the driver didn't
1080 memset(ivi
.mac
, 0, sizeof(ivi
.mac
));
1081 /* The default value for VF link state is "auto"
1082 * IFLA_VF_LINK_STATE_AUTO which equals zero
1085 if (dev
->netdev_ops
->ndo_get_vf_config(dev
, i
, &ivi
))
1092 vf_linkstate
.vf
= ivi
.vf
;
1094 memcpy(vf_mac
.mac
, ivi
.mac
, sizeof(ivi
.mac
));
1095 vf_vlan
.vlan
= ivi
.vlan
;
1096 vf_vlan
.qos
= ivi
.qos
;
1097 vf_tx_rate
.rate
= ivi
.max_tx_rate
;
1098 vf_rate
.min_tx_rate
= ivi
.min_tx_rate
;
1099 vf_rate
.max_tx_rate
= ivi
.max_tx_rate
;
1100 vf_spoofchk
.setting
= ivi
.spoofchk
;
1101 vf_linkstate
.link_state
= ivi
.linkstate
;
1102 vf
= nla_nest_start(skb
, IFLA_VF_INFO
);
1104 nla_nest_cancel(skb
, vfinfo
);
1105 goto nla_put_failure
;
1107 if (nla_put(skb
, IFLA_VF_MAC
, sizeof(vf_mac
), &vf_mac
) ||
1108 nla_put(skb
, IFLA_VF_VLAN
, sizeof(vf_vlan
), &vf_vlan
) ||
1109 nla_put(skb
, IFLA_VF_RATE
, sizeof(vf_rate
),
1111 nla_put(skb
, IFLA_VF_TX_RATE
, sizeof(vf_tx_rate
),
1113 nla_put(skb
, IFLA_VF_SPOOFCHK
, sizeof(vf_spoofchk
),
1115 nla_put(skb
, IFLA_VF_LINK_STATE
, sizeof(vf_linkstate
),
1117 goto nla_put_failure
;
1118 nla_nest_end(skb
, vf
);
1120 nla_nest_end(skb
, vfinfo
);
1123 if (rtnl_port_fill(skb
, dev
, ext_filter_mask
))
1124 goto nla_put_failure
;
1126 if (dev
->rtnl_link_ops
|| rtnl_have_link_slave_info(dev
)) {
1127 if (rtnl_link_fill(skb
, dev
) < 0)
1128 goto nla_put_failure
;
1131 if (!(af_spec
= nla_nest_start(skb
, IFLA_AF_SPEC
)))
1132 goto nla_put_failure
;
1134 list_for_each_entry(af_ops
, &rtnl_af_ops
, list
) {
1135 if (af_ops
->fill_link_af
) {
1139 if (!(af
= nla_nest_start(skb
, af_ops
->family
)))
1140 goto nla_put_failure
;
1142 err
= af_ops
->fill_link_af(skb
, dev
);
1145 * Caller may return ENODATA to indicate that there
1146 * was no data to be dumped. This is not an error, it
1147 * means we should trim the attribute header and
1150 if (err
== -ENODATA
)
1151 nla_nest_cancel(skb
, af
);
1153 goto nla_put_failure
;
1155 nla_nest_end(skb
, af
);
1159 nla_nest_end(skb
, af_spec
);
1161 return nlmsg_end(skb
, nlh
);
1164 nlmsg_cancel(skb
, nlh
);
1168 static const struct nla_policy ifla_policy
[IFLA_MAX
+1] = {
1169 [IFLA_IFNAME
] = { .type
= NLA_STRING
, .len
= IFNAMSIZ
-1 },
1170 [IFLA_ADDRESS
] = { .type
= NLA_BINARY
, .len
= MAX_ADDR_LEN
},
1171 [IFLA_BROADCAST
] = { .type
= NLA_BINARY
, .len
= MAX_ADDR_LEN
},
1172 [IFLA_MAP
] = { .len
= sizeof(struct rtnl_link_ifmap
) },
1173 [IFLA_MTU
] = { .type
= NLA_U32
},
1174 [IFLA_LINK
] = { .type
= NLA_U32
},
1175 [IFLA_MASTER
] = { .type
= NLA_U32
},
1176 [IFLA_CARRIER
] = { .type
= NLA_U8
},
1177 [IFLA_TXQLEN
] = { .type
= NLA_U32
},
1178 [IFLA_WEIGHT
] = { .type
= NLA_U32
},
1179 [IFLA_OPERSTATE
] = { .type
= NLA_U8
},
1180 [IFLA_LINKMODE
] = { .type
= NLA_U8
},
1181 [IFLA_LINKINFO
] = { .type
= NLA_NESTED
},
1182 [IFLA_NET_NS_PID
] = { .type
= NLA_U32
},
1183 [IFLA_NET_NS_FD
] = { .type
= NLA_U32
},
1184 [IFLA_IFALIAS
] = { .type
= NLA_STRING
, .len
= IFALIASZ
-1 },
1185 [IFLA_VFINFO_LIST
] = {. type
= NLA_NESTED
},
1186 [IFLA_VF_PORTS
] = { .type
= NLA_NESTED
},
1187 [IFLA_PORT_SELF
] = { .type
= NLA_NESTED
},
1188 [IFLA_AF_SPEC
] = { .type
= NLA_NESTED
},
1189 [IFLA_EXT_MASK
] = { .type
= NLA_U32
},
1190 [IFLA_PROMISCUITY
] = { .type
= NLA_U32
},
1191 [IFLA_NUM_TX_QUEUES
] = { .type
= NLA_U32
},
1192 [IFLA_NUM_RX_QUEUES
] = { .type
= NLA_U32
},
1193 [IFLA_PHYS_PORT_ID
] = { .type
= NLA_BINARY
, .len
= MAX_PHYS_PORT_ID_LEN
},
1194 [IFLA_CARRIER_CHANGES
] = { .type
= NLA_U32
}, /* ignored */
1197 static const struct nla_policy ifla_info_policy
[IFLA_INFO_MAX
+1] = {
1198 [IFLA_INFO_KIND
] = { .type
= NLA_STRING
},
1199 [IFLA_INFO_DATA
] = { .type
= NLA_NESTED
},
1200 [IFLA_INFO_SLAVE_KIND
] = { .type
= NLA_STRING
},
1201 [IFLA_INFO_SLAVE_DATA
] = { .type
= NLA_NESTED
},
1204 static const struct nla_policy ifla_vfinfo_policy
[IFLA_VF_INFO_MAX
+1] = {
1205 [IFLA_VF_INFO
] = { .type
= NLA_NESTED
},
1208 static const struct nla_policy ifla_vf_policy
[IFLA_VF_MAX
+1] = {
1209 [IFLA_VF_MAC
] = { .type
= NLA_BINARY
,
1210 .len
= sizeof(struct ifla_vf_mac
) },
1211 [IFLA_VF_VLAN
] = { .type
= NLA_BINARY
,
1212 .len
= sizeof(struct ifla_vf_vlan
) },
1213 [IFLA_VF_TX_RATE
] = { .type
= NLA_BINARY
,
1214 .len
= sizeof(struct ifla_vf_tx_rate
) },
1215 [IFLA_VF_SPOOFCHK
] = { .type
= NLA_BINARY
,
1216 .len
= sizeof(struct ifla_vf_spoofchk
) },
1217 [IFLA_VF_RATE
] = { .type
= NLA_BINARY
,
1218 .len
= sizeof(struct ifla_vf_rate
) },
1219 [IFLA_VF_LINK_STATE
] = { .type
= NLA_BINARY
,
1220 .len
= sizeof(struct ifla_vf_link_state
) },
1223 static const struct nla_policy ifla_port_policy
[IFLA_PORT_MAX
+1] = {
1224 [IFLA_PORT_VF
] = { .type
= NLA_U32
},
1225 [IFLA_PORT_PROFILE
] = { .type
= NLA_STRING
,
1226 .len
= PORT_PROFILE_MAX
},
1227 [IFLA_PORT_VSI_TYPE
] = { .type
= NLA_BINARY
,
1228 .len
= sizeof(struct ifla_port_vsi
)},
1229 [IFLA_PORT_INSTANCE_UUID
] = { .type
= NLA_BINARY
,
1230 .len
= PORT_UUID_MAX
},
1231 [IFLA_PORT_HOST_UUID
] = { .type
= NLA_STRING
,
1232 .len
= PORT_UUID_MAX
},
1233 [IFLA_PORT_REQUEST
] = { .type
= NLA_U8
, },
1234 [IFLA_PORT_RESPONSE
] = { .type
= NLA_U16
, },
1237 static int rtnl_dump_ifinfo(struct sk_buff
*skb
, struct netlink_callback
*cb
)
1239 struct net
*net
= sock_net(skb
->sk
);
1242 struct net_device
*dev
;
1243 struct hlist_head
*head
;
1244 struct nlattr
*tb
[IFLA_MAX
+1];
1245 u32 ext_filter_mask
= 0;
1250 s_idx
= cb
->args
[1];
1253 cb
->seq
= net
->dev_base_seq
;
1255 /* A hack to preserve kernel<->userspace interface.
1256 * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
1257 * However, before Linux v3.9 the code here assumed rtgenmsg and that's
1258 * what iproute2 < v3.9.0 used.
1259 * We can detect the old iproute2. Even including the IFLA_EXT_MASK
1260 * attribute, its netlink message is shorter than struct ifinfomsg.
1262 hdrlen
= nlmsg_len(cb
->nlh
) < sizeof(struct ifinfomsg
) ?
1263 sizeof(struct rtgenmsg
) : sizeof(struct ifinfomsg
);
1265 if (nlmsg_parse(cb
->nlh
, hdrlen
, tb
, IFLA_MAX
, ifla_policy
) >= 0) {
1267 if (tb
[IFLA_EXT_MASK
])
1268 ext_filter_mask
= nla_get_u32(tb
[IFLA_EXT_MASK
]);
1271 for (h
= s_h
; h
< NETDEV_HASHENTRIES
; h
++, s_idx
= 0) {
1273 head
= &net
->dev_index_head
[h
];
1274 hlist_for_each_entry_rcu(dev
, head
, index_hlist
) {
1277 err
= rtnl_fill_ifinfo(skb
, dev
, RTM_NEWLINK
,
1278 NETLINK_CB(cb
->skb
).portid
,
1279 cb
->nlh
->nlmsg_seq
, 0,
1282 /* If we ran out of room on the first message,
1285 WARN_ON((err
== -EMSGSIZE
) && (skb
->len
== 0));
1290 nl_dump_check_consistent(cb
, nlmsg_hdr(skb
));
1303 int rtnl_nla_parse_ifla(struct nlattr
**tb
, const struct nlattr
*head
, int len
)
1305 return nla_parse(tb
, IFLA_MAX
, head
, len
, ifla_policy
);
1307 EXPORT_SYMBOL(rtnl_nla_parse_ifla
);
1309 struct net
*rtnl_link_get_net(struct net
*src_net
, struct nlattr
*tb
[])
1312 /* Examine the link attributes and figure out which
1313 * network namespace we are talking about.
1315 if (tb
[IFLA_NET_NS_PID
])
1316 net
= get_net_ns_by_pid(nla_get_u32(tb
[IFLA_NET_NS_PID
]));
1317 else if (tb
[IFLA_NET_NS_FD
])
1318 net
= get_net_ns_by_fd(nla_get_u32(tb
[IFLA_NET_NS_FD
]));
1320 net
= get_net(src_net
);
1323 EXPORT_SYMBOL(rtnl_link_get_net
);
1325 static int validate_linkmsg(struct net_device
*dev
, struct nlattr
*tb
[])
1328 if (tb
[IFLA_ADDRESS
] &&
1329 nla_len(tb
[IFLA_ADDRESS
]) < dev
->addr_len
)
1332 if (tb
[IFLA_BROADCAST
] &&
1333 nla_len(tb
[IFLA_BROADCAST
]) < dev
->addr_len
)
1337 if (tb
[IFLA_AF_SPEC
]) {
1341 nla_for_each_nested(af
, tb
[IFLA_AF_SPEC
], rem
) {
1342 const struct rtnl_af_ops
*af_ops
;
1344 if (!(af_ops
= rtnl_af_lookup(nla_type(af
))))
1345 return -EAFNOSUPPORT
;
1347 if (!af_ops
->set_link_af
)
1350 if (af_ops
->validate_link_af
) {
1351 err
= af_ops
->validate_link_af(dev
, af
);
1361 static int do_setvfinfo(struct net_device
*dev
, struct nlattr
*attr
)
1363 int rem
, err
= -EINVAL
;
1365 const struct net_device_ops
*ops
= dev
->netdev_ops
;
1367 nla_for_each_nested(vf
, attr
, rem
) {
1368 switch (nla_type(vf
)) {
1370 struct ifla_vf_mac
*ivm
;
1373 if (ops
->ndo_set_vf_mac
)
1374 err
= ops
->ndo_set_vf_mac(dev
, ivm
->vf
,
1378 case IFLA_VF_VLAN
: {
1379 struct ifla_vf_vlan
*ivv
;
1382 if (ops
->ndo_set_vf_vlan
)
1383 err
= ops
->ndo_set_vf_vlan(dev
, ivv
->vf
,
1388 case IFLA_VF_TX_RATE
: {
1389 struct ifla_vf_tx_rate
*ivt
;
1390 struct ifla_vf_info ivf
;
1393 if (ops
->ndo_get_vf_config
)
1394 err
= ops
->ndo_get_vf_config(dev
, ivt
->vf
,
1399 if (ops
->ndo_set_vf_rate
)
1400 err
= ops
->ndo_set_vf_rate(dev
, ivt
->vf
,
1405 case IFLA_VF_RATE
: {
1406 struct ifla_vf_rate
*ivt
;
1409 if (ops
->ndo_set_vf_rate
)
1410 err
= ops
->ndo_set_vf_rate(dev
, ivt
->vf
,
1415 case IFLA_VF_SPOOFCHK
: {
1416 struct ifla_vf_spoofchk
*ivs
;
1419 if (ops
->ndo_set_vf_spoofchk
)
1420 err
= ops
->ndo_set_vf_spoofchk(dev
, ivs
->vf
,
1424 case IFLA_VF_LINK_STATE
: {
1425 struct ifla_vf_link_state
*ivl
;
1428 if (ops
->ndo_set_vf_link_state
)
1429 err
= ops
->ndo_set_vf_link_state(dev
, ivl
->vf
,
1443 static int do_set_master(struct net_device
*dev
, int ifindex
)
1445 struct net_device
*upper_dev
= netdev_master_upper_dev_get(dev
);
1446 const struct net_device_ops
*ops
;
1450 if (upper_dev
->ifindex
== ifindex
)
1452 ops
= upper_dev
->netdev_ops
;
1453 if (ops
->ndo_del_slave
) {
1454 err
= ops
->ndo_del_slave(upper_dev
, dev
);
1463 upper_dev
= __dev_get_by_index(dev_net(dev
), ifindex
);
1466 ops
= upper_dev
->netdev_ops
;
1467 if (ops
->ndo_add_slave
) {
1468 err
= ops
->ndo_add_slave(upper_dev
, dev
);
1478 static int do_setlink(const struct sk_buff
*skb
,
1479 struct net_device
*dev
, struct ifinfomsg
*ifm
,
1480 struct nlattr
**tb
, char *ifname
, int modified
)
1482 const struct net_device_ops
*ops
= dev
->netdev_ops
;
1485 if (tb
[IFLA_NET_NS_PID
] || tb
[IFLA_NET_NS_FD
]) {
1486 struct net
*net
= rtnl_link_get_net(dev_net(dev
), tb
);
1491 if (!netlink_ns_capable(skb
, net
->user_ns
, CAP_NET_ADMIN
)) {
1495 err
= dev_change_net_namespace(dev
, net
, ifname
);
1503 struct rtnl_link_ifmap
*u_map
;
1506 if (!ops
->ndo_set_config
) {
1511 if (!netif_device_present(dev
)) {
1516 u_map
= nla_data(tb
[IFLA_MAP
]);
1517 k_map
.mem_start
= (unsigned long) u_map
->mem_start
;
1518 k_map
.mem_end
= (unsigned long) u_map
->mem_end
;
1519 k_map
.base_addr
= (unsigned short) u_map
->base_addr
;
1520 k_map
.irq
= (unsigned char) u_map
->irq
;
1521 k_map
.dma
= (unsigned char) u_map
->dma
;
1522 k_map
.port
= (unsigned char) u_map
->port
;
1524 err
= ops
->ndo_set_config(dev
, &k_map
);
1531 if (tb
[IFLA_ADDRESS
]) {
1532 struct sockaddr
*sa
;
1535 len
= sizeof(sa_family_t
) + dev
->addr_len
;
1536 sa
= kmalloc(len
, GFP_KERNEL
);
1541 sa
->sa_family
= dev
->type
;
1542 memcpy(sa
->sa_data
, nla_data(tb
[IFLA_ADDRESS
]),
1544 err
= dev_set_mac_address(dev
, sa
);
1552 err
= dev_set_mtu(dev
, nla_get_u32(tb
[IFLA_MTU
]));
1558 if (tb
[IFLA_GROUP
]) {
1559 dev_set_group(dev
, nla_get_u32(tb
[IFLA_GROUP
]));
1564 * Interface selected by interface index but interface
1565 * name provided implies that a name change has been
1568 if (ifm
->ifi_index
> 0 && ifname
[0]) {
1569 err
= dev_change_name(dev
, ifname
);
1575 if (tb
[IFLA_IFALIAS
]) {
1576 err
= dev_set_alias(dev
, nla_data(tb
[IFLA_IFALIAS
]),
1577 nla_len(tb
[IFLA_IFALIAS
]));
1583 if (tb
[IFLA_BROADCAST
]) {
1584 nla_memcpy(dev
->broadcast
, tb
[IFLA_BROADCAST
], dev
->addr_len
);
1585 call_netdevice_notifiers(NETDEV_CHANGEADDR
, dev
);
1588 if (ifm
->ifi_flags
|| ifm
->ifi_change
) {
1589 err
= dev_change_flags(dev
, rtnl_dev_combine_flags(dev
, ifm
));
1594 if (tb
[IFLA_MASTER
]) {
1595 err
= do_set_master(dev
, nla_get_u32(tb
[IFLA_MASTER
]));
1601 if (tb
[IFLA_CARRIER
]) {
1602 err
= dev_change_carrier(dev
, nla_get_u8(tb
[IFLA_CARRIER
]));
1608 if (tb
[IFLA_TXQLEN
])
1609 dev
->tx_queue_len
= nla_get_u32(tb
[IFLA_TXQLEN
]);
1611 if (tb
[IFLA_OPERSTATE
])
1612 set_operstate(dev
, nla_get_u8(tb
[IFLA_OPERSTATE
]));
1614 if (tb
[IFLA_LINKMODE
]) {
1615 write_lock_bh(&dev_base_lock
);
1616 dev
->link_mode
= nla_get_u8(tb
[IFLA_LINKMODE
]);
1617 write_unlock_bh(&dev_base_lock
);
1620 if (tb
[IFLA_VFINFO_LIST
]) {
1621 struct nlattr
*attr
;
1623 nla_for_each_nested(attr
, tb
[IFLA_VFINFO_LIST
], rem
) {
1624 if (nla_type(attr
) != IFLA_VF_INFO
) {
1628 err
= do_setvfinfo(dev
, attr
);
1636 if (tb
[IFLA_VF_PORTS
]) {
1637 struct nlattr
*port
[IFLA_PORT_MAX
+1];
1638 struct nlattr
*attr
;
1643 if (!ops
->ndo_set_vf_port
)
1646 nla_for_each_nested(attr
, tb
[IFLA_VF_PORTS
], rem
) {
1647 if (nla_type(attr
) != IFLA_VF_PORT
)
1649 err
= nla_parse_nested(port
, IFLA_PORT_MAX
,
1650 attr
, ifla_port_policy
);
1653 if (!port
[IFLA_PORT_VF
]) {
1657 vf
= nla_get_u32(port
[IFLA_PORT_VF
]);
1658 err
= ops
->ndo_set_vf_port(dev
, vf
, port
);
1666 if (tb
[IFLA_PORT_SELF
]) {
1667 struct nlattr
*port
[IFLA_PORT_MAX
+1];
1669 err
= nla_parse_nested(port
, IFLA_PORT_MAX
,
1670 tb
[IFLA_PORT_SELF
], ifla_port_policy
);
1675 if (ops
->ndo_set_vf_port
)
1676 err
= ops
->ndo_set_vf_port(dev
, PORT_SELF_VF
, port
);
1682 if (tb
[IFLA_AF_SPEC
]) {
1686 nla_for_each_nested(af
, tb
[IFLA_AF_SPEC
], rem
) {
1687 const struct rtnl_af_ops
*af_ops
;
1689 if (!(af_ops
= rtnl_af_lookup(nla_type(af
))))
1692 err
= af_ops
->set_link_af(dev
, af
);
1702 if (err
< 0 && modified
)
1703 net_warn_ratelimited("A link change request failed with some changes committed already. Interface %s may have been left with an inconsistent configuration, please check.\n",
1709 static int rtnl_setlink(struct sk_buff
*skb
, struct nlmsghdr
*nlh
)
1711 struct net
*net
= sock_net(skb
->sk
);
1712 struct ifinfomsg
*ifm
;
1713 struct net_device
*dev
;
1715 struct nlattr
*tb
[IFLA_MAX
+1];
1716 char ifname
[IFNAMSIZ
];
1718 err
= nlmsg_parse(nlh
, sizeof(*ifm
), tb
, IFLA_MAX
, ifla_policy
);
1722 if (tb
[IFLA_IFNAME
])
1723 nla_strlcpy(ifname
, tb
[IFLA_IFNAME
], IFNAMSIZ
);
1728 ifm
= nlmsg_data(nlh
);
1729 if (ifm
->ifi_index
> 0)
1730 dev
= __dev_get_by_index(net
, ifm
->ifi_index
);
1731 else if (tb
[IFLA_IFNAME
])
1732 dev
= __dev_get_by_name(net
, ifname
);
1741 err
= validate_linkmsg(dev
, tb
);
1745 err
= do_setlink(skb
, dev
, ifm
, tb
, ifname
, 0);
1750 static int rtnl_dellink(struct sk_buff
*skb
, struct nlmsghdr
*nlh
)
1752 struct net
*net
= sock_net(skb
->sk
);
1753 const struct rtnl_link_ops
*ops
;
1754 struct net_device
*dev
;
1755 struct ifinfomsg
*ifm
;
1756 char ifname
[IFNAMSIZ
];
1757 struct nlattr
*tb
[IFLA_MAX
+1];
1759 LIST_HEAD(list_kill
);
1761 err
= nlmsg_parse(nlh
, sizeof(*ifm
), tb
, IFLA_MAX
, ifla_policy
);
1765 if (tb
[IFLA_IFNAME
])
1766 nla_strlcpy(ifname
, tb
[IFLA_IFNAME
], IFNAMSIZ
);
1768 ifm
= nlmsg_data(nlh
);
1769 if (ifm
->ifi_index
> 0)
1770 dev
= __dev_get_by_index(net
, ifm
->ifi_index
);
1771 else if (tb
[IFLA_IFNAME
])
1772 dev
= __dev_get_by_name(net
, ifname
);
1779 ops
= dev
->rtnl_link_ops
;
1783 ops
->dellink(dev
, &list_kill
);
1784 unregister_netdevice_many(&list_kill
);
1788 int rtnl_configure_link(struct net_device
*dev
, const struct ifinfomsg
*ifm
)
1790 unsigned int old_flags
;
1793 old_flags
= dev
->flags
;
1794 if (ifm
&& (ifm
->ifi_flags
|| ifm
->ifi_change
)) {
1795 err
= __dev_change_flags(dev
, rtnl_dev_combine_flags(dev
, ifm
));
1800 dev
->rtnl_link_state
= RTNL_LINK_INITIALIZED
;
1802 __dev_notify_flags(dev
, old_flags
, ~0U);
1805 EXPORT_SYMBOL(rtnl_configure_link
);
1807 struct net_device
*rtnl_create_link(struct net
*net
,
1808 char *ifname
, const struct rtnl_link_ops
*ops
, struct nlattr
*tb
[])
1811 struct net_device
*dev
;
1812 unsigned int num_tx_queues
= 1;
1813 unsigned int num_rx_queues
= 1;
1815 if (tb
[IFLA_NUM_TX_QUEUES
])
1816 num_tx_queues
= nla_get_u32(tb
[IFLA_NUM_TX_QUEUES
]);
1817 else if (ops
->get_num_tx_queues
)
1818 num_tx_queues
= ops
->get_num_tx_queues();
1820 if (tb
[IFLA_NUM_RX_QUEUES
])
1821 num_rx_queues
= nla_get_u32(tb
[IFLA_NUM_RX_QUEUES
]);
1822 else if (ops
->get_num_rx_queues
)
1823 num_rx_queues
= ops
->get_num_rx_queues();
1826 dev
= alloc_netdev_mqs(ops
->priv_size
, ifname
, ops
->setup
,
1827 num_tx_queues
, num_rx_queues
);
1831 dev_net_set(dev
, net
);
1832 dev
->rtnl_link_ops
= ops
;
1833 dev
->rtnl_link_state
= RTNL_LINK_INITIALIZING
;
1836 dev
->mtu
= nla_get_u32(tb
[IFLA_MTU
]);
1837 if (tb
[IFLA_ADDRESS
]) {
1838 memcpy(dev
->dev_addr
, nla_data(tb
[IFLA_ADDRESS
]),
1839 nla_len(tb
[IFLA_ADDRESS
]));
1840 dev
->addr_assign_type
= NET_ADDR_SET
;
1842 if (tb
[IFLA_BROADCAST
])
1843 memcpy(dev
->broadcast
, nla_data(tb
[IFLA_BROADCAST
]),
1844 nla_len(tb
[IFLA_BROADCAST
]));
1845 if (tb
[IFLA_TXQLEN
])
1846 dev
->tx_queue_len
= nla_get_u32(tb
[IFLA_TXQLEN
]);
1847 if (tb
[IFLA_OPERSTATE
])
1848 set_operstate(dev
, nla_get_u8(tb
[IFLA_OPERSTATE
]));
1849 if (tb
[IFLA_LINKMODE
])
1850 dev
->link_mode
= nla_get_u8(tb
[IFLA_LINKMODE
]);
1852 dev_set_group(dev
, nla_get_u32(tb
[IFLA_GROUP
]));
1857 return ERR_PTR(err
);
1859 EXPORT_SYMBOL(rtnl_create_link
);
1861 static int rtnl_group_changelink(const struct sk_buff
*skb
,
1862 struct net
*net
, int group
,
1863 struct ifinfomsg
*ifm
,
1866 struct net_device
*dev
;
1869 for_each_netdev(net
, dev
) {
1870 if (dev
->group
== group
) {
1871 err
= do_setlink(skb
, dev
, ifm
, tb
, NULL
, 0);
1880 static int rtnl_newlink(struct sk_buff
*skb
, struct nlmsghdr
*nlh
)
1882 struct net
*net
= sock_net(skb
->sk
);
1883 const struct rtnl_link_ops
*ops
;
1884 const struct rtnl_link_ops
*m_ops
= NULL
;
1885 struct net_device
*dev
;
1886 struct net_device
*master_dev
= NULL
;
1887 struct ifinfomsg
*ifm
;
1888 char kind
[MODULE_NAME_LEN
];
1889 char ifname
[IFNAMSIZ
];
1890 struct nlattr
*tb
[IFLA_MAX
+1];
1891 struct nlattr
*linkinfo
[IFLA_INFO_MAX
+1];
1894 #ifdef CONFIG_MODULES
1897 err
= nlmsg_parse(nlh
, sizeof(*ifm
), tb
, IFLA_MAX
, ifla_policy
);
1901 if (tb
[IFLA_IFNAME
])
1902 nla_strlcpy(ifname
, tb
[IFLA_IFNAME
], IFNAMSIZ
);
1906 ifm
= nlmsg_data(nlh
);
1907 if (ifm
->ifi_index
> 0)
1908 dev
= __dev_get_by_index(net
, ifm
->ifi_index
);
1911 dev
= __dev_get_by_name(net
, ifname
);
1917 master_dev
= netdev_master_upper_dev_get(dev
);
1919 m_ops
= master_dev
->rtnl_link_ops
;
1922 err
= validate_linkmsg(dev
, tb
);
1926 if (tb
[IFLA_LINKINFO
]) {
1927 err
= nla_parse_nested(linkinfo
, IFLA_INFO_MAX
,
1928 tb
[IFLA_LINKINFO
], ifla_info_policy
);
1932 memset(linkinfo
, 0, sizeof(linkinfo
));
1934 if (linkinfo
[IFLA_INFO_KIND
]) {
1935 nla_strlcpy(kind
, linkinfo
[IFLA_INFO_KIND
], sizeof(kind
));
1936 ops
= rtnl_link_ops_get(kind
);
1943 struct nlattr
*attr
[ops
? ops
->maxtype
+ 1 : 0];
1944 struct nlattr
*slave_attr
[m_ops
? m_ops
->slave_maxtype
+ 1 : 0];
1945 struct nlattr
**data
= NULL
;
1946 struct nlattr
**slave_data
= NULL
;
1947 struct net
*dest_net
;
1950 if (ops
->maxtype
&& linkinfo
[IFLA_INFO_DATA
]) {
1951 err
= nla_parse_nested(attr
, ops
->maxtype
,
1952 linkinfo
[IFLA_INFO_DATA
],
1958 if (ops
->validate
) {
1959 err
= ops
->validate(tb
, data
);
1966 if (m_ops
->slave_maxtype
&&
1967 linkinfo
[IFLA_INFO_SLAVE_DATA
]) {
1968 err
= nla_parse_nested(slave_attr
,
1969 m_ops
->slave_maxtype
,
1970 linkinfo
[IFLA_INFO_SLAVE_DATA
],
1971 m_ops
->slave_policy
);
1974 slave_data
= slave_attr
;
1976 if (m_ops
->slave_validate
) {
1977 err
= m_ops
->slave_validate(tb
, slave_data
);
1986 if (nlh
->nlmsg_flags
& NLM_F_EXCL
)
1988 if (nlh
->nlmsg_flags
& NLM_F_REPLACE
)
1991 if (linkinfo
[IFLA_INFO_DATA
]) {
1992 if (!ops
|| ops
!= dev
->rtnl_link_ops
||
1996 err
= ops
->changelink(dev
, tb
, data
);
2002 if (linkinfo
[IFLA_INFO_SLAVE_DATA
]) {
2003 if (!m_ops
|| !m_ops
->slave_changelink
)
2006 err
= m_ops
->slave_changelink(master_dev
, dev
,
2013 return do_setlink(skb
, dev
, ifm
, tb
, ifname
, modified
);
2016 if (!(nlh
->nlmsg_flags
& NLM_F_CREATE
)) {
2017 if (ifm
->ifi_index
== 0 && tb
[IFLA_GROUP
])
2018 return rtnl_group_changelink(skb
, net
,
2019 nla_get_u32(tb
[IFLA_GROUP
]),
2024 if (tb
[IFLA_MAP
] || tb
[IFLA_MASTER
] || tb
[IFLA_PROTINFO
])
2028 #ifdef CONFIG_MODULES
2031 request_module("rtnl-link-%s", kind
);
2033 ops
= rtnl_link_ops_get(kind
);
2042 snprintf(ifname
, IFNAMSIZ
, "%s%%d", ops
->kind
);
2044 dest_net
= rtnl_link_get_net(net
, tb
);
2045 if (IS_ERR(dest_net
))
2046 return PTR_ERR(dest_net
);
2048 dev
= rtnl_create_link(dest_net
, ifname
, ops
, tb
);
2054 dev
->ifindex
= ifm
->ifi_index
;
2057 err
= ops
->newlink(net
, dev
, tb
, data
);
2058 /* Drivers should call free_netdev() in ->destructor
2059 * and unregister it on failure after registration
2060 * so that device could be finally freed in rtnl_unlock.
2063 /* If device is not registered at all, free it now */
2064 if (dev
->reg_state
== NETREG_UNINITIALIZED
)
2069 err
= register_netdevice(dev
);
2075 err
= rtnl_configure_link(dev
, ifm
);
2077 unregister_netdevice(dev
);
2084 static int rtnl_getlink(struct sk_buff
*skb
, struct nlmsghdr
* nlh
)
2086 struct net
*net
= sock_net(skb
->sk
);
2087 struct ifinfomsg
*ifm
;
2088 char ifname
[IFNAMSIZ
];
2089 struct nlattr
*tb
[IFLA_MAX
+1];
2090 struct net_device
*dev
= NULL
;
2091 struct sk_buff
*nskb
;
2093 u32 ext_filter_mask
= 0;
2095 err
= nlmsg_parse(nlh
, sizeof(*ifm
), tb
, IFLA_MAX
, ifla_policy
);
2099 if (tb
[IFLA_IFNAME
])
2100 nla_strlcpy(ifname
, tb
[IFLA_IFNAME
], IFNAMSIZ
);
2102 if (tb
[IFLA_EXT_MASK
])
2103 ext_filter_mask
= nla_get_u32(tb
[IFLA_EXT_MASK
]);
2105 ifm
= nlmsg_data(nlh
);
2106 if (ifm
->ifi_index
> 0)
2107 dev
= __dev_get_by_index(net
, ifm
->ifi_index
);
2108 else if (tb
[IFLA_IFNAME
])
2109 dev
= __dev_get_by_name(net
, ifname
);
2116 nskb
= nlmsg_new(if_nlmsg_size(dev
, ext_filter_mask
), GFP_KERNEL
);
2120 err
= rtnl_fill_ifinfo(nskb
, dev
, RTM_NEWLINK
, NETLINK_CB(skb
).portid
,
2121 nlh
->nlmsg_seq
, 0, 0, ext_filter_mask
);
2123 /* -EMSGSIZE implies BUG in if_nlmsg_size */
2124 WARN_ON(err
== -EMSGSIZE
);
2127 err
= rtnl_unicast(nskb
, net
, NETLINK_CB(skb
).portid
);
2132 static u16
rtnl_calcit(struct sk_buff
*skb
, struct nlmsghdr
*nlh
)
2134 struct net
*net
= sock_net(skb
->sk
);
2135 struct net_device
*dev
;
2136 struct nlattr
*tb
[IFLA_MAX
+1];
2137 u32 ext_filter_mask
= 0;
2138 u16 min_ifinfo_dump_size
= 0;
2141 /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
2142 hdrlen
= nlmsg_len(nlh
) < sizeof(struct ifinfomsg
) ?
2143 sizeof(struct rtgenmsg
) : sizeof(struct ifinfomsg
);
2145 if (nlmsg_parse(nlh
, hdrlen
, tb
, IFLA_MAX
, ifla_policy
) >= 0) {
2146 if (tb
[IFLA_EXT_MASK
])
2147 ext_filter_mask
= nla_get_u32(tb
[IFLA_EXT_MASK
]);
2150 if (!ext_filter_mask
)
2151 return NLMSG_GOODSIZE
;
2153 * traverse the list of net devices and compute the minimum
2154 * buffer size based upon the filter mask.
2156 list_for_each_entry(dev
, &net
->dev_base_head
, dev_list
) {
2157 min_ifinfo_dump_size
= max_t(u16
, min_ifinfo_dump_size
,
2162 return min_ifinfo_dump_size
;
2165 static int rtnl_dump_all(struct sk_buff
*skb
, struct netlink_callback
*cb
)
2168 int s_idx
= cb
->family
;
2172 for (idx
= 1; idx
<= RTNL_FAMILY_MAX
; idx
++) {
2173 int type
= cb
->nlh
->nlmsg_type
-RTM_BASE
;
2174 if (idx
< s_idx
|| idx
== PF_PACKET
)
2176 if (rtnl_msg_handlers
[idx
] == NULL
||
2177 rtnl_msg_handlers
[idx
][type
].dumpit
== NULL
)
2180 memset(&cb
->args
[0], 0, sizeof(cb
->args
));
2184 if (rtnl_msg_handlers
[idx
][type
].dumpit(skb
, cb
))
2192 void rtmsg_ifinfo(int type
, struct net_device
*dev
, unsigned int change
,
2195 struct net
*net
= dev_net(dev
);
2196 struct sk_buff
*skb
;
2198 size_t if_info_size
;
2200 skb
= nlmsg_new((if_info_size
= if_nlmsg_size(dev
, 0)), flags
);
2204 err
= rtnl_fill_ifinfo(skb
, dev
, type
, 0, 0, change
, 0, 0);
2206 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
2207 WARN_ON(err
== -EMSGSIZE
);
2211 rtnl_notify(skb
, net
, 0, RTNLGRP_LINK
, NULL
, flags
);
2215 rtnl_set_sk_err(net
, RTNLGRP_LINK
, err
);
2217 EXPORT_SYMBOL(rtmsg_ifinfo
);
2219 static int nlmsg_populate_fdb_fill(struct sk_buff
*skb
,
2220 struct net_device
*dev
,
2221 u8
*addr
, u32 pid
, u32 seq
,
2222 int type
, unsigned int flags
,
2225 struct nlmsghdr
*nlh
;
2228 nlh
= nlmsg_put(skb
, pid
, seq
, type
, sizeof(*ndm
), nlflags
);
2232 ndm
= nlmsg_data(nlh
);
2233 ndm
->ndm_family
= AF_BRIDGE
;
2236 ndm
->ndm_flags
= flags
;
2238 ndm
->ndm_ifindex
= dev
->ifindex
;
2239 ndm
->ndm_state
= NUD_PERMANENT
;
2241 if (nla_put(skb
, NDA_LLADDR
, ETH_ALEN
, addr
))
2242 goto nla_put_failure
;
2244 return nlmsg_end(skb
, nlh
);
2247 nlmsg_cancel(skb
, nlh
);
2251 static inline size_t rtnl_fdb_nlmsg_size(void)
2253 return NLMSG_ALIGN(sizeof(struct ndmsg
)) + nla_total_size(ETH_ALEN
);
2256 static void rtnl_fdb_notify(struct net_device
*dev
, u8
*addr
, int type
)
2258 struct net
*net
= dev_net(dev
);
2259 struct sk_buff
*skb
;
2262 skb
= nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC
);
2266 err
= nlmsg_populate_fdb_fill(skb
, dev
, addr
, 0, 0, type
, NTF_SELF
, 0);
2272 rtnl_notify(skb
, net
, 0, RTNLGRP_NEIGH
, NULL
, GFP_ATOMIC
);
2275 rtnl_set_sk_err(net
, RTNLGRP_NEIGH
, err
);
2279 * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
2281 int ndo_dflt_fdb_add(struct ndmsg
*ndm
,
2282 struct nlattr
*tb
[],
2283 struct net_device
*dev
,
2284 const unsigned char *addr
,
2289 /* If aging addresses are supported device will need to
2290 * implement its own handler for this.
2292 if (ndm
->ndm_state
&& !(ndm
->ndm_state
& NUD_PERMANENT
)) {
2293 pr_info("%s: FDB only supports static addresses\n", dev
->name
);
2297 if (is_unicast_ether_addr(addr
) || is_link_local_ether_addr(addr
))
2298 err
= dev_uc_add_excl(dev
, addr
);
2299 else if (is_multicast_ether_addr(addr
))
2300 err
= dev_mc_add_excl(dev
, addr
);
2302 /* Only return duplicate errors if NLM_F_EXCL is set */
2303 if (err
== -EEXIST
&& !(flags
& NLM_F_EXCL
))
2308 EXPORT_SYMBOL(ndo_dflt_fdb_add
);
2310 static int rtnl_fdb_add(struct sk_buff
*skb
, struct nlmsghdr
*nlh
)
2312 struct net
*net
= sock_net(skb
->sk
);
2314 struct nlattr
*tb
[NDA_MAX
+1];
2315 struct net_device
*dev
;
2319 err
= nlmsg_parse(nlh
, sizeof(*ndm
), tb
, NDA_MAX
, NULL
);
2323 ndm
= nlmsg_data(nlh
);
2324 if (ndm
->ndm_ifindex
== 0) {
2325 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n");
2329 dev
= __dev_get_by_index(net
, ndm
->ndm_ifindex
);
2331 pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n");
2335 if (!tb
[NDA_LLADDR
] || nla_len(tb
[NDA_LLADDR
]) != ETH_ALEN
) {
2336 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n");
2340 addr
= nla_data(tb
[NDA_LLADDR
]);
2344 /* Support fdb on master device the net/bridge default case */
2345 if ((!ndm
->ndm_flags
|| ndm
->ndm_flags
& NTF_MASTER
) &&
2346 (dev
->priv_flags
& IFF_BRIDGE_PORT
)) {
2347 struct net_device
*br_dev
= netdev_master_upper_dev_get(dev
);
2348 const struct net_device_ops
*ops
= br_dev
->netdev_ops
;
2350 err
= ops
->ndo_fdb_add(ndm
, tb
, dev
, addr
, nlh
->nlmsg_flags
);
2354 ndm
->ndm_flags
&= ~NTF_MASTER
;
2357 /* Embedded bridge, macvlan, and any other device support */
2358 if ((ndm
->ndm_flags
& NTF_SELF
)) {
2359 if (dev
->netdev_ops
->ndo_fdb_add
)
2360 err
= dev
->netdev_ops
->ndo_fdb_add(ndm
, tb
, dev
, addr
,
2363 err
= ndo_dflt_fdb_add(ndm
, tb
, dev
, addr
,
2367 rtnl_fdb_notify(dev
, addr
, RTM_NEWNEIGH
);
2368 ndm
->ndm_flags
&= ~NTF_SELF
;
2376 * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
2378 int ndo_dflt_fdb_del(struct ndmsg
*ndm
,
2379 struct nlattr
*tb
[],
2380 struct net_device
*dev
,
2381 const unsigned char *addr
)
2383 int err
= -EOPNOTSUPP
;
2385 /* If aging addresses are supported device will need to
2386 * implement its own handler for this.
2388 if (!(ndm
->ndm_state
& NUD_PERMANENT
)) {
2389 pr_info("%s: FDB only supports static addresses\n", dev
->name
);
2393 if (is_unicast_ether_addr(addr
) || is_link_local_ether_addr(addr
))
2394 err
= dev_uc_del(dev
, addr
);
2395 else if (is_multicast_ether_addr(addr
))
2396 err
= dev_mc_del(dev
, addr
);
2402 EXPORT_SYMBOL(ndo_dflt_fdb_del
);
2404 static int rtnl_fdb_del(struct sk_buff
*skb
, struct nlmsghdr
*nlh
)
2406 struct net
*net
= sock_net(skb
->sk
);
2408 struct nlattr
*tb
[NDA_MAX
+1];
2409 struct net_device
*dev
;
2413 if (!netlink_capable(skb
, CAP_NET_ADMIN
))
2416 err
= nlmsg_parse(nlh
, sizeof(*ndm
), tb
, NDA_MAX
, NULL
);
2420 ndm
= nlmsg_data(nlh
);
2421 if (ndm
->ndm_ifindex
== 0) {
2422 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n");
2426 dev
= __dev_get_by_index(net
, ndm
->ndm_ifindex
);
2428 pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n");
2432 if (!tb
[NDA_LLADDR
] || nla_len(tb
[NDA_LLADDR
]) != ETH_ALEN
) {
2433 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid address\n");
2437 addr
= nla_data(tb
[NDA_LLADDR
]);
2441 /* Support fdb on master device the net/bridge default case */
2442 if ((!ndm
->ndm_flags
|| ndm
->ndm_flags
& NTF_MASTER
) &&
2443 (dev
->priv_flags
& IFF_BRIDGE_PORT
)) {
2444 struct net_device
*br_dev
= netdev_master_upper_dev_get(dev
);
2445 const struct net_device_ops
*ops
= br_dev
->netdev_ops
;
2447 if (ops
->ndo_fdb_del
)
2448 err
= ops
->ndo_fdb_del(ndm
, tb
, dev
, addr
);
2453 ndm
->ndm_flags
&= ~NTF_MASTER
;
2456 /* Embedded bridge, macvlan, and any other device support */
2457 if (ndm
->ndm_flags
& NTF_SELF
) {
2458 if (dev
->netdev_ops
->ndo_fdb_del
)
2459 err
= dev
->netdev_ops
->ndo_fdb_del(ndm
, tb
, dev
, addr
);
2461 err
= ndo_dflt_fdb_del(ndm
, tb
, dev
, addr
);
2464 rtnl_fdb_notify(dev
, addr
, RTM_DELNEIGH
);
2465 ndm
->ndm_flags
&= ~NTF_SELF
;
2472 static int nlmsg_populate_fdb(struct sk_buff
*skb
,
2473 struct netlink_callback
*cb
,
2474 struct net_device
*dev
,
2476 struct netdev_hw_addr_list
*list
)
2478 struct netdev_hw_addr
*ha
;
2482 portid
= NETLINK_CB(cb
->skb
).portid
;
2483 seq
= cb
->nlh
->nlmsg_seq
;
2485 list_for_each_entry(ha
, &list
->list
, list
) {
2486 if (*idx
< cb
->args
[0])
2489 err
= nlmsg_populate_fdb_fill(skb
, dev
, ha
->addr
,
2491 RTM_NEWNEIGH
, NTF_SELF
,
2502 * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
2503 * @nlh: netlink message header
2506 * Default netdevice operation to dump the existing unicast address list.
2507 * Returns number of addresses from list put in skb.
2509 int ndo_dflt_fdb_dump(struct sk_buff
*skb
,
2510 struct netlink_callback
*cb
,
2511 struct net_device
*dev
,
2516 netif_addr_lock_bh(dev
);
2517 err
= nlmsg_populate_fdb(skb
, cb
, dev
, &idx
, &dev
->uc
);
2520 nlmsg_populate_fdb(skb
, cb
, dev
, &idx
, &dev
->mc
);
2522 netif_addr_unlock_bh(dev
);
2525 EXPORT_SYMBOL(ndo_dflt_fdb_dump
);
2527 static int rtnl_fdb_dump(struct sk_buff
*skb
, struct netlink_callback
*cb
)
2530 struct net
*net
= sock_net(skb
->sk
);
2531 struct net_device
*dev
;
2534 for_each_netdev_rcu(net
, dev
) {
2535 if (dev
->priv_flags
& IFF_BRIDGE_PORT
) {
2536 struct net_device
*br_dev
;
2537 const struct net_device_ops
*ops
;
2539 br_dev
= netdev_master_upper_dev_get(dev
);
2540 ops
= br_dev
->netdev_ops
;
2541 if (ops
->ndo_fdb_dump
)
2542 idx
= ops
->ndo_fdb_dump(skb
, cb
, dev
, idx
);
2545 if (dev
->netdev_ops
->ndo_fdb_dump
)
2546 idx
= dev
->netdev_ops
->ndo_fdb_dump(skb
, cb
, dev
, idx
);
2548 idx
= ndo_dflt_fdb_dump(skb
, cb
, dev
, idx
);
2556 int ndo_dflt_bridge_getlink(struct sk_buff
*skb
, u32 pid
, u32 seq
,
2557 struct net_device
*dev
, u16 mode
)
2559 struct nlmsghdr
*nlh
;
2560 struct ifinfomsg
*ifm
;
2561 struct nlattr
*br_afspec
;
2562 u8 operstate
= netif_running(dev
) ? dev
->operstate
: IF_OPER_DOWN
;
2563 struct net_device
*br_dev
= netdev_master_upper_dev_get(dev
);
2565 nlh
= nlmsg_put(skb
, pid
, seq
, RTM_NEWLINK
, sizeof(*ifm
), NLM_F_MULTI
);
2569 ifm
= nlmsg_data(nlh
);
2570 ifm
->ifi_family
= AF_BRIDGE
;
2572 ifm
->ifi_type
= dev
->type
;
2573 ifm
->ifi_index
= dev
->ifindex
;
2574 ifm
->ifi_flags
= dev_get_flags(dev
);
2575 ifm
->ifi_change
= 0;
2578 if (nla_put_string(skb
, IFLA_IFNAME
, dev
->name
) ||
2579 nla_put_u32(skb
, IFLA_MTU
, dev
->mtu
) ||
2580 nla_put_u8(skb
, IFLA_OPERSTATE
, operstate
) ||
2582 nla_put_u32(skb
, IFLA_MASTER
, br_dev
->ifindex
)) ||
2584 nla_put(skb
, IFLA_ADDRESS
, dev
->addr_len
, dev
->dev_addr
)) ||
2585 (dev
->ifindex
!= dev
->iflink
&&
2586 nla_put_u32(skb
, IFLA_LINK
, dev
->iflink
)))
2587 goto nla_put_failure
;
2589 br_afspec
= nla_nest_start(skb
, IFLA_AF_SPEC
);
2591 goto nla_put_failure
;
2593 if (nla_put_u16(skb
, IFLA_BRIDGE_FLAGS
, BRIDGE_FLAGS_SELF
) ||
2594 nla_put_u16(skb
, IFLA_BRIDGE_MODE
, mode
)) {
2595 nla_nest_cancel(skb
, br_afspec
);
2596 goto nla_put_failure
;
2598 nla_nest_end(skb
, br_afspec
);
2600 return nlmsg_end(skb
, nlh
);
2602 nlmsg_cancel(skb
, nlh
);
2605 EXPORT_SYMBOL(ndo_dflt_bridge_getlink
);
2607 static int rtnl_bridge_getlink(struct sk_buff
*skb
, struct netlink_callback
*cb
)
2609 struct net
*net
= sock_net(skb
->sk
);
2610 struct net_device
*dev
;
2612 u32 portid
= NETLINK_CB(cb
->skb
).portid
;
2613 u32 seq
= cb
->nlh
->nlmsg_seq
;
2614 struct nlattr
*extfilt
;
2615 u32 filter_mask
= 0;
2617 extfilt
= nlmsg_find_attr(cb
->nlh
, sizeof(struct ifinfomsg
),
2620 filter_mask
= nla_get_u32(extfilt
);
2623 for_each_netdev_rcu(net
, dev
) {
2624 const struct net_device_ops
*ops
= dev
->netdev_ops
;
2625 struct net_device
*br_dev
= netdev_master_upper_dev_get(dev
);
2627 if (br_dev
&& br_dev
->netdev_ops
->ndo_bridge_getlink
) {
2628 if (idx
>= cb
->args
[0] &&
2629 br_dev
->netdev_ops
->ndo_bridge_getlink(
2630 skb
, portid
, seq
, dev
, filter_mask
) < 0)
2635 if (ops
->ndo_bridge_getlink
) {
2636 if (idx
>= cb
->args
[0] &&
2637 ops
->ndo_bridge_getlink(skb
, portid
, seq
, dev
,
2649 static inline size_t bridge_nlmsg_size(void)
2651 return NLMSG_ALIGN(sizeof(struct ifinfomsg
))
2652 + nla_total_size(IFNAMSIZ
) /* IFLA_IFNAME */
2653 + nla_total_size(MAX_ADDR_LEN
) /* IFLA_ADDRESS */
2654 + nla_total_size(sizeof(u32
)) /* IFLA_MASTER */
2655 + nla_total_size(sizeof(u32
)) /* IFLA_MTU */
2656 + nla_total_size(sizeof(u32
)) /* IFLA_LINK */
2657 + nla_total_size(sizeof(u32
)) /* IFLA_OPERSTATE */
2658 + nla_total_size(sizeof(u8
)) /* IFLA_PROTINFO */
2659 + nla_total_size(sizeof(struct nlattr
)) /* IFLA_AF_SPEC */
2660 + nla_total_size(sizeof(u16
)) /* IFLA_BRIDGE_FLAGS */
2661 + nla_total_size(sizeof(u16
)); /* IFLA_BRIDGE_MODE */
2664 static int rtnl_bridge_notify(struct net_device
*dev
, u16 flags
)
2666 struct net
*net
= dev_net(dev
);
2667 struct net_device
*br_dev
= netdev_master_upper_dev_get(dev
);
2668 struct sk_buff
*skb
;
2669 int err
= -EOPNOTSUPP
;
2671 skb
= nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC
);
2677 if ((!flags
|| (flags
& BRIDGE_FLAGS_MASTER
)) &&
2678 br_dev
&& br_dev
->netdev_ops
->ndo_bridge_getlink
) {
2679 err
= br_dev
->netdev_ops
->ndo_bridge_getlink(skb
, 0, 0, dev
, 0);
2684 if ((flags
& BRIDGE_FLAGS_SELF
) &&
2685 dev
->netdev_ops
->ndo_bridge_getlink
) {
2686 err
= dev
->netdev_ops
->ndo_bridge_getlink(skb
, 0, 0, dev
, 0);
2691 rtnl_notify(skb
, net
, 0, RTNLGRP_LINK
, NULL
, GFP_ATOMIC
);
2694 WARN_ON(err
== -EMSGSIZE
);
2696 rtnl_set_sk_err(net
, RTNLGRP_LINK
, err
);
2700 static int rtnl_bridge_setlink(struct sk_buff
*skb
, struct nlmsghdr
*nlh
)
2702 struct net
*net
= sock_net(skb
->sk
);
2703 struct ifinfomsg
*ifm
;
2704 struct net_device
*dev
;
2705 struct nlattr
*br_spec
, *attr
= NULL
;
2706 int rem
, err
= -EOPNOTSUPP
;
2707 u16 oflags
, flags
= 0;
2708 bool have_flags
= false;
2710 if (nlmsg_len(nlh
) < sizeof(*ifm
))
2713 ifm
= nlmsg_data(nlh
);
2714 if (ifm
->ifi_family
!= AF_BRIDGE
)
2715 return -EPFNOSUPPORT
;
2717 dev
= __dev_get_by_index(net
, ifm
->ifi_index
);
2719 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
2723 br_spec
= nlmsg_find_attr(nlh
, sizeof(struct ifinfomsg
), IFLA_AF_SPEC
);
2725 nla_for_each_nested(attr
, br_spec
, rem
) {
2726 if (nla_type(attr
) == IFLA_BRIDGE_FLAGS
) {
2728 flags
= nla_get_u16(attr
);
2736 if (!flags
|| (flags
& BRIDGE_FLAGS_MASTER
)) {
2737 struct net_device
*br_dev
= netdev_master_upper_dev_get(dev
);
2739 if (!br_dev
|| !br_dev
->netdev_ops
->ndo_bridge_setlink
) {
2744 err
= br_dev
->netdev_ops
->ndo_bridge_setlink(dev
, nlh
);
2748 flags
&= ~BRIDGE_FLAGS_MASTER
;
2751 if ((flags
& BRIDGE_FLAGS_SELF
)) {
2752 if (!dev
->netdev_ops
->ndo_bridge_setlink
)
2755 err
= dev
->netdev_ops
->ndo_bridge_setlink(dev
, nlh
);
2758 flags
&= ~BRIDGE_FLAGS_SELF
;
2762 memcpy(nla_data(attr
), &flags
, sizeof(flags
));
2763 /* Generate event to notify upper layer of bridge change */
2765 err
= rtnl_bridge_notify(dev
, oflags
);
2770 static int rtnl_bridge_dellink(struct sk_buff
*skb
, struct nlmsghdr
*nlh
)
2772 struct net
*net
= sock_net(skb
->sk
);
2773 struct ifinfomsg
*ifm
;
2774 struct net_device
*dev
;
2775 struct nlattr
*br_spec
, *attr
= NULL
;
2776 int rem
, err
= -EOPNOTSUPP
;
2777 u16 oflags
, flags
= 0;
2778 bool have_flags
= false;
2780 if (nlmsg_len(nlh
) < sizeof(*ifm
))
2783 ifm
= nlmsg_data(nlh
);
2784 if (ifm
->ifi_family
!= AF_BRIDGE
)
2785 return -EPFNOSUPPORT
;
2787 dev
= __dev_get_by_index(net
, ifm
->ifi_index
);
2789 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
2793 br_spec
= nlmsg_find_attr(nlh
, sizeof(struct ifinfomsg
), IFLA_AF_SPEC
);
2795 nla_for_each_nested(attr
, br_spec
, rem
) {
2796 if (nla_type(attr
) == IFLA_BRIDGE_FLAGS
) {
2798 flags
= nla_get_u16(attr
);
2806 if (!flags
|| (flags
& BRIDGE_FLAGS_MASTER
)) {
2807 struct net_device
*br_dev
= netdev_master_upper_dev_get(dev
);
2809 if (!br_dev
|| !br_dev
->netdev_ops
->ndo_bridge_dellink
) {
2814 err
= br_dev
->netdev_ops
->ndo_bridge_dellink(dev
, nlh
);
2818 flags
&= ~BRIDGE_FLAGS_MASTER
;
2821 if ((flags
& BRIDGE_FLAGS_SELF
)) {
2822 if (!dev
->netdev_ops
->ndo_bridge_dellink
)
2825 err
= dev
->netdev_ops
->ndo_bridge_dellink(dev
, nlh
);
2828 flags
&= ~BRIDGE_FLAGS_SELF
;
2832 memcpy(nla_data(attr
), &flags
, sizeof(flags
));
2833 /* Generate event to notify upper layer of bridge change */
2835 err
= rtnl_bridge_notify(dev
, oflags
);
2840 /* Process one rtnetlink message. */
2842 static int rtnetlink_rcv_msg(struct sk_buff
*skb
, struct nlmsghdr
*nlh
)
2844 struct net
*net
= sock_net(skb
->sk
);
2845 rtnl_doit_func doit
;
2851 type
= nlh
->nlmsg_type
;
2857 /* All the messages must have at least 1 byte length */
2858 if (nlmsg_len(nlh
) < sizeof(struct rtgenmsg
))
2861 family
= ((struct rtgenmsg
*)nlmsg_data(nlh
))->rtgen_family
;
2865 if (kind
!= 2 && !netlink_net_capable(skb
, CAP_NET_ADMIN
))
2868 if (kind
== 2 && nlh
->nlmsg_flags
&NLM_F_DUMP
) {
2870 rtnl_dumpit_func dumpit
;
2871 rtnl_calcit_func calcit
;
2872 u16 min_dump_alloc
= 0;
2874 dumpit
= rtnl_get_dumpit(family
, type
);
2877 calcit
= rtnl_get_calcit(family
, type
);
2879 min_dump_alloc
= calcit(skb
, nlh
);
2884 struct netlink_dump_control c
= {
2886 .min_dump_alloc
= min_dump_alloc
,
2888 err
= netlink_dump_start(rtnl
, skb
, nlh
, &c
);
2894 doit
= rtnl_get_doit(family
, type
);
2898 return doit(skb
, nlh
);
2901 static void rtnetlink_rcv(struct sk_buff
*skb
)
2904 netlink_rcv_skb(skb
, &rtnetlink_rcv_msg
);
2908 static int rtnetlink_event(struct notifier_block
*this, unsigned long event
, void *ptr
)
2910 struct net_device
*dev
= netdev_notifier_info_to_dev(ptr
);
2916 case NETDEV_POST_INIT
:
2917 case NETDEV_REGISTER
:
2919 case NETDEV_PRE_TYPE_CHANGE
:
2920 case NETDEV_GOING_DOWN
:
2921 case NETDEV_UNREGISTER
:
2922 case NETDEV_UNREGISTER_FINAL
:
2923 case NETDEV_RELEASE
:
2927 rtmsg_ifinfo(RTM_NEWLINK
, dev
, 0, GFP_KERNEL
);
2933 static struct notifier_block rtnetlink_dev_notifier
= {
2934 .notifier_call
= rtnetlink_event
,
2938 static int __net_init
rtnetlink_net_init(struct net
*net
)
2941 struct netlink_kernel_cfg cfg
= {
2942 .groups
= RTNLGRP_MAX
,
2943 .input
= rtnetlink_rcv
,
2944 .cb_mutex
= &rtnl_mutex
,
2945 .flags
= NL_CFG_F_NONROOT_RECV
,
2948 sk
= netlink_kernel_create(net
, NETLINK_ROUTE
, &cfg
);
2955 static void __net_exit
rtnetlink_net_exit(struct net
*net
)
2957 netlink_kernel_release(net
->rtnl
);
2961 static struct pernet_operations rtnetlink_net_ops
= {
2962 .init
= rtnetlink_net_init
,
2963 .exit
= rtnetlink_net_exit
,
2966 void __init
rtnetlink_init(void)
2968 if (register_pernet_subsys(&rtnetlink_net_ops
))
2969 panic("rtnetlink_init: cannot initialize rtnetlink\n");
2971 register_netdevice_notifier(&rtnetlink_dev_notifier
);
2973 rtnl_register(PF_UNSPEC
, RTM_GETLINK
, rtnl_getlink
,
2974 rtnl_dump_ifinfo
, rtnl_calcit
);
2975 rtnl_register(PF_UNSPEC
, RTM_SETLINK
, rtnl_setlink
, NULL
, NULL
);
2976 rtnl_register(PF_UNSPEC
, RTM_NEWLINK
, rtnl_newlink
, NULL
, NULL
);
2977 rtnl_register(PF_UNSPEC
, RTM_DELLINK
, rtnl_dellink
, NULL
, NULL
);
2979 rtnl_register(PF_UNSPEC
, RTM_GETADDR
, NULL
, rtnl_dump_all
, NULL
);
2980 rtnl_register(PF_UNSPEC
, RTM_GETROUTE
, NULL
, rtnl_dump_all
, NULL
);
2982 rtnl_register(PF_BRIDGE
, RTM_NEWNEIGH
, rtnl_fdb_add
, NULL
, NULL
);
2983 rtnl_register(PF_BRIDGE
, RTM_DELNEIGH
, rtnl_fdb_del
, NULL
, NULL
);
2984 rtnl_register(PF_BRIDGE
, RTM_GETNEIGH
, NULL
, rtnl_fdb_dump
, NULL
);
2986 rtnl_register(PF_BRIDGE
, RTM_GETLINK
, NULL
, rtnl_bridge_getlink
, NULL
);
2987 rtnl_register(PF_BRIDGE
, RTM_DELLINK
, rtnl_bridge_dellink
, NULL
, NULL
);
2988 rtnl_register(PF_BRIDGE
, RTM_SETLINK
, rtnl_bridge_setlink
, NULL
, NULL
);