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/if_vlan.h>
40 #include <linux/pci.h>
41 #include <linux/etherdevice.h>
43 #include <linux/uaccess.h>
45 #include <linux/inet.h>
46 #include <linux/netdevice.h>
47 #include <net/switchdev.h>
49 #include <net/protocol.h>
51 #include <net/route.h>
55 #include <net/pkt_sched.h>
56 #include <net/fib_rules.h>
57 #include <net/rtnetlink.h>
58 #include <net/net_namespace.h>
62 rtnl_dumpit_func dumpit
;
63 rtnl_calcit_func calcit
;
66 static DEFINE_MUTEX(rtnl_mutex
);
70 mutex_lock(&rtnl_mutex
);
72 EXPORT_SYMBOL(rtnl_lock
);
74 static struct sk_buff
*defer_kfree_skb_list
;
75 void rtnl_kfree_skbs(struct sk_buff
*head
, struct sk_buff
*tail
)
78 tail
->next
= defer_kfree_skb_list
;
79 defer_kfree_skb_list
= head
;
82 EXPORT_SYMBOL(rtnl_kfree_skbs
);
84 void __rtnl_unlock(void)
86 struct sk_buff
*head
= defer_kfree_skb_list
;
88 defer_kfree_skb_list
= NULL
;
90 mutex_unlock(&rtnl_mutex
);
93 struct sk_buff
*next
= head
->next
;
101 void rtnl_unlock(void)
103 /* This fellow will unlock it for us. */
106 EXPORT_SYMBOL(rtnl_unlock
);
108 int rtnl_trylock(void)
110 return mutex_trylock(&rtnl_mutex
);
112 EXPORT_SYMBOL(rtnl_trylock
);
114 int rtnl_is_locked(void)
116 return mutex_is_locked(&rtnl_mutex
);
118 EXPORT_SYMBOL(rtnl_is_locked
);
120 #ifdef CONFIG_PROVE_LOCKING
121 bool lockdep_rtnl_is_held(void)
123 return lockdep_is_held(&rtnl_mutex
);
125 EXPORT_SYMBOL(lockdep_rtnl_is_held
);
126 #endif /* #ifdef CONFIG_PROVE_LOCKING */
128 static struct rtnl_link
*rtnl_msg_handlers
[RTNL_FAMILY_MAX
+ 1];
130 static inline int rtm_msgindex(int msgtype
)
132 int msgindex
= msgtype
- RTM_BASE
;
135 * msgindex < 0 implies someone tried to register a netlink
136 * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
137 * the message type has not been added to linux/rtnetlink.h
139 BUG_ON(msgindex
< 0 || msgindex
>= RTM_NR_MSGTYPES
);
144 static rtnl_doit_func
rtnl_get_doit(int protocol
, int msgindex
)
146 struct rtnl_link
*tab
;
148 if (protocol
<= RTNL_FAMILY_MAX
)
149 tab
= rtnl_msg_handlers
[protocol
];
153 if (tab
== NULL
|| tab
[msgindex
].doit
== NULL
)
154 tab
= rtnl_msg_handlers
[PF_UNSPEC
];
156 return tab
[msgindex
].doit
;
159 static rtnl_dumpit_func
rtnl_get_dumpit(int protocol
, int msgindex
)
161 struct rtnl_link
*tab
;
163 if (protocol
<= RTNL_FAMILY_MAX
)
164 tab
= rtnl_msg_handlers
[protocol
];
168 if (tab
== NULL
|| tab
[msgindex
].dumpit
== NULL
)
169 tab
= rtnl_msg_handlers
[PF_UNSPEC
];
171 return tab
[msgindex
].dumpit
;
174 static rtnl_calcit_func
rtnl_get_calcit(int protocol
, int msgindex
)
176 struct rtnl_link
*tab
;
178 if (protocol
<= RTNL_FAMILY_MAX
)
179 tab
= rtnl_msg_handlers
[protocol
];
183 if (tab
== NULL
|| tab
[msgindex
].calcit
== NULL
)
184 tab
= rtnl_msg_handlers
[PF_UNSPEC
];
186 return tab
[msgindex
].calcit
;
190 * __rtnl_register - Register a rtnetlink message type
191 * @protocol: Protocol family or PF_UNSPEC
192 * @msgtype: rtnetlink message type
193 * @doit: Function pointer called for each request message
194 * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
195 * @calcit: Function pointer to calc size of dump message
197 * Registers the specified function pointers (at least one of them has
198 * to be non-NULL) to be called whenever a request message for the
199 * specified protocol family and message type is received.
201 * The special protocol family PF_UNSPEC may be used to define fallback
202 * function pointers for the case when no entry for the specific protocol
205 * Returns 0 on success or a negative error code.
207 int __rtnl_register(int protocol
, int msgtype
,
208 rtnl_doit_func doit
, rtnl_dumpit_func dumpit
,
209 rtnl_calcit_func calcit
)
211 struct rtnl_link
*tab
;
214 BUG_ON(protocol
< 0 || protocol
> RTNL_FAMILY_MAX
);
215 msgindex
= rtm_msgindex(msgtype
);
217 tab
= rtnl_msg_handlers
[protocol
];
219 tab
= kcalloc(RTM_NR_MSGTYPES
, sizeof(*tab
), GFP_KERNEL
);
223 rtnl_msg_handlers
[protocol
] = tab
;
227 tab
[msgindex
].doit
= doit
;
230 tab
[msgindex
].dumpit
= dumpit
;
233 tab
[msgindex
].calcit
= calcit
;
237 EXPORT_SYMBOL_GPL(__rtnl_register
);
240 * rtnl_register - Register a rtnetlink message type
242 * Identical to __rtnl_register() but panics on failure. This is useful
243 * as failure of this function is very unlikely, it can only happen due
244 * to lack of memory when allocating the chain to store all message
245 * handlers for a protocol. Meant for use in init functions where lack
246 * of memory implies no sense in continuing.
248 void rtnl_register(int protocol
, int msgtype
,
249 rtnl_doit_func doit
, rtnl_dumpit_func dumpit
,
250 rtnl_calcit_func calcit
)
252 if (__rtnl_register(protocol
, msgtype
, doit
, dumpit
, calcit
) < 0)
253 panic("Unable to register rtnetlink message handler, "
254 "protocol = %d, message type = %d\n",
257 EXPORT_SYMBOL_GPL(rtnl_register
);
260 * rtnl_unregister - Unregister a rtnetlink message type
261 * @protocol: Protocol family or PF_UNSPEC
262 * @msgtype: rtnetlink message type
264 * Returns 0 on success or a negative error code.
266 int rtnl_unregister(int protocol
, int msgtype
)
270 BUG_ON(protocol
< 0 || protocol
> RTNL_FAMILY_MAX
);
271 msgindex
= rtm_msgindex(msgtype
);
273 if (rtnl_msg_handlers
[protocol
] == NULL
)
276 rtnl_msg_handlers
[protocol
][msgindex
].doit
= NULL
;
277 rtnl_msg_handlers
[protocol
][msgindex
].dumpit
= NULL
;
278 rtnl_msg_handlers
[protocol
][msgindex
].calcit
= NULL
;
282 EXPORT_SYMBOL_GPL(rtnl_unregister
);
285 * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
286 * @protocol : Protocol family or PF_UNSPEC
288 * Identical to calling rtnl_unregster() for all registered message types
289 * of a certain protocol family.
291 void rtnl_unregister_all(int protocol
)
293 BUG_ON(protocol
< 0 || protocol
> RTNL_FAMILY_MAX
);
295 kfree(rtnl_msg_handlers
[protocol
]);
296 rtnl_msg_handlers
[protocol
] = NULL
;
298 EXPORT_SYMBOL_GPL(rtnl_unregister_all
);
300 static LIST_HEAD(link_ops
);
302 static const struct rtnl_link_ops
*rtnl_link_ops_get(const char *kind
)
304 const struct rtnl_link_ops
*ops
;
306 list_for_each_entry(ops
, &link_ops
, list
) {
307 if (!strcmp(ops
->kind
, kind
))
314 * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
315 * @ops: struct rtnl_link_ops * to register
317 * The caller must hold the rtnl_mutex. This function should be used
318 * by drivers that create devices during module initialization. It
319 * must be called before registering the devices.
321 * Returns 0 on success or a negative error code.
323 int __rtnl_link_register(struct rtnl_link_ops
*ops
)
325 if (rtnl_link_ops_get(ops
->kind
))
328 /* The check for setup is here because if ops
329 * does not have that filled up, it is not possible
330 * to use the ops for creating device. So do not
331 * fill up dellink as well. That disables rtnl_dellink.
333 if (ops
->setup
&& !ops
->dellink
)
334 ops
->dellink
= unregister_netdevice_queue
;
336 list_add_tail(&ops
->list
, &link_ops
);
339 EXPORT_SYMBOL_GPL(__rtnl_link_register
);
342 * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
343 * @ops: struct rtnl_link_ops * to register
345 * Returns 0 on success or a negative error code.
347 int rtnl_link_register(struct rtnl_link_ops
*ops
)
352 err
= __rtnl_link_register(ops
);
356 EXPORT_SYMBOL_GPL(rtnl_link_register
);
358 static void __rtnl_kill_links(struct net
*net
, struct rtnl_link_ops
*ops
)
360 struct net_device
*dev
;
361 LIST_HEAD(list_kill
);
363 for_each_netdev(net
, dev
) {
364 if (dev
->rtnl_link_ops
== ops
)
365 ops
->dellink(dev
, &list_kill
);
367 unregister_netdevice_many(&list_kill
);
371 * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
372 * @ops: struct rtnl_link_ops * to unregister
374 * The caller must hold the rtnl_mutex.
376 void __rtnl_link_unregister(struct rtnl_link_ops
*ops
)
381 __rtnl_kill_links(net
, ops
);
383 list_del(&ops
->list
);
385 EXPORT_SYMBOL_GPL(__rtnl_link_unregister
);
387 /* Return with the rtnl_lock held when there are no network
388 * devices unregistering in any network namespace.
390 static void rtnl_lock_unregistering_all(void)
394 DEFINE_WAIT_FUNC(wait
, woken_wake_function
);
396 add_wait_queue(&netdev_unregistering_wq
, &wait
);
398 unregistering
= false;
401 if (net
->dev_unreg_count
> 0) {
402 unregistering
= true;
410 wait_woken(&wait
, TASK_UNINTERRUPTIBLE
, MAX_SCHEDULE_TIMEOUT
);
412 remove_wait_queue(&netdev_unregistering_wq
, &wait
);
416 * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
417 * @ops: struct rtnl_link_ops * to unregister
419 void rtnl_link_unregister(struct rtnl_link_ops
*ops
)
421 /* Close the race with cleanup_net() */
422 mutex_lock(&net_mutex
);
423 rtnl_lock_unregistering_all();
424 __rtnl_link_unregister(ops
);
426 mutex_unlock(&net_mutex
);
428 EXPORT_SYMBOL_GPL(rtnl_link_unregister
);
430 static size_t rtnl_link_get_slave_info_data_size(const struct net_device
*dev
)
432 struct net_device
*master_dev
;
433 const struct rtnl_link_ops
*ops
;
435 master_dev
= netdev_master_upper_dev_get((struct net_device
*) dev
);
438 ops
= master_dev
->rtnl_link_ops
;
439 if (!ops
|| !ops
->get_slave_size
)
441 /* IFLA_INFO_SLAVE_DATA + nested data */
442 return nla_total_size(sizeof(struct nlattr
)) +
443 ops
->get_slave_size(master_dev
, dev
);
446 static size_t rtnl_link_get_size(const struct net_device
*dev
)
448 const struct rtnl_link_ops
*ops
= dev
->rtnl_link_ops
;
454 size
= nla_total_size(sizeof(struct nlattr
)) + /* IFLA_LINKINFO */
455 nla_total_size(strlen(ops
->kind
) + 1); /* IFLA_INFO_KIND */
458 /* IFLA_INFO_DATA + nested data */
459 size
+= nla_total_size(sizeof(struct nlattr
)) +
462 if (ops
->get_xstats_size
)
463 /* IFLA_INFO_XSTATS */
464 size
+= nla_total_size(ops
->get_xstats_size(dev
));
466 size
+= rtnl_link_get_slave_info_data_size(dev
);
471 static LIST_HEAD(rtnl_af_ops
);
473 static const struct rtnl_af_ops
*rtnl_af_lookup(const int family
)
475 const struct rtnl_af_ops
*ops
;
477 list_for_each_entry(ops
, &rtnl_af_ops
, list
) {
478 if (ops
->family
== family
)
486 * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
487 * @ops: struct rtnl_af_ops * to register
489 * Returns 0 on success or a negative error code.
491 void rtnl_af_register(struct rtnl_af_ops
*ops
)
494 list_add_tail(&ops
->list
, &rtnl_af_ops
);
497 EXPORT_SYMBOL_GPL(rtnl_af_register
);
500 * __rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
501 * @ops: struct rtnl_af_ops * to unregister
503 * The caller must hold the rtnl_mutex.
505 void __rtnl_af_unregister(struct rtnl_af_ops
*ops
)
507 list_del(&ops
->list
);
509 EXPORT_SYMBOL_GPL(__rtnl_af_unregister
);
512 * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
513 * @ops: struct rtnl_af_ops * to unregister
515 void rtnl_af_unregister(struct rtnl_af_ops
*ops
)
518 __rtnl_af_unregister(ops
);
521 EXPORT_SYMBOL_GPL(rtnl_af_unregister
);
523 static size_t rtnl_link_get_af_size(const struct net_device
*dev
,
526 struct rtnl_af_ops
*af_ops
;
530 size
= nla_total_size(sizeof(struct nlattr
));
532 list_for_each_entry(af_ops
, &rtnl_af_ops
, list
) {
533 if (af_ops
->get_link_af_size
) {
534 /* AF_* + nested data */
535 size
+= nla_total_size(sizeof(struct nlattr
)) +
536 af_ops
->get_link_af_size(dev
, ext_filter_mask
);
543 static bool rtnl_have_link_slave_info(const struct net_device
*dev
)
545 struct net_device
*master_dev
;
547 master_dev
= netdev_master_upper_dev_get((struct net_device
*) dev
);
548 if (master_dev
&& master_dev
->rtnl_link_ops
)
553 static int rtnl_link_slave_info_fill(struct sk_buff
*skb
,
554 const struct net_device
*dev
)
556 struct net_device
*master_dev
;
557 const struct rtnl_link_ops
*ops
;
558 struct nlattr
*slave_data
;
561 master_dev
= netdev_master_upper_dev_get((struct net_device
*) dev
);
564 ops
= master_dev
->rtnl_link_ops
;
567 if (nla_put_string(skb
, IFLA_INFO_SLAVE_KIND
, ops
->kind
) < 0)
569 if (ops
->fill_slave_info
) {
570 slave_data
= nla_nest_start(skb
, IFLA_INFO_SLAVE_DATA
);
573 err
= ops
->fill_slave_info(skb
, master_dev
, dev
);
575 goto err_cancel_slave_data
;
576 nla_nest_end(skb
, slave_data
);
580 err_cancel_slave_data
:
581 nla_nest_cancel(skb
, slave_data
);
585 static int rtnl_link_info_fill(struct sk_buff
*skb
,
586 const struct net_device
*dev
)
588 const struct rtnl_link_ops
*ops
= dev
->rtnl_link_ops
;
594 if (nla_put_string(skb
, IFLA_INFO_KIND
, ops
->kind
) < 0)
596 if (ops
->fill_xstats
) {
597 err
= ops
->fill_xstats(skb
, dev
);
601 if (ops
->fill_info
) {
602 data
= nla_nest_start(skb
, IFLA_INFO_DATA
);
605 err
= ops
->fill_info(skb
, dev
);
607 goto err_cancel_data
;
608 nla_nest_end(skb
, data
);
613 nla_nest_cancel(skb
, data
);
617 static int rtnl_link_fill(struct sk_buff
*skb
, const struct net_device
*dev
)
619 struct nlattr
*linkinfo
;
622 linkinfo
= nla_nest_start(skb
, IFLA_LINKINFO
);
623 if (linkinfo
== NULL
)
626 err
= rtnl_link_info_fill(skb
, dev
);
628 goto err_cancel_link
;
630 err
= rtnl_link_slave_info_fill(skb
, dev
);
632 goto err_cancel_link
;
634 nla_nest_end(skb
, linkinfo
);
638 nla_nest_cancel(skb
, linkinfo
);
643 int rtnetlink_send(struct sk_buff
*skb
, struct net
*net
, u32 pid
, unsigned int group
, int echo
)
645 struct sock
*rtnl
= net
->rtnl
;
648 NETLINK_CB(skb
).dst_group
= group
;
650 atomic_inc(&skb
->users
);
651 netlink_broadcast(rtnl
, skb
, pid
, group
, GFP_KERNEL
);
653 err
= netlink_unicast(rtnl
, skb
, pid
, MSG_DONTWAIT
);
657 int rtnl_unicast(struct sk_buff
*skb
, struct net
*net
, u32 pid
)
659 struct sock
*rtnl
= net
->rtnl
;
661 return nlmsg_unicast(rtnl
, skb
, pid
);
663 EXPORT_SYMBOL(rtnl_unicast
);
665 void rtnl_notify(struct sk_buff
*skb
, struct net
*net
, u32 pid
, u32 group
,
666 struct nlmsghdr
*nlh
, gfp_t flags
)
668 struct sock
*rtnl
= net
->rtnl
;
672 report
= nlmsg_report(nlh
);
674 nlmsg_notify(rtnl
, skb
, pid
, group
, report
, flags
);
676 EXPORT_SYMBOL(rtnl_notify
);
678 void rtnl_set_sk_err(struct net
*net
, u32 group
, int error
)
680 struct sock
*rtnl
= net
->rtnl
;
682 netlink_set_err(rtnl
, 0, group
, error
);
684 EXPORT_SYMBOL(rtnl_set_sk_err
);
686 int rtnetlink_put_metrics(struct sk_buff
*skb
, u32
*metrics
)
691 mx
= nla_nest_start(skb
, RTA_METRICS
);
695 for (i
= 0; i
< RTAX_MAX
; i
++) {
697 if (i
== RTAX_CC_ALGO
- 1) {
698 char tmp
[TCP_CA_NAME_MAX
], *name
;
700 name
= tcp_ca_get_name_by_key(metrics
[i
], tmp
);
703 if (nla_put_string(skb
, i
+ 1, name
))
704 goto nla_put_failure
;
705 } else if (i
== RTAX_FEATURES
- 1) {
706 u32 user_features
= metrics
[i
] & RTAX_FEATURE_MASK
;
710 BUILD_BUG_ON(RTAX_FEATURE_MASK
& DST_FEATURE_MASK
);
711 if (nla_put_u32(skb
, i
+ 1, user_features
))
712 goto nla_put_failure
;
714 if (nla_put_u32(skb
, i
+ 1, metrics
[i
]))
715 goto nla_put_failure
;
722 nla_nest_cancel(skb
, mx
);
726 return nla_nest_end(skb
, mx
);
729 nla_nest_cancel(skb
, mx
);
732 EXPORT_SYMBOL(rtnetlink_put_metrics
);
734 int rtnl_put_cacheinfo(struct sk_buff
*skb
, struct dst_entry
*dst
, u32 id
,
735 long expires
, u32 error
)
737 struct rta_cacheinfo ci
= {
738 .rta_lastuse
= jiffies_delta_to_clock_t(jiffies
- dst
->lastuse
),
739 .rta_used
= dst
->__use
,
740 .rta_clntref
= atomic_read(&(dst
->__refcnt
)),
748 clock
= jiffies_to_clock_t(abs(expires
));
749 clock
= min_t(unsigned long, clock
, INT_MAX
);
750 ci
.rta_expires
= (expires
> 0) ? clock
: -clock
;
752 return nla_put(skb
, RTA_CACHEINFO
, sizeof(ci
), &ci
);
754 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo
);
756 static void set_operstate(struct net_device
*dev
, unsigned char transition
)
758 unsigned char operstate
= dev
->operstate
;
760 switch (transition
) {
762 if ((operstate
== IF_OPER_DORMANT
||
763 operstate
== IF_OPER_UNKNOWN
) &&
765 operstate
= IF_OPER_UP
;
768 case IF_OPER_DORMANT
:
769 if (operstate
== IF_OPER_UP
||
770 operstate
== IF_OPER_UNKNOWN
)
771 operstate
= IF_OPER_DORMANT
;
775 if (dev
->operstate
!= operstate
) {
776 write_lock_bh(&dev_base_lock
);
777 dev
->operstate
= operstate
;
778 write_unlock_bh(&dev_base_lock
);
779 netdev_state_change(dev
);
783 static unsigned int rtnl_dev_get_flags(const struct net_device
*dev
)
785 return (dev
->flags
& ~(IFF_PROMISC
| IFF_ALLMULTI
)) |
786 (dev
->gflags
& (IFF_PROMISC
| IFF_ALLMULTI
));
789 static unsigned int rtnl_dev_combine_flags(const struct net_device
*dev
,
790 const struct ifinfomsg
*ifm
)
792 unsigned int flags
= ifm
->ifi_flags
;
794 /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
796 flags
= (flags
& ifm
->ifi_change
) |
797 (rtnl_dev_get_flags(dev
) & ~ifm
->ifi_change
);
802 static void copy_rtnl_link_stats(struct rtnl_link_stats
*a
,
803 const struct rtnl_link_stats64
*b
)
805 a
->rx_packets
= b
->rx_packets
;
806 a
->tx_packets
= b
->tx_packets
;
807 a
->rx_bytes
= b
->rx_bytes
;
808 a
->tx_bytes
= b
->tx_bytes
;
809 a
->rx_errors
= b
->rx_errors
;
810 a
->tx_errors
= b
->tx_errors
;
811 a
->rx_dropped
= b
->rx_dropped
;
812 a
->tx_dropped
= b
->tx_dropped
;
814 a
->multicast
= b
->multicast
;
815 a
->collisions
= b
->collisions
;
817 a
->rx_length_errors
= b
->rx_length_errors
;
818 a
->rx_over_errors
= b
->rx_over_errors
;
819 a
->rx_crc_errors
= b
->rx_crc_errors
;
820 a
->rx_frame_errors
= b
->rx_frame_errors
;
821 a
->rx_fifo_errors
= b
->rx_fifo_errors
;
822 a
->rx_missed_errors
= b
->rx_missed_errors
;
824 a
->tx_aborted_errors
= b
->tx_aborted_errors
;
825 a
->tx_carrier_errors
= b
->tx_carrier_errors
;
826 a
->tx_fifo_errors
= b
->tx_fifo_errors
;
827 a
->tx_heartbeat_errors
= b
->tx_heartbeat_errors
;
828 a
->tx_window_errors
= b
->tx_window_errors
;
830 a
->rx_compressed
= b
->rx_compressed
;
831 a
->tx_compressed
= b
->tx_compressed
;
833 a
->rx_nohandler
= b
->rx_nohandler
;
837 static inline int rtnl_vfinfo_size(const struct net_device
*dev
,
840 if (dev
->dev
.parent
&& (ext_filter_mask
& RTEXT_FILTER_VF
)) {
841 int num_vfs
= dev_num_vf(dev
->dev
.parent
);
842 size_t size
= nla_total_size(0);
845 nla_total_size(sizeof(struct ifla_vf_mac
)) +
846 nla_total_size(sizeof(struct ifla_vf_vlan
)) +
847 nla_total_size(0) + /* nest IFLA_VF_VLAN_LIST */
848 nla_total_size(MAX_VLAN_LIST_LEN
*
849 sizeof(struct ifla_vf_vlan_info
)) +
850 nla_total_size(sizeof(struct ifla_vf_spoofchk
)) +
851 nla_total_size(sizeof(struct ifla_vf_tx_rate
)) +
852 nla_total_size(sizeof(struct ifla_vf_rate
)) +
853 nla_total_size(sizeof(struct ifla_vf_link_state
)) +
854 nla_total_size(sizeof(struct ifla_vf_rss_query_en
)) +
855 nla_total_size(0) + /* nest IFLA_VF_STATS */
856 /* IFLA_VF_STATS_RX_PACKETS */
857 nla_total_size_64bit(sizeof(__u64
)) +
858 /* IFLA_VF_STATS_TX_PACKETS */
859 nla_total_size_64bit(sizeof(__u64
)) +
860 /* IFLA_VF_STATS_RX_BYTES */
861 nla_total_size_64bit(sizeof(__u64
)) +
862 /* IFLA_VF_STATS_TX_BYTES */
863 nla_total_size_64bit(sizeof(__u64
)) +
864 /* IFLA_VF_STATS_BROADCAST */
865 nla_total_size_64bit(sizeof(__u64
)) +
866 /* IFLA_VF_STATS_MULTICAST */
867 nla_total_size_64bit(sizeof(__u64
)) +
868 nla_total_size(sizeof(struct ifla_vf_trust
)));
874 static size_t rtnl_port_size(const struct net_device
*dev
,
877 size_t port_size
= nla_total_size(4) /* PORT_VF */
878 + nla_total_size(PORT_PROFILE_MAX
) /* PORT_PROFILE */
879 + nla_total_size(PORT_UUID_MAX
) /* PORT_INSTANCE_UUID */
880 + nla_total_size(PORT_UUID_MAX
) /* PORT_HOST_UUID */
881 + nla_total_size(1) /* PROT_VDP_REQUEST */
882 + nla_total_size(2); /* PORT_VDP_RESPONSE */
883 size_t vf_ports_size
= nla_total_size(sizeof(struct nlattr
));
884 size_t vf_port_size
= nla_total_size(sizeof(struct nlattr
))
886 size_t port_self_size
= nla_total_size(sizeof(struct nlattr
))
889 if (!dev
->netdev_ops
->ndo_get_vf_port
|| !dev
->dev
.parent
||
890 !(ext_filter_mask
& RTEXT_FILTER_VF
))
892 if (dev_num_vf(dev
->dev
.parent
))
893 return port_self_size
+ vf_ports_size
+
894 vf_port_size
* dev_num_vf(dev
->dev
.parent
);
896 return port_self_size
;
899 static size_t rtnl_xdp_size(void)
901 size_t xdp_size
= nla_total_size(0) + /* nest IFLA_XDP */
902 nla_total_size(1); /* XDP_ATTACHED */
907 static noinline
size_t if_nlmsg_size(const struct net_device
*dev
,
910 return NLMSG_ALIGN(sizeof(struct ifinfomsg
))
911 + nla_total_size(IFNAMSIZ
) /* IFLA_IFNAME */
912 + nla_total_size(IFALIASZ
) /* IFLA_IFALIAS */
913 + nla_total_size(IFNAMSIZ
) /* IFLA_QDISC */
914 + nla_total_size_64bit(sizeof(struct rtnl_link_ifmap
))
915 + nla_total_size(sizeof(struct rtnl_link_stats
))
916 + nla_total_size_64bit(sizeof(struct rtnl_link_stats64
))
917 + nla_total_size(MAX_ADDR_LEN
) /* IFLA_ADDRESS */
918 + nla_total_size(MAX_ADDR_LEN
) /* IFLA_BROADCAST */
919 + nla_total_size(4) /* IFLA_TXQLEN */
920 + nla_total_size(4) /* IFLA_WEIGHT */
921 + nla_total_size(4) /* IFLA_MTU */
922 + nla_total_size(4) /* IFLA_LINK */
923 + nla_total_size(4) /* IFLA_MASTER */
924 + nla_total_size(1) /* IFLA_CARRIER */
925 + nla_total_size(4) /* IFLA_PROMISCUITY */
926 + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
927 + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
928 + nla_total_size(4) /* IFLA_GSO_MAX_SEGS */
929 + nla_total_size(4) /* IFLA_GSO_MAX_SIZE */
930 + nla_total_size(1) /* IFLA_OPERSTATE */
931 + nla_total_size(1) /* IFLA_LINKMODE */
932 + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
933 + nla_total_size(4) /* IFLA_LINK_NETNSID */
934 + nla_total_size(ext_filter_mask
935 & RTEXT_FILTER_VF
? 4 : 0) /* IFLA_NUM_VF */
936 + rtnl_vfinfo_size(dev
, ext_filter_mask
) /* IFLA_VFINFO_LIST */
937 + rtnl_port_size(dev
, ext_filter_mask
) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
938 + rtnl_link_get_size(dev
) /* IFLA_LINKINFO */
939 + rtnl_link_get_af_size(dev
, ext_filter_mask
) /* IFLA_AF_SPEC */
940 + nla_total_size(MAX_PHYS_ITEM_ID_LEN
) /* IFLA_PHYS_PORT_ID */
941 + nla_total_size(MAX_PHYS_ITEM_ID_LEN
) /* IFLA_PHYS_SWITCH_ID */
942 + nla_total_size(IFNAMSIZ
) /* IFLA_PHYS_PORT_NAME */
943 + rtnl_xdp_size() /* IFLA_XDP */
944 + nla_total_size(1); /* IFLA_PROTO_DOWN */
948 static int rtnl_vf_ports_fill(struct sk_buff
*skb
, struct net_device
*dev
)
950 struct nlattr
*vf_ports
;
951 struct nlattr
*vf_port
;
955 vf_ports
= nla_nest_start(skb
, IFLA_VF_PORTS
);
959 for (vf
= 0; vf
< dev_num_vf(dev
->dev
.parent
); vf
++) {
960 vf_port
= nla_nest_start(skb
, IFLA_VF_PORT
);
962 goto nla_put_failure
;
963 if (nla_put_u32(skb
, IFLA_PORT_VF
, vf
))
964 goto nla_put_failure
;
965 err
= dev
->netdev_ops
->ndo_get_vf_port(dev
, vf
, skb
);
966 if (err
== -EMSGSIZE
)
967 goto nla_put_failure
;
969 nla_nest_cancel(skb
, vf_port
);
972 nla_nest_end(skb
, vf_port
);
975 nla_nest_end(skb
, vf_ports
);
980 nla_nest_cancel(skb
, vf_ports
);
984 static int rtnl_port_self_fill(struct sk_buff
*skb
, struct net_device
*dev
)
986 struct nlattr
*port_self
;
989 port_self
= nla_nest_start(skb
, IFLA_PORT_SELF
);
993 err
= dev
->netdev_ops
->ndo_get_vf_port(dev
, PORT_SELF_VF
, skb
);
995 nla_nest_cancel(skb
, port_self
);
996 return (err
== -EMSGSIZE
) ? err
: 0;
999 nla_nest_end(skb
, port_self
);
1004 static int rtnl_port_fill(struct sk_buff
*skb
, struct net_device
*dev
,
1005 u32 ext_filter_mask
)
1009 if (!dev
->netdev_ops
->ndo_get_vf_port
|| !dev
->dev
.parent
||
1010 !(ext_filter_mask
& RTEXT_FILTER_VF
))
1013 err
= rtnl_port_self_fill(skb
, dev
);
1017 if (dev_num_vf(dev
->dev
.parent
)) {
1018 err
= rtnl_vf_ports_fill(skb
, dev
);
1026 static int rtnl_phys_port_id_fill(struct sk_buff
*skb
, struct net_device
*dev
)
1029 struct netdev_phys_item_id ppid
;
1031 err
= dev_get_phys_port_id(dev
, &ppid
);
1033 if (err
== -EOPNOTSUPP
)
1038 if (nla_put(skb
, IFLA_PHYS_PORT_ID
, ppid
.id_len
, ppid
.id
))
1044 static int rtnl_phys_port_name_fill(struct sk_buff
*skb
, struct net_device
*dev
)
1046 char name
[IFNAMSIZ
];
1049 err
= dev_get_phys_port_name(dev
, name
, sizeof(name
));
1051 if (err
== -EOPNOTSUPP
)
1056 if (nla_put_string(skb
, IFLA_PHYS_PORT_NAME
, name
))
1062 static int rtnl_phys_switch_id_fill(struct sk_buff
*skb
, struct net_device
*dev
)
1065 struct switchdev_attr attr
= {
1067 .id
= SWITCHDEV_ATTR_ID_PORT_PARENT_ID
,
1068 .flags
= SWITCHDEV_F_NO_RECURSE
,
1071 err
= switchdev_port_attr_get(dev
, &attr
);
1073 if (err
== -EOPNOTSUPP
)
1078 if (nla_put(skb
, IFLA_PHYS_SWITCH_ID
, attr
.u
.ppid
.id_len
,
1085 static noinline_for_stack
int rtnl_fill_stats(struct sk_buff
*skb
,
1086 struct net_device
*dev
)
1088 struct rtnl_link_stats64
*sp
;
1089 struct nlattr
*attr
;
1091 attr
= nla_reserve_64bit(skb
, IFLA_STATS64
,
1092 sizeof(struct rtnl_link_stats64
), IFLA_PAD
);
1096 sp
= nla_data(attr
);
1097 dev_get_stats(dev
, sp
);
1099 attr
= nla_reserve(skb
, IFLA_STATS
,
1100 sizeof(struct rtnl_link_stats
));
1104 copy_rtnl_link_stats(nla_data(attr
), sp
);
1109 static noinline_for_stack
int rtnl_fill_vfinfo(struct sk_buff
*skb
,
1110 struct net_device
*dev
,
1112 struct nlattr
*vfinfo
)
1114 struct ifla_vf_rss_query_en vf_rss_query_en
;
1115 struct nlattr
*vf
, *vfstats
, *vfvlanlist
;
1116 struct ifla_vf_link_state vf_linkstate
;
1117 struct ifla_vf_vlan_info vf_vlan_info
;
1118 struct ifla_vf_spoofchk vf_spoofchk
;
1119 struct ifla_vf_tx_rate vf_tx_rate
;
1120 struct ifla_vf_stats vf_stats
;
1121 struct ifla_vf_trust vf_trust
;
1122 struct ifla_vf_vlan vf_vlan
;
1123 struct ifla_vf_rate vf_rate
;
1124 struct ifla_vf_mac vf_mac
;
1125 struct ifla_vf_info ivi
;
1127 /* Not all SR-IOV capable drivers support the
1128 * spoofcheck and "RSS query enable" query. Preset to
1129 * -1 so the user space tool can detect that the driver
1130 * didn't report anything.
1133 ivi
.rss_query_en
= -1;
1135 memset(ivi
.mac
, 0, sizeof(ivi
.mac
));
1136 /* The default value for VF link state is "auto"
1137 * IFLA_VF_LINK_STATE_AUTO which equals zero
1140 /* VLAN Protocol by default is 802.1Q */
1141 ivi
.vlan_proto
= htons(ETH_P_8021Q
);
1142 if (dev
->netdev_ops
->ndo_get_vf_config(dev
, vfs_num
, &ivi
))
1145 memset(&vf_vlan_info
, 0, sizeof(vf_vlan_info
));
1154 vf_rss_query_en
.vf
=
1155 vf_trust
.vf
= ivi
.vf
;
1157 memcpy(vf_mac
.mac
, ivi
.mac
, sizeof(ivi
.mac
));
1158 vf_vlan
.vlan
= ivi
.vlan
;
1159 vf_vlan
.qos
= ivi
.qos
;
1160 vf_vlan_info
.vlan
= ivi
.vlan
;
1161 vf_vlan_info
.qos
= ivi
.qos
;
1162 vf_vlan_info
.vlan_proto
= ivi
.vlan_proto
;
1163 vf_tx_rate
.rate
= ivi
.max_tx_rate
;
1164 vf_rate
.min_tx_rate
= ivi
.min_tx_rate
;
1165 vf_rate
.max_tx_rate
= ivi
.max_tx_rate
;
1166 vf_spoofchk
.setting
= ivi
.spoofchk
;
1167 vf_linkstate
.link_state
= ivi
.linkstate
;
1168 vf_rss_query_en
.setting
= ivi
.rss_query_en
;
1169 vf_trust
.setting
= ivi
.trusted
;
1170 vf
= nla_nest_start(skb
, IFLA_VF_INFO
);
1172 goto nla_put_vfinfo_failure
;
1173 if (nla_put(skb
, IFLA_VF_MAC
, sizeof(vf_mac
), &vf_mac
) ||
1174 nla_put(skb
, IFLA_VF_VLAN
, sizeof(vf_vlan
), &vf_vlan
) ||
1175 nla_put(skb
, IFLA_VF_RATE
, sizeof(vf_rate
),
1177 nla_put(skb
, IFLA_VF_TX_RATE
, sizeof(vf_tx_rate
),
1179 nla_put(skb
, IFLA_VF_SPOOFCHK
, sizeof(vf_spoofchk
),
1181 nla_put(skb
, IFLA_VF_LINK_STATE
, sizeof(vf_linkstate
),
1183 nla_put(skb
, IFLA_VF_RSS_QUERY_EN
,
1184 sizeof(vf_rss_query_en
),
1185 &vf_rss_query_en
) ||
1186 nla_put(skb
, IFLA_VF_TRUST
,
1187 sizeof(vf_trust
), &vf_trust
))
1188 goto nla_put_vf_failure
;
1189 vfvlanlist
= nla_nest_start(skb
, IFLA_VF_VLAN_LIST
);
1191 goto nla_put_vf_failure
;
1192 if (nla_put(skb
, IFLA_VF_VLAN_INFO
, sizeof(vf_vlan_info
),
1194 nla_nest_cancel(skb
, vfvlanlist
);
1195 goto nla_put_vf_failure
;
1197 nla_nest_end(skb
, vfvlanlist
);
1198 memset(&vf_stats
, 0, sizeof(vf_stats
));
1199 if (dev
->netdev_ops
->ndo_get_vf_stats
)
1200 dev
->netdev_ops
->ndo_get_vf_stats(dev
, vfs_num
,
1202 vfstats
= nla_nest_start(skb
, IFLA_VF_STATS
);
1204 goto nla_put_vf_failure
;
1205 if (nla_put_u64_64bit(skb
, IFLA_VF_STATS_RX_PACKETS
,
1206 vf_stats
.rx_packets
, IFLA_VF_STATS_PAD
) ||
1207 nla_put_u64_64bit(skb
, IFLA_VF_STATS_TX_PACKETS
,
1208 vf_stats
.tx_packets
, IFLA_VF_STATS_PAD
) ||
1209 nla_put_u64_64bit(skb
, IFLA_VF_STATS_RX_BYTES
,
1210 vf_stats
.rx_bytes
, IFLA_VF_STATS_PAD
) ||
1211 nla_put_u64_64bit(skb
, IFLA_VF_STATS_TX_BYTES
,
1212 vf_stats
.tx_bytes
, IFLA_VF_STATS_PAD
) ||
1213 nla_put_u64_64bit(skb
, IFLA_VF_STATS_BROADCAST
,
1214 vf_stats
.broadcast
, IFLA_VF_STATS_PAD
) ||
1215 nla_put_u64_64bit(skb
, IFLA_VF_STATS_MULTICAST
,
1216 vf_stats
.multicast
, IFLA_VF_STATS_PAD
)) {
1217 nla_nest_cancel(skb
, vfstats
);
1218 goto nla_put_vf_failure
;
1220 nla_nest_end(skb
, vfstats
);
1221 nla_nest_end(skb
, vf
);
1225 nla_nest_cancel(skb
, vf
);
1226 nla_put_vfinfo_failure
:
1227 nla_nest_cancel(skb
, vfinfo
);
1231 static int rtnl_fill_link_ifmap(struct sk_buff
*skb
, struct net_device
*dev
)
1233 struct rtnl_link_ifmap map
;
1235 memset(&map
, 0, sizeof(map
));
1236 map
.mem_start
= dev
->mem_start
;
1237 map
.mem_end
= dev
->mem_end
;
1238 map
.base_addr
= dev
->base_addr
;
1241 map
.port
= dev
->if_port
;
1243 if (nla_put_64bit(skb
, IFLA_MAP
, sizeof(map
), &map
, IFLA_PAD
))
1249 static u8
rtnl_xdp_attached_mode(struct net_device
*dev
)
1251 const struct net_device_ops
*ops
= dev
->netdev_ops
;
1255 if (rcu_access_pointer(dev
->xdp_prog
))
1256 return XDP_ATTACHED_SKB
;
1257 if (ops
->ndo_xdp
&& __dev_xdp_attached(dev
, ops
->ndo_xdp
))
1258 return XDP_ATTACHED_DRV
;
1260 return XDP_ATTACHED_NONE
;
1263 static int rtnl_xdp_fill(struct sk_buff
*skb
, struct net_device
*dev
)
1268 xdp
= nla_nest_start(skb
, IFLA_XDP
);
1272 err
= nla_put_u8(skb
, IFLA_XDP_ATTACHED
,
1273 rtnl_xdp_attached_mode(dev
));
1277 nla_nest_end(skb
, xdp
);
1281 nla_nest_cancel(skb
, xdp
);
1285 static int rtnl_fill_ifinfo(struct sk_buff
*skb
, struct net_device
*dev
,
1286 int type
, u32 pid
, u32 seq
, u32 change
,
1287 unsigned int flags
, u32 ext_filter_mask
)
1289 struct ifinfomsg
*ifm
;
1290 struct nlmsghdr
*nlh
;
1291 struct nlattr
*af_spec
;
1292 struct rtnl_af_ops
*af_ops
;
1293 struct net_device
*upper_dev
= netdev_master_upper_dev_get(dev
);
1296 nlh
= nlmsg_put(skb
, pid
, seq
, type
, sizeof(*ifm
), flags
);
1300 ifm
= nlmsg_data(nlh
);
1301 ifm
->ifi_family
= AF_UNSPEC
;
1303 ifm
->ifi_type
= dev
->type
;
1304 ifm
->ifi_index
= dev
->ifindex
;
1305 ifm
->ifi_flags
= dev_get_flags(dev
);
1306 ifm
->ifi_change
= change
;
1308 if (nla_put_string(skb
, IFLA_IFNAME
, dev
->name
) ||
1309 nla_put_u32(skb
, IFLA_TXQLEN
, dev
->tx_queue_len
) ||
1310 nla_put_u8(skb
, IFLA_OPERSTATE
,
1311 netif_running(dev
) ? dev
->operstate
: IF_OPER_DOWN
) ||
1312 nla_put_u8(skb
, IFLA_LINKMODE
, dev
->link_mode
) ||
1313 nla_put_u32(skb
, IFLA_MTU
, dev
->mtu
) ||
1314 nla_put_u32(skb
, IFLA_GROUP
, dev
->group
) ||
1315 nla_put_u32(skb
, IFLA_PROMISCUITY
, dev
->promiscuity
) ||
1316 nla_put_u32(skb
, IFLA_NUM_TX_QUEUES
, dev
->num_tx_queues
) ||
1317 nla_put_u32(skb
, IFLA_GSO_MAX_SEGS
, dev
->gso_max_segs
) ||
1318 nla_put_u32(skb
, IFLA_GSO_MAX_SIZE
, dev
->gso_max_size
) ||
1320 nla_put_u32(skb
, IFLA_NUM_RX_QUEUES
, dev
->num_rx_queues
) ||
1322 (dev
->ifindex
!= dev_get_iflink(dev
) &&
1323 nla_put_u32(skb
, IFLA_LINK
, dev_get_iflink(dev
))) ||
1325 nla_put_u32(skb
, IFLA_MASTER
, upper_dev
->ifindex
)) ||
1326 nla_put_u8(skb
, IFLA_CARRIER
, netif_carrier_ok(dev
)) ||
1328 nla_put_string(skb
, IFLA_QDISC
, dev
->qdisc
->ops
->id
)) ||
1330 nla_put_string(skb
, IFLA_IFALIAS
, dev
->ifalias
)) ||
1331 nla_put_u32(skb
, IFLA_CARRIER_CHANGES
,
1332 atomic_read(&dev
->carrier_changes
)) ||
1333 nla_put_u8(skb
, IFLA_PROTO_DOWN
, dev
->proto_down
))
1334 goto nla_put_failure
;
1336 if (rtnl_fill_link_ifmap(skb
, dev
))
1337 goto nla_put_failure
;
1339 if (dev
->addr_len
) {
1340 if (nla_put(skb
, IFLA_ADDRESS
, dev
->addr_len
, dev
->dev_addr
) ||
1341 nla_put(skb
, IFLA_BROADCAST
, dev
->addr_len
, dev
->broadcast
))
1342 goto nla_put_failure
;
1345 if (rtnl_phys_port_id_fill(skb
, dev
))
1346 goto nla_put_failure
;
1348 if (rtnl_phys_port_name_fill(skb
, dev
))
1349 goto nla_put_failure
;
1351 if (rtnl_phys_switch_id_fill(skb
, dev
))
1352 goto nla_put_failure
;
1354 if (rtnl_fill_stats(skb
, dev
))
1355 goto nla_put_failure
;
1357 if (dev
->dev
.parent
&& (ext_filter_mask
& RTEXT_FILTER_VF
) &&
1358 nla_put_u32(skb
, IFLA_NUM_VF
, dev_num_vf(dev
->dev
.parent
)))
1359 goto nla_put_failure
;
1361 if (dev
->netdev_ops
->ndo_get_vf_config
&& dev
->dev
.parent
&&
1362 ext_filter_mask
& RTEXT_FILTER_VF
) {
1364 struct nlattr
*vfinfo
;
1365 int num_vfs
= dev_num_vf(dev
->dev
.parent
);
1367 vfinfo
= nla_nest_start(skb
, IFLA_VFINFO_LIST
);
1369 goto nla_put_failure
;
1370 for (i
= 0; i
< num_vfs
; i
++) {
1371 if (rtnl_fill_vfinfo(skb
, dev
, i
, vfinfo
))
1372 goto nla_put_failure
;
1375 nla_nest_end(skb
, vfinfo
);
1378 if (rtnl_port_fill(skb
, dev
, ext_filter_mask
))
1379 goto nla_put_failure
;
1381 if (rtnl_xdp_fill(skb
, dev
))
1382 goto nla_put_failure
;
1384 if (dev
->rtnl_link_ops
|| rtnl_have_link_slave_info(dev
)) {
1385 if (rtnl_link_fill(skb
, dev
) < 0)
1386 goto nla_put_failure
;
1389 if (dev
->rtnl_link_ops
&&
1390 dev
->rtnl_link_ops
->get_link_net
) {
1391 struct net
*link_net
= dev
->rtnl_link_ops
->get_link_net(dev
);
1393 if (!net_eq(dev_net(dev
), link_net
)) {
1394 int id
= peernet2id_alloc(dev_net(dev
), link_net
);
1396 if (nla_put_s32(skb
, IFLA_LINK_NETNSID
, id
))
1397 goto nla_put_failure
;
1401 if (!(af_spec
= nla_nest_start(skb
, IFLA_AF_SPEC
)))
1402 goto nla_put_failure
;
1404 list_for_each_entry(af_ops
, &rtnl_af_ops
, list
) {
1405 if (af_ops
->fill_link_af
) {
1409 if (!(af
= nla_nest_start(skb
, af_ops
->family
)))
1410 goto nla_put_failure
;
1412 err
= af_ops
->fill_link_af(skb
, dev
, ext_filter_mask
);
1415 * Caller may return ENODATA to indicate that there
1416 * was no data to be dumped. This is not an error, it
1417 * means we should trim the attribute header and
1420 if (err
== -ENODATA
)
1421 nla_nest_cancel(skb
, af
);
1423 goto nla_put_failure
;
1425 nla_nest_end(skb
, af
);
1429 nla_nest_end(skb
, af_spec
);
1431 nlmsg_end(skb
, nlh
);
1435 nlmsg_cancel(skb
, nlh
);
1439 static const struct nla_policy ifla_policy
[IFLA_MAX
+1] = {
1440 [IFLA_IFNAME
] = { .type
= NLA_STRING
, .len
= IFNAMSIZ
-1 },
1441 [IFLA_ADDRESS
] = { .type
= NLA_BINARY
, .len
= MAX_ADDR_LEN
},
1442 [IFLA_BROADCAST
] = { .type
= NLA_BINARY
, .len
= MAX_ADDR_LEN
},
1443 [IFLA_MAP
] = { .len
= sizeof(struct rtnl_link_ifmap
) },
1444 [IFLA_MTU
] = { .type
= NLA_U32
},
1445 [IFLA_LINK
] = { .type
= NLA_U32
},
1446 [IFLA_MASTER
] = { .type
= NLA_U32
},
1447 [IFLA_CARRIER
] = { .type
= NLA_U8
},
1448 [IFLA_TXQLEN
] = { .type
= NLA_U32
},
1449 [IFLA_WEIGHT
] = { .type
= NLA_U32
},
1450 [IFLA_OPERSTATE
] = { .type
= NLA_U8
},
1451 [IFLA_LINKMODE
] = { .type
= NLA_U8
},
1452 [IFLA_LINKINFO
] = { .type
= NLA_NESTED
},
1453 [IFLA_NET_NS_PID
] = { .type
= NLA_U32
},
1454 [IFLA_NET_NS_FD
] = { .type
= NLA_U32
},
1455 [IFLA_IFALIAS
] = { .type
= NLA_STRING
, .len
= IFALIASZ
-1 },
1456 [IFLA_VFINFO_LIST
] = {. type
= NLA_NESTED
},
1457 [IFLA_VF_PORTS
] = { .type
= NLA_NESTED
},
1458 [IFLA_PORT_SELF
] = { .type
= NLA_NESTED
},
1459 [IFLA_AF_SPEC
] = { .type
= NLA_NESTED
},
1460 [IFLA_EXT_MASK
] = { .type
= NLA_U32
},
1461 [IFLA_PROMISCUITY
] = { .type
= NLA_U32
},
1462 [IFLA_NUM_TX_QUEUES
] = { .type
= NLA_U32
},
1463 [IFLA_NUM_RX_QUEUES
] = { .type
= NLA_U32
},
1464 [IFLA_PHYS_PORT_ID
] = { .type
= NLA_BINARY
, .len
= MAX_PHYS_ITEM_ID_LEN
},
1465 [IFLA_CARRIER_CHANGES
] = { .type
= NLA_U32
}, /* ignored */
1466 [IFLA_PHYS_SWITCH_ID
] = { .type
= NLA_BINARY
, .len
= MAX_PHYS_ITEM_ID_LEN
},
1467 [IFLA_LINK_NETNSID
] = { .type
= NLA_S32
},
1468 [IFLA_PROTO_DOWN
] = { .type
= NLA_U8
},
1469 [IFLA_XDP
] = { .type
= NLA_NESTED
},
1472 static const struct nla_policy ifla_info_policy
[IFLA_INFO_MAX
+1] = {
1473 [IFLA_INFO_KIND
] = { .type
= NLA_STRING
},
1474 [IFLA_INFO_DATA
] = { .type
= NLA_NESTED
},
1475 [IFLA_INFO_SLAVE_KIND
] = { .type
= NLA_STRING
},
1476 [IFLA_INFO_SLAVE_DATA
] = { .type
= NLA_NESTED
},
1479 static const struct nla_policy ifla_vf_policy
[IFLA_VF_MAX
+1] = {
1480 [IFLA_VF_MAC
] = { .len
= sizeof(struct ifla_vf_mac
) },
1481 [IFLA_VF_VLAN
] = { .len
= sizeof(struct ifla_vf_vlan
) },
1482 [IFLA_VF_VLAN_LIST
] = { .type
= NLA_NESTED
},
1483 [IFLA_VF_TX_RATE
] = { .len
= sizeof(struct ifla_vf_tx_rate
) },
1484 [IFLA_VF_SPOOFCHK
] = { .len
= sizeof(struct ifla_vf_spoofchk
) },
1485 [IFLA_VF_RATE
] = { .len
= sizeof(struct ifla_vf_rate
) },
1486 [IFLA_VF_LINK_STATE
] = { .len
= sizeof(struct ifla_vf_link_state
) },
1487 [IFLA_VF_RSS_QUERY_EN
] = { .len
= sizeof(struct ifla_vf_rss_query_en
) },
1488 [IFLA_VF_STATS
] = { .type
= NLA_NESTED
},
1489 [IFLA_VF_TRUST
] = { .len
= sizeof(struct ifla_vf_trust
) },
1490 [IFLA_VF_IB_NODE_GUID
] = { .len
= sizeof(struct ifla_vf_guid
) },
1491 [IFLA_VF_IB_PORT_GUID
] = { .len
= sizeof(struct ifla_vf_guid
) },
1494 static const struct nla_policy ifla_port_policy
[IFLA_PORT_MAX
+1] = {
1495 [IFLA_PORT_VF
] = { .type
= NLA_U32
},
1496 [IFLA_PORT_PROFILE
] = { .type
= NLA_STRING
,
1497 .len
= PORT_PROFILE_MAX
},
1498 [IFLA_PORT_INSTANCE_UUID
] = { .type
= NLA_BINARY
,
1499 .len
= PORT_UUID_MAX
},
1500 [IFLA_PORT_HOST_UUID
] = { .type
= NLA_STRING
,
1501 .len
= PORT_UUID_MAX
},
1502 [IFLA_PORT_REQUEST
] = { .type
= NLA_U8
, },
1503 [IFLA_PORT_RESPONSE
] = { .type
= NLA_U16
, },
1505 /* Unused, but we need to keep it here since user space could
1506 * fill it. It's also broken with regard to NLA_BINARY use in
1507 * combination with structs.
1509 [IFLA_PORT_VSI_TYPE
] = { .type
= NLA_BINARY
,
1510 .len
= sizeof(struct ifla_port_vsi
) },
1513 static const struct nla_policy ifla_xdp_policy
[IFLA_XDP_MAX
+ 1] = {
1514 [IFLA_XDP_FD
] = { .type
= NLA_S32
},
1515 [IFLA_XDP_ATTACHED
] = { .type
= NLA_U8
},
1516 [IFLA_XDP_FLAGS
] = { .type
= NLA_U32
},
1519 static const struct rtnl_link_ops
*linkinfo_to_kind_ops(const struct nlattr
*nla
)
1521 const struct rtnl_link_ops
*ops
= NULL
;
1522 struct nlattr
*linfo
[IFLA_INFO_MAX
+ 1];
1524 if (nla_parse_nested(linfo
, IFLA_INFO_MAX
, nla
,
1525 ifla_info_policy
, NULL
) < 0)
1528 if (linfo
[IFLA_INFO_KIND
]) {
1529 char kind
[MODULE_NAME_LEN
];
1531 nla_strlcpy(kind
, linfo
[IFLA_INFO_KIND
], sizeof(kind
));
1532 ops
= rtnl_link_ops_get(kind
);
1538 static bool link_master_filtered(struct net_device
*dev
, int master_idx
)
1540 struct net_device
*master
;
1545 master
= netdev_master_upper_dev_get(dev
);
1546 if (!master
|| master
->ifindex
!= master_idx
)
1552 static bool link_kind_filtered(const struct net_device
*dev
,
1553 const struct rtnl_link_ops
*kind_ops
)
1555 if (kind_ops
&& dev
->rtnl_link_ops
!= kind_ops
)
1561 static bool link_dump_filtered(struct net_device
*dev
,
1563 const struct rtnl_link_ops
*kind_ops
)
1565 if (link_master_filtered(dev
, master_idx
) ||
1566 link_kind_filtered(dev
, kind_ops
))
1572 static int rtnl_dump_ifinfo(struct sk_buff
*skb
, struct netlink_callback
*cb
)
1574 struct net
*net
= sock_net(skb
->sk
);
1577 struct net_device
*dev
;
1578 struct hlist_head
*head
;
1579 struct nlattr
*tb
[IFLA_MAX
+1];
1580 u32 ext_filter_mask
= 0;
1581 const struct rtnl_link_ops
*kind_ops
= NULL
;
1582 unsigned int flags
= NLM_F_MULTI
;
1588 s_idx
= cb
->args
[1];
1590 cb
->seq
= net
->dev_base_seq
;
1592 /* A hack to preserve kernel<->userspace interface.
1593 * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
1594 * However, before Linux v3.9 the code here assumed rtgenmsg and that's
1595 * what iproute2 < v3.9.0 used.
1596 * We can detect the old iproute2. Even including the IFLA_EXT_MASK
1597 * attribute, its netlink message is shorter than struct ifinfomsg.
1599 hdrlen
= nlmsg_len(cb
->nlh
) < sizeof(struct ifinfomsg
) ?
1600 sizeof(struct rtgenmsg
) : sizeof(struct ifinfomsg
);
1602 if (nlmsg_parse(cb
->nlh
, hdrlen
, tb
, IFLA_MAX
,
1603 ifla_policy
, NULL
) >= 0) {
1604 if (tb
[IFLA_EXT_MASK
])
1605 ext_filter_mask
= nla_get_u32(tb
[IFLA_EXT_MASK
]);
1607 if (tb
[IFLA_MASTER
])
1608 master_idx
= nla_get_u32(tb
[IFLA_MASTER
]);
1610 if (tb
[IFLA_LINKINFO
])
1611 kind_ops
= linkinfo_to_kind_ops(tb
[IFLA_LINKINFO
]);
1613 if (master_idx
|| kind_ops
)
1614 flags
|= NLM_F_DUMP_FILTERED
;
1617 for (h
= s_h
; h
< NETDEV_HASHENTRIES
; h
++, s_idx
= 0) {
1619 head
= &net
->dev_index_head
[h
];
1620 hlist_for_each_entry(dev
, head
, index_hlist
) {
1621 if (link_dump_filtered(dev
, master_idx
, kind_ops
))
1625 err
= rtnl_fill_ifinfo(skb
, dev
, RTM_NEWLINK
,
1626 NETLINK_CB(cb
->skb
).portid
,
1627 cb
->nlh
->nlmsg_seq
, 0,
1632 if (likely(skb
->len
))
1638 nl_dump_check_consistent(cb
, nlmsg_hdr(skb
));
1652 int rtnl_nla_parse_ifla(struct nlattr
**tb
, const struct nlattr
*head
, int len
,
1653 struct netlink_ext_ack
*exterr
)
1655 return nla_parse(tb
, IFLA_MAX
, head
, len
, ifla_policy
, exterr
);
1657 EXPORT_SYMBOL(rtnl_nla_parse_ifla
);
1659 struct net
*rtnl_link_get_net(struct net
*src_net
, struct nlattr
*tb
[])
1662 /* Examine the link attributes and figure out which
1663 * network namespace we are talking about.
1665 if (tb
[IFLA_NET_NS_PID
])
1666 net
= get_net_ns_by_pid(nla_get_u32(tb
[IFLA_NET_NS_PID
]));
1667 else if (tb
[IFLA_NET_NS_FD
])
1668 net
= get_net_ns_by_fd(nla_get_u32(tb
[IFLA_NET_NS_FD
]));
1670 net
= get_net(src_net
);
1673 EXPORT_SYMBOL(rtnl_link_get_net
);
1675 static int validate_linkmsg(struct net_device
*dev
, struct nlattr
*tb
[])
1678 if (tb
[IFLA_ADDRESS
] &&
1679 nla_len(tb
[IFLA_ADDRESS
]) < dev
->addr_len
)
1682 if (tb
[IFLA_BROADCAST
] &&
1683 nla_len(tb
[IFLA_BROADCAST
]) < dev
->addr_len
)
1687 if (tb
[IFLA_AF_SPEC
]) {
1691 nla_for_each_nested(af
, tb
[IFLA_AF_SPEC
], rem
) {
1692 const struct rtnl_af_ops
*af_ops
;
1694 if (!(af_ops
= rtnl_af_lookup(nla_type(af
))))
1695 return -EAFNOSUPPORT
;
1697 if (!af_ops
->set_link_af
)
1700 if (af_ops
->validate_link_af
) {
1701 err
= af_ops
->validate_link_af(dev
, af
);
1711 static int handle_infiniband_guid(struct net_device
*dev
, struct ifla_vf_guid
*ivt
,
1714 const struct net_device_ops
*ops
= dev
->netdev_ops
;
1716 return ops
->ndo_set_vf_guid(dev
, ivt
->vf
, ivt
->guid
, guid_type
);
1719 static int handle_vf_guid(struct net_device
*dev
, struct ifla_vf_guid
*ivt
, int guid_type
)
1721 if (dev
->type
!= ARPHRD_INFINIBAND
)
1724 return handle_infiniband_guid(dev
, ivt
, guid_type
);
1727 static int do_setvfinfo(struct net_device
*dev
, struct nlattr
**tb
)
1729 const struct net_device_ops
*ops
= dev
->netdev_ops
;
1732 if (tb
[IFLA_VF_MAC
]) {
1733 struct ifla_vf_mac
*ivm
= nla_data(tb
[IFLA_VF_MAC
]);
1736 if (ops
->ndo_set_vf_mac
)
1737 err
= ops
->ndo_set_vf_mac(dev
, ivm
->vf
,
1743 if (tb
[IFLA_VF_VLAN
]) {
1744 struct ifla_vf_vlan
*ivv
= nla_data(tb
[IFLA_VF_VLAN
]);
1747 if (ops
->ndo_set_vf_vlan
)
1748 err
= ops
->ndo_set_vf_vlan(dev
, ivv
->vf
, ivv
->vlan
,
1750 htons(ETH_P_8021Q
));
1755 if (tb
[IFLA_VF_VLAN_LIST
]) {
1756 struct ifla_vf_vlan_info
*ivvl
[MAX_VLAN_LIST_LEN
];
1757 struct nlattr
*attr
;
1761 if (!ops
->ndo_set_vf_vlan
)
1764 nla_for_each_nested(attr
, tb
[IFLA_VF_VLAN_LIST
], rem
) {
1765 if (nla_type(attr
) != IFLA_VF_VLAN_INFO
||
1766 nla_len(attr
) < NLA_HDRLEN
) {
1769 if (len
>= MAX_VLAN_LIST_LEN
)
1771 ivvl
[len
] = nla_data(attr
);
1778 err
= ops
->ndo_set_vf_vlan(dev
, ivvl
[0]->vf
, ivvl
[0]->vlan
,
1779 ivvl
[0]->qos
, ivvl
[0]->vlan_proto
);
1784 if (tb
[IFLA_VF_TX_RATE
]) {
1785 struct ifla_vf_tx_rate
*ivt
= nla_data(tb
[IFLA_VF_TX_RATE
]);
1786 struct ifla_vf_info ivf
;
1789 if (ops
->ndo_get_vf_config
)
1790 err
= ops
->ndo_get_vf_config(dev
, ivt
->vf
, &ivf
);
1795 if (ops
->ndo_set_vf_rate
)
1796 err
= ops
->ndo_set_vf_rate(dev
, ivt
->vf
,
1803 if (tb
[IFLA_VF_RATE
]) {
1804 struct ifla_vf_rate
*ivt
= nla_data(tb
[IFLA_VF_RATE
]);
1807 if (ops
->ndo_set_vf_rate
)
1808 err
= ops
->ndo_set_vf_rate(dev
, ivt
->vf
,
1815 if (tb
[IFLA_VF_SPOOFCHK
]) {
1816 struct ifla_vf_spoofchk
*ivs
= nla_data(tb
[IFLA_VF_SPOOFCHK
]);
1819 if (ops
->ndo_set_vf_spoofchk
)
1820 err
= ops
->ndo_set_vf_spoofchk(dev
, ivs
->vf
,
1826 if (tb
[IFLA_VF_LINK_STATE
]) {
1827 struct ifla_vf_link_state
*ivl
= nla_data(tb
[IFLA_VF_LINK_STATE
]);
1830 if (ops
->ndo_set_vf_link_state
)
1831 err
= ops
->ndo_set_vf_link_state(dev
, ivl
->vf
,
1837 if (tb
[IFLA_VF_RSS_QUERY_EN
]) {
1838 struct ifla_vf_rss_query_en
*ivrssq_en
;
1841 ivrssq_en
= nla_data(tb
[IFLA_VF_RSS_QUERY_EN
]);
1842 if (ops
->ndo_set_vf_rss_query_en
)
1843 err
= ops
->ndo_set_vf_rss_query_en(dev
, ivrssq_en
->vf
,
1844 ivrssq_en
->setting
);
1849 if (tb
[IFLA_VF_TRUST
]) {
1850 struct ifla_vf_trust
*ivt
= nla_data(tb
[IFLA_VF_TRUST
]);
1853 if (ops
->ndo_set_vf_trust
)
1854 err
= ops
->ndo_set_vf_trust(dev
, ivt
->vf
, ivt
->setting
);
1859 if (tb
[IFLA_VF_IB_NODE_GUID
]) {
1860 struct ifla_vf_guid
*ivt
= nla_data(tb
[IFLA_VF_IB_NODE_GUID
]);
1862 if (!ops
->ndo_set_vf_guid
)
1865 return handle_vf_guid(dev
, ivt
, IFLA_VF_IB_NODE_GUID
);
1868 if (tb
[IFLA_VF_IB_PORT_GUID
]) {
1869 struct ifla_vf_guid
*ivt
= nla_data(tb
[IFLA_VF_IB_PORT_GUID
]);
1871 if (!ops
->ndo_set_vf_guid
)
1874 return handle_vf_guid(dev
, ivt
, IFLA_VF_IB_PORT_GUID
);
1880 static int do_set_master(struct net_device
*dev
, int ifindex
)
1882 struct net_device
*upper_dev
= netdev_master_upper_dev_get(dev
);
1883 const struct net_device_ops
*ops
;
1887 if (upper_dev
->ifindex
== ifindex
)
1889 ops
= upper_dev
->netdev_ops
;
1890 if (ops
->ndo_del_slave
) {
1891 err
= ops
->ndo_del_slave(upper_dev
, dev
);
1900 upper_dev
= __dev_get_by_index(dev_net(dev
), ifindex
);
1903 ops
= upper_dev
->netdev_ops
;
1904 if (ops
->ndo_add_slave
) {
1905 err
= ops
->ndo_add_slave(upper_dev
, dev
);
1915 #define DO_SETLINK_MODIFIED 0x01
1916 /* notify flag means notify + modified. */
1917 #define DO_SETLINK_NOTIFY 0x03
1918 static int do_setlink(const struct sk_buff
*skb
,
1919 struct net_device
*dev
, struct ifinfomsg
*ifm
,
1920 struct netlink_ext_ack
*extack
,
1921 struct nlattr
**tb
, char *ifname
, int status
)
1923 const struct net_device_ops
*ops
= dev
->netdev_ops
;
1926 if (tb
[IFLA_NET_NS_PID
] || tb
[IFLA_NET_NS_FD
]) {
1927 struct net
*net
= rtnl_link_get_net(dev_net(dev
), tb
);
1932 if (!netlink_ns_capable(skb
, net
->user_ns
, CAP_NET_ADMIN
)) {
1937 err
= dev_change_net_namespace(dev
, net
, ifname
);
1941 status
|= DO_SETLINK_MODIFIED
;
1945 struct rtnl_link_ifmap
*u_map
;
1948 if (!ops
->ndo_set_config
) {
1953 if (!netif_device_present(dev
)) {
1958 u_map
= nla_data(tb
[IFLA_MAP
]);
1959 k_map
.mem_start
= (unsigned long) u_map
->mem_start
;
1960 k_map
.mem_end
= (unsigned long) u_map
->mem_end
;
1961 k_map
.base_addr
= (unsigned short) u_map
->base_addr
;
1962 k_map
.irq
= (unsigned char) u_map
->irq
;
1963 k_map
.dma
= (unsigned char) u_map
->dma
;
1964 k_map
.port
= (unsigned char) u_map
->port
;
1966 err
= ops
->ndo_set_config(dev
, &k_map
);
1970 status
|= DO_SETLINK_NOTIFY
;
1973 if (tb
[IFLA_ADDRESS
]) {
1974 struct sockaddr
*sa
;
1977 len
= sizeof(sa_family_t
) + dev
->addr_len
;
1978 sa
= kmalloc(len
, GFP_KERNEL
);
1983 sa
->sa_family
= dev
->type
;
1984 memcpy(sa
->sa_data
, nla_data(tb
[IFLA_ADDRESS
]),
1986 err
= dev_set_mac_address(dev
, sa
);
1990 status
|= DO_SETLINK_MODIFIED
;
1994 err
= dev_set_mtu(dev
, nla_get_u32(tb
[IFLA_MTU
]));
1997 status
|= DO_SETLINK_MODIFIED
;
2000 if (tb
[IFLA_GROUP
]) {
2001 dev_set_group(dev
, nla_get_u32(tb
[IFLA_GROUP
]));
2002 status
|= DO_SETLINK_NOTIFY
;
2006 * Interface selected by interface index but interface
2007 * name provided implies that a name change has been
2010 if (ifm
->ifi_index
> 0 && ifname
[0]) {
2011 err
= dev_change_name(dev
, ifname
);
2014 status
|= DO_SETLINK_MODIFIED
;
2017 if (tb
[IFLA_IFALIAS
]) {
2018 err
= dev_set_alias(dev
, nla_data(tb
[IFLA_IFALIAS
]),
2019 nla_len(tb
[IFLA_IFALIAS
]));
2022 status
|= DO_SETLINK_NOTIFY
;
2025 if (tb
[IFLA_BROADCAST
]) {
2026 nla_memcpy(dev
->broadcast
, tb
[IFLA_BROADCAST
], dev
->addr_len
);
2027 call_netdevice_notifiers(NETDEV_CHANGEADDR
, dev
);
2030 if (ifm
->ifi_flags
|| ifm
->ifi_change
) {
2031 err
= dev_change_flags(dev
, rtnl_dev_combine_flags(dev
, ifm
));
2036 if (tb
[IFLA_MASTER
]) {
2037 err
= do_set_master(dev
, nla_get_u32(tb
[IFLA_MASTER
]));
2040 status
|= DO_SETLINK_MODIFIED
;
2043 if (tb
[IFLA_CARRIER
]) {
2044 err
= dev_change_carrier(dev
, nla_get_u8(tb
[IFLA_CARRIER
]));
2047 status
|= DO_SETLINK_MODIFIED
;
2050 if (tb
[IFLA_TXQLEN
]) {
2051 unsigned long value
= nla_get_u32(tb
[IFLA_TXQLEN
]);
2052 unsigned long orig_len
= dev
->tx_queue_len
;
2054 if (dev
->tx_queue_len
^ value
) {
2055 dev
->tx_queue_len
= value
;
2056 err
= call_netdevice_notifiers(
2057 NETDEV_CHANGE_TX_QUEUE_LEN
, dev
);
2058 err
= notifier_to_errno(err
);
2060 dev
->tx_queue_len
= orig_len
;
2063 status
|= DO_SETLINK_NOTIFY
;
2067 if (tb
[IFLA_OPERSTATE
])
2068 set_operstate(dev
, nla_get_u8(tb
[IFLA_OPERSTATE
]));
2070 if (tb
[IFLA_LINKMODE
]) {
2071 unsigned char value
= nla_get_u8(tb
[IFLA_LINKMODE
]);
2073 write_lock_bh(&dev_base_lock
);
2074 if (dev
->link_mode
^ value
)
2075 status
|= DO_SETLINK_NOTIFY
;
2076 dev
->link_mode
= value
;
2077 write_unlock_bh(&dev_base_lock
);
2080 if (tb
[IFLA_VFINFO_LIST
]) {
2081 struct nlattr
*vfinfo
[IFLA_VF_MAX
+ 1];
2082 struct nlattr
*attr
;
2085 nla_for_each_nested(attr
, tb
[IFLA_VFINFO_LIST
], rem
) {
2086 if (nla_type(attr
) != IFLA_VF_INFO
||
2087 nla_len(attr
) < NLA_HDRLEN
) {
2091 err
= nla_parse_nested(vfinfo
, IFLA_VF_MAX
, attr
,
2092 ifla_vf_policy
, NULL
);
2095 err
= do_setvfinfo(dev
, vfinfo
);
2098 status
|= DO_SETLINK_NOTIFY
;
2103 if (tb
[IFLA_VF_PORTS
]) {
2104 struct nlattr
*port
[IFLA_PORT_MAX
+1];
2105 struct nlattr
*attr
;
2110 if (!ops
->ndo_set_vf_port
)
2113 nla_for_each_nested(attr
, tb
[IFLA_VF_PORTS
], rem
) {
2114 if (nla_type(attr
) != IFLA_VF_PORT
||
2115 nla_len(attr
) < NLA_HDRLEN
) {
2119 err
= nla_parse_nested(port
, IFLA_PORT_MAX
, attr
,
2120 ifla_port_policy
, NULL
);
2123 if (!port
[IFLA_PORT_VF
]) {
2127 vf
= nla_get_u32(port
[IFLA_PORT_VF
]);
2128 err
= ops
->ndo_set_vf_port(dev
, vf
, port
);
2131 status
|= DO_SETLINK_NOTIFY
;
2136 if (tb
[IFLA_PORT_SELF
]) {
2137 struct nlattr
*port
[IFLA_PORT_MAX
+1];
2139 err
= nla_parse_nested(port
, IFLA_PORT_MAX
,
2140 tb
[IFLA_PORT_SELF
], ifla_port_policy
,
2146 if (ops
->ndo_set_vf_port
)
2147 err
= ops
->ndo_set_vf_port(dev
, PORT_SELF_VF
, port
);
2150 status
|= DO_SETLINK_NOTIFY
;
2153 if (tb
[IFLA_AF_SPEC
]) {
2157 nla_for_each_nested(af
, tb
[IFLA_AF_SPEC
], rem
) {
2158 const struct rtnl_af_ops
*af_ops
;
2160 if (!(af_ops
= rtnl_af_lookup(nla_type(af
))))
2163 err
= af_ops
->set_link_af(dev
, af
);
2167 status
|= DO_SETLINK_NOTIFY
;
2172 if (tb
[IFLA_PROTO_DOWN
]) {
2173 err
= dev_change_proto_down(dev
,
2174 nla_get_u8(tb
[IFLA_PROTO_DOWN
]));
2177 status
|= DO_SETLINK_NOTIFY
;
2181 struct nlattr
*xdp
[IFLA_XDP_MAX
+ 1];
2184 err
= nla_parse_nested(xdp
, IFLA_XDP_MAX
, tb
[IFLA_XDP
],
2185 ifla_xdp_policy
, NULL
);
2189 if (xdp
[IFLA_XDP_ATTACHED
]) {
2194 if (xdp
[IFLA_XDP_FLAGS
]) {
2195 xdp_flags
= nla_get_u32(xdp
[IFLA_XDP_FLAGS
]);
2196 if (xdp_flags
& ~XDP_FLAGS_MASK
) {
2200 if ((xdp_flags
& XDP_FLAGS_SKB_MODE
) &&
2201 (xdp_flags
& XDP_FLAGS_DRV_MODE
)) {
2207 if (xdp
[IFLA_XDP_FD
]) {
2208 err
= dev_change_xdp_fd(dev
, extack
,
2209 nla_get_s32(xdp
[IFLA_XDP_FD
]),
2213 status
|= DO_SETLINK_NOTIFY
;
2218 if (status
& DO_SETLINK_MODIFIED
) {
2219 if (status
& DO_SETLINK_NOTIFY
)
2220 netdev_state_change(dev
);
2223 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",
2230 static int rtnl_setlink(struct sk_buff
*skb
, struct nlmsghdr
*nlh
,
2231 struct netlink_ext_ack
*extack
)
2233 struct net
*net
= sock_net(skb
->sk
);
2234 struct ifinfomsg
*ifm
;
2235 struct net_device
*dev
;
2237 struct nlattr
*tb
[IFLA_MAX
+1];
2238 char ifname
[IFNAMSIZ
];
2240 err
= nlmsg_parse(nlh
, sizeof(*ifm
), tb
, IFLA_MAX
, ifla_policy
,
2245 if (tb
[IFLA_IFNAME
])
2246 nla_strlcpy(ifname
, tb
[IFLA_IFNAME
], IFNAMSIZ
);
2251 ifm
= nlmsg_data(nlh
);
2252 if (ifm
->ifi_index
> 0)
2253 dev
= __dev_get_by_index(net
, ifm
->ifi_index
);
2254 else if (tb
[IFLA_IFNAME
])
2255 dev
= __dev_get_by_name(net
, ifname
);
2264 err
= validate_linkmsg(dev
, tb
);
2268 err
= do_setlink(skb
, dev
, ifm
, extack
, tb
, ifname
, 0);
2273 static int rtnl_group_dellink(const struct net
*net
, int group
)
2275 struct net_device
*dev
, *aux
;
2276 LIST_HEAD(list_kill
);
2282 for_each_netdev(net
, dev
) {
2283 if (dev
->group
== group
) {
2284 const struct rtnl_link_ops
*ops
;
2287 ops
= dev
->rtnl_link_ops
;
2288 if (!ops
|| !ops
->dellink
)
2296 for_each_netdev_safe(net
, dev
, aux
) {
2297 if (dev
->group
== group
) {
2298 const struct rtnl_link_ops
*ops
;
2300 ops
= dev
->rtnl_link_ops
;
2301 ops
->dellink(dev
, &list_kill
);
2304 unregister_netdevice_many(&list_kill
);
2309 int rtnl_delete_link(struct net_device
*dev
)
2311 const struct rtnl_link_ops
*ops
;
2312 LIST_HEAD(list_kill
);
2314 ops
= dev
->rtnl_link_ops
;
2315 if (!ops
|| !ops
->dellink
)
2318 ops
->dellink(dev
, &list_kill
);
2319 unregister_netdevice_many(&list_kill
);
2323 EXPORT_SYMBOL_GPL(rtnl_delete_link
);
2325 static int rtnl_dellink(struct sk_buff
*skb
, struct nlmsghdr
*nlh
,
2326 struct netlink_ext_ack
*extack
)
2328 struct net
*net
= sock_net(skb
->sk
);
2329 struct net_device
*dev
;
2330 struct ifinfomsg
*ifm
;
2331 char ifname
[IFNAMSIZ
];
2332 struct nlattr
*tb
[IFLA_MAX
+1];
2335 err
= nlmsg_parse(nlh
, sizeof(*ifm
), tb
, IFLA_MAX
, ifla_policy
, extack
);
2339 if (tb
[IFLA_IFNAME
])
2340 nla_strlcpy(ifname
, tb
[IFLA_IFNAME
], IFNAMSIZ
);
2342 ifm
= nlmsg_data(nlh
);
2343 if (ifm
->ifi_index
> 0)
2344 dev
= __dev_get_by_index(net
, ifm
->ifi_index
);
2345 else if (tb
[IFLA_IFNAME
])
2346 dev
= __dev_get_by_name(net
, ifname
);
2347 else if (tb
[IFLA_GROUP
])
2348 return rtnl_group_dellink(net
, nla_get_u32(tb
[IFLA_GROUP
]));
2355 return rtnl_delete_link(dev
);
2358 int rtnl_configure_link(struct net_device
*dev
, const struct ifinfomsg
*ifm
)
2360 unsigned int old_flags
;
2363 old_flags
= dev
->flags
;
2364 if (ifm
&& (ifm
->ifi_flags
|| ifm
->ifi_change
)) {
2365 err
= __dev_change_flags(dev
, rtnl_dev_combine_flags(dev
, ifm
));
2370 dev
->rtnl_link_state
= RTNL_LINK_INITIALIZED
;
2372 __dev_notify_flags(dev
, old_flags
, ~0U);
2375 EXPORT_SYMBOL(rtnl_configure_link
);
2377 struct net_device
*rtnl_create_link(struct net
*net
,
2378 const char *ifname
, unsigned char name_assign_type
,
2379 const struct rtnl_link_ops
*ops
, struct nlattr
*tb
[])
2381 struct net_device
*dev
;
2382 unsigned int num_tx_queues
= 1;
2383 unsigned int num_rx_queues
= 1;
2385 if (tb
[IFLA_NUM_TX_QUEUES
])
2386 num_tx_queues
= nla_get_u32(tb
[IFLA_NUM_TX_QUEUES
]);
2387 else if (ops
->get_num_tx_queues
)
2388 num_tx_queues
= ops
->get_num_tx_queues();
2390 if (tb
[IFLA_NUM_RX_QUEUES
])
2391 num_rx_queues
= nla_get_u32(tb
[IFLA_NUM_RX_QUEUES
]);
2392 else if (ops
->get_num_rx_queues
)
2393 num_rx_queues
= ops
->get_num_rx_queues();
2395 dev
= alloc_netdev_mqs(ops
->priv_size
, ifname
, name_assign_type
,
2396 ops
->setup
, num_tx_queues
, num_rx_queues
);
2398 return ERR_PTR(-ENOMEM
);
2400 dev_net_set(dev
, net
);
2401 dev
->rtnl_link_ops
= ops
;
2402 dev
->rtnl_link_state
= RTNL_LINK_INITIALIZING
;
2405 dev
->mtu
= nla_get_u32(tb
[IFLA_MTU
]);
2406 if (tb
[IFLA_ADDRESS
]) {
2407 memcpy(dev
->dev_addr
, nla_data(tb
[IFLA_ADDRESS
]),
2408 nla_len(tb
[IFLA_ADDRESS
]));
2409 dev
->addr_assign_type
= NET_ADDR_SET
;
2411 if (tb
[IFLA_BROADCAST
])
2412 memcpy(dev
->broadcast
, nla_data(tb
[IFLA_BROADCAST
]),
2413 nla_len(tb
[IFLA_BROADCAST
]));
2414 if (tb
[IFLA_TXQLEN
])
2415 dev
->tx_queue_len
= nla_get_u32(tb
[IFLA_TXQLEN
]);
2416 if (tb
[IFLA_OPERSTATE
])
2417 set_operstate(dev
, nla_get_u8(tb
[IFLA_OPERSTATE
]));
2418 if (tb
[IFLA_LINKMODE
])
2419 dev
->link_mode
= nla_get_u8(tb
[IFLA_LINKMODE
]);
2421 dev_set_group(dev
, nla_get_u32(tb
[IFLA_GROUP
]));
2425 EXPORT_SYMBOL(rtnl_create_link
);
2427 static int rtnl_group_changelink(const struct sk_buff
*skb
,
2428 struct net
*net
, int group
,
2429 struct ifinfomsg
*ifm
,
2430 struct netlink_ext_ack
*extack
,
2433 struct net_device
*dev
, *aux
;
2436 for_each_netdev_safe(net
, dev
, aux
) {
2437 if (dev
->group
== group
) {
2438 err
= do_setlink(skb
, dev
, ifm
, extack
, tb
, NULL
, 0);
2447 static int rtnl_newlink(struct sk_buff
*skb
, struct nlmsghdr
*nlh
,
2448 struct netlink_ext_ack
*extack
)
2450 struct net
*net
= sock_net(skb
->sk
);
2451 const struct rtnl_link_ops
*ops
;
2452 const struct rtnl_link_ops
*m_ops
= NULL
;
2453 struct net_device
*dev
;
2454 struct net_device
*master_dev
= NULL
;
2455 struct ifinfomsg
*ifm
;
2456 char kind
[MODULE_NAME_LEN
];
2457 char ifname
[IFNAMSIZ
];
2458 struct nlattr
*tb
[IFLA_MAX
+1];
2459 struct nlattr
*linkinfo
[IFLA_INFO_MAX
+1];
2460 unsigned char name_assign_type
= NET_NAME_USER
;
2463 #ifdef CONFIG_MODULES
2466 err
= nlmsg_parse(nlh
, sizeof(*ifm
), tb
, IFLA_MAX
, ifla_policy
, extack
);
2470 if (tb
[IFLA_IFNAME
])
2471 nla_strlcpy(ifname
, tb
[IFLA_IFNAME
], IFNAMSIZ
);
2475 ifm
= nlmsg_data(nlh
);
2476 if (ifm
->ifi_index
> 0)
2477 dev
= __dev_get_by_index(net
, ifm
->ifi_index
);
2480 dev
= __dev_get_by_name(net
, ifname
);
2486 master_dev
= netdev_master_upper_dev_get(dev
);
2488 m_ops
= master_dev
->rtnl_link_ops
;
2491 err
= validate_linkmsg(dev
, tb
);
2495 if (tb
[IFLA_LINKINFO
]) {
2496 err
= nla_parse_nested(linkinfo
, IFLA_INFO_MAX
,
2497 tb
[IFLA_LINKINFO
], ifla_info_policy
,
2502 memset(linkinfo
, 0, sizeof(linkinfo
));
2504 if (linkinfo
[IFLA_INFO_KIND
]) {
2505 nla_strlcpy(kind
, linkinfo
[IFLA_INFO_KIND
], sizeof(kind
));
2506 ops
= rtnl_link_ops_get(kind
);
2513 struct nlattr
*attr
[ops
? ops
->maxtype
+ 1 : 1];
2514 struct nlattr
*slave_attr
[m_ops
? m_ops
->slave_maxtype
+ 1 : 1];
2515 struct nlattr
**data
= NULL
;
2516 struct nlattr
**slave_data
= NULL
;
2517 struct net
*dest_net
, *link_net
= NULL
;
2520 if (ops
->maxtype
&& linkinfo
[IFLA_INFO_DATA
]) {
2521 err
= nla_parse_nested(attr
, ops
->maxtype
,
2522 linkinfo
[IFLA_INFO_DATA
],
2528 if (ops
->validate
) {
2529 err
= ops
->validate(tb
, data
);
2536 if (m_ops
->slave_maxtype
&&
2537 linkinfo
[IFLA_INFO_SLAVE_DATA
]) {
2538 err
= nla_parse_nested(slave_attr
,
2539 m_ops
->slave_maxtype
,
2540 linkinfo
[IFLA_INFO_SLAVE_DATA
],
2541 m_ops
->slave_policy
,
2545 slave_data
= slave_attr
;
2547 if (m_ops
->slave_validate
) {
2548 err
= m_ops
->slave_validate(tb
, slave_data
);
2557 if (nlh
->nlmsg_flags
& NLM_F_EXCL
)
2559 if (nlh
->nlmsg_flags
& NLM_F_REPLACE
)
2562 if (linkinfo
[IFLA_INFO_DATA
]) {
2563 if (!ops
|| ops
!= dev
->rtnl_link_ops
||
2567 err
= ops
->changelink(dev
, tb
, data
);
2570 status
|= DO_SETLINK_NOTIFY
;
2573 if (linkinfo
[IFLA_INFO_SLAVE_DATA
]) {
2574 if (!m_ops
|| !m_ops
->slave_changelink
)
2577 err
= m_ops
->slave_changelink(master_dev
, dev
,
2581 status
|= DO_SETLINK_NOTIFY
;
2584 return do_setlink(skb
, dev
, ifm
, extack
, tb
, ifname
,
2588 if (!(nlh
->nlmsg_flags
& NLM_F_CREATE
)) {
2589 if (ifm
->ifi_index
== 0 && tb
[IFLA_GROUP
])
2590 return rtnl_group_changelink(skb
, net
,
2591 nla_get_u32(tb
[IFLA_GROUP
]),
2596 if (tb
[IFLA_MAP
] || tb
[IFLA_PROTINFO
])
2600 #ifdef CONFIG_MODULES
2603 request_module("rtnl-link-%s", kind
);
2605 ops
= rtnl_link_ops_get(kind
);
2617 snprintf(ifname
, IFNAMSIZ
, "%s%%d", ops
->kind
);
2618 name_assign_type
= NET_NAME_ENUM
;
2621 dest_net
= rtnl_link_get_net(net
, tb
);
2622 if (IS_ERR(dest_net
))
2623 return PTR_ERR(dest_net
);
2626 if (!netlink_ns_capable(skb
, dest_net
->user_ns
, CAP_NET_ADMIN
))
2629 if (tb
[IFLA_LINK_NETNSID
]) {
2630 int id
= nla_get_s32(tb
[IFLA_LINK_NETNSID
]);
2632 link_net
= get_net_ns_by_id(dest_net
, id
);
2638 if (!netlink_ns_capable(skb
, link_net
->user_ns
, CAP_NET_ADMIN
))
2642 dev
= rtnl_create_link(link_net
? : dest_net
, ifname
,
2643 name_assign_type
, ops
, tb
);
2649 dev
->ifindex
= ifm
->ifi_index
;
2652 err
= ops
->newlink(link_net
? : net
, dev
, tb
, data
);
2653 /* Drivers should call free_netdev() in ->destructor
2654 * and unregister it on failure after registration
2655 * so that device could be finally freed in rtnl_unlock.
2658 /* If device is not registered at all, free it now */
2659 if (dev
->reg_state
== NETREG_UNINITIALIZED
)
2664 err
= register_netdevice(dev
);
2670 err
= rtnl_configure_link(dev
, ifm
);
2672 goto out_unregister
;
2674 err
= dev_change_net_namespace(dev
, dest_net
, ifname
);
2676 goto out_unregister
;
2678 if (tb
[IFLA_MASTER
]) {
2679 err
= do_set_master(dev
, nla_get_u32(tb
[IFLA_MASTER
]));
2681 goto out_unregister
;
2690 LIST_HEAD(list_kill
);
2692 ops
->dellink(dev
, &list_kill
);
2693 unregister_netdevice_many(&list_kill
);
2695 unregister_netdevice(dev
);
2701 static int rtnl_getlink(struct sk_buff
*skb
, struct nlmsghdr
*nlh
,
2702 struct netlink_ext_ack
*extack
)
2704 struct net
*net
= sock_net(skb
->sk
);
2705 struct ifinfomsg
*ifm
;
2706 char ifname
[IFNAMSIZ
];
2707 struct nlattr
*tb
[IFLA_MAX
+1];
2708 struct net_device
*dev
= NULL
;
2709 struct sk_buff
*nskb
;
2711 u32 ext_filter_mask
= 0;
2713 err
= nlmsg_parse(nlh
, sizeof(*ifm
), tb
, IFLA_MAX
, ifla_policy
, extack
);
2717 if (tb
[IFLA_IFNAME
])
2718 nla_strlcpy(ifname
, tb
[IFLA_IFNAME
], IFNAMSIZ
);
2720 if (tb
[IFLA_EXT_MASK
])
2721 ext_filter_mask
= nla_get_u32(tb
[IFLA_EXT_MASK
]);
2723 ifm
= nlmsg_data(nlh
);
2724 if (ifm
->ifi_index
> 0)
2725 dev
= __dev_get_by_index(net
, ifm
->ifi_index
);
2726 else if (tb
[IFLA_IFNAME
])
2727 dev
= __dev_get_by_name(net
, ifname
);
2734 nskb
= nlmsg_new(if_nlmsg_size(dev
, ext_filter_mask
), GFP_KERNEL
);
2738 err
= rtnl_fill_ifinfo(nskb
, dev
, RTM_NEWLINK
, NETLINK_CB(skb
).portid
,
2739 nlh
->nlmsg_seq
, 0, 0, ext_filter_mask
);
2741 /* -EMSGSIZE implies BUG in if_nlmsg_size */
2742 WARN_ON(err
== -EMSGSIZE
);
2745 err
= rtnl_unicast(nskb
, net
, NETLINK_CB(skb
).portid
);
2750 static u16
rtnl_calcit(struct sk_buff
*skb
, struct nlmsghdr
*nlh
)
2752 struct net
*net
= sock_net(skb
->sk
);
2753 struct net_device
*dev
;
2754 struct nlattr
*tb
[IFLA_MAX
+1];
2755 u32 ext_filter_mask
= 0;
2756 u16 min_ifinfo_dump_size
= 0;
2759 /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
2760 hdrlen
= nlmsg_len(nlh
) < sizeof(struct ifinfomsg
) ?
2761 sizeof(struct rtgenmsg
) : sizeof(struct ifinfomsg
);
2763 if (nlmsg_parse(nlh
, hdrlen
, tb
, IFLA_MAX
, ifla_policy
, NULL
) >= 0) {
2764 if (tb
[IFLA_EXT_MASK
])
2765 ext_filter_mask
= nla_get_u32(tb
[IFLA_EXT_MASK
]);
2768 if (!ext_filter_mask
)
2769 return NLMSG_GOODSIZE
;
2771 * traverse the list of net devices and compute the minimum
2772 * buffer size based upon the filter mask.
2774 list_for_each_entry(dev
, &net
->dev_base_head
, dev_list
) {
2775 min_ifinfo_dump_size
= max_t(u16
, min_ifinfo_dump_size
,
2780 return nlmsg_total_size(min_ifinfo_dump_size
);
2783 static int rtnl_dump_all(struct sk_buff
*skb
, struct netlink_callback
*cb
)
2786 int s_idx
= cb
->family
;
2790 for (idx
= 1; idx
<= RTNL_FAMILY_MAX
; idx
++) {
2791 int type
= cb
->nlh
->nlmsg_type
-RTM_BASE
;
2792 if (idx
< s_idx
|| idx
== PF_PACKET
)
2794 if (rtnl_msg_handlers
[idx
] == NULL
||
2795 rtnl_msg_handlers
[idx
][type
].dumpit
== NULL
)
2798 memset(&cb
->args
[0], 0, sizeof(cb
->args
));
2802 if (rtnl_msg_handlers
[idx
][type
].dumpit(skb
, cb
))
2810 struct sk_buff
*rtmsg_ifinfo_build_skb(int type
, struct net_device
*dev
,
2811 unsigned int change
, gfp_t flags
)
2813 struct net
*net
= dev_net(dev
);
2814 struct sk_buff
*skb
;
2816 size_t if_info_size
;
2818 skb
= nlmsg_new((if_info_size
= if_nlmsg_size(dev
, 0)), flags
);
2822 err
= rtnl_fill_ifinfo(skb
, dev
, type
, 0, 0, change
, 0, 0);
2824 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
2825 WARN_ON(err
== -EMSGSIZE
);
2832 rtnl_set_sk_err(net
, RTNLGRP_LINK
, err
);
2836 void rtmsg_ifinfo_send(struct sk_buff
*skb
, struct net_device
*dev
, gfp_t flags
)
2838 struct net
*net
= dev_net(dev
);
2840 rtnl_notify(skb
, net
, 0, RTNLGRP_LINK
, NULL
, flags
);
2843 void rtmsg_ifinfo(int type
, struct net_device
*dev
, unsigned int change
,
2846 struct sk_buff
*skb
;
2848 if (dev
->reg_state
!= NETREG_REGISTERED
)
2851 skb
= rtmsg_ifinfo_build_skb(type
, dev
, change
, flags
);
2853 rtmsg_ifinfo_send(skb
, dev
, flags
);
2855 EXPORT_SYMBOL(rtmsg_ifinfo
);
2857 static int nlmsg_populate_fdb_fill(struct sk_buff
*skb
,
2858 struct net_device
*dev
,
2859 u8
*addr
, u16 vid
, u32 pid
, u32 seq
,
2860 int type
, unsigned int flags
,
2861 int nlflags
, u16 ndm_state
)
2863 struct nlmsghdr
*nlh
;
2866 nlh
= nlmsg_put(skb
, pid
, seq
, type
, sizeof(*ndm
), nlflags
);
2870 ndm
= nlmsg_data(nlh
);
2871 ndm
->ndm_family
= AF_BRIDGE
;
2874 ndm
->ndm_flags
= flags
;
2876 ndm
->ndm_ifindex
= dev
->ifindex
;
2877 ndm
->ndm_state
= ndm_state
;
2879 if (nla_put(skb
, NDA_LLADDR
, ETH_ALEN
, addr
))
2880 goto nla_put_failure
;
2882 if (nla_put(skb
, NDA_VLAN
, sizeof(u16
), &vid
))
2883 goto nla_put_failure
;
2885 nlmsg_end(skb
, nlh
);
2889 nlmsg_cancel(skb
, nlh
);
2893 static inline size_t rtnl_fdb_nlmsg_size(void)
2895 return NLMSG_ALIGN(sizeof(struct ndmsg
)) +
2896 nla_total_size(ETH_ALEN
) + /* NDA_LLADDR */
2897 nla_total_size(sizeof(u16
)) + /* NDA_VLAN */
2901 static void rtnl_fdb_notify(struct net_device
*dev
, u8
*addr
, u16 vid
, int type
,
2904 struct net
*net
= dev_net(dev
);
2905 struct sk_buff
*skb
;
2908 skb
= nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC
);
2912 err
= nlmsg_populate_fdb_fill(skb
, dev
, addr
, vid
,
2913 0, 0, type
, NTF_SELF
, 0, ndm_state
);
2919 rtnl_notify(skb
, net
, 0, RTNLGRP_NEIGH
, NULL
, GFP_ATOMIC
);
2922 rtnl_set_sk_err(net
, RTNLGRP_NEIGH
, err
);
2926 * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
2928 int ndo_dflt_fdb_add(struct ndmsg
*ndm
,
2929 struct nlattr
*tb
[],
2930 struct net_device
*dev
,
2931 const unsigned char *addr
, u16 vid
,
2936 /* If aging addresses are supported device will need to
2937 * implement its own handler for this.
2939 if (ndm
->ndm_state
&& !(ndm
->ndm_state
& NUD_PERMANENT
)) {
2940 pr_info("%s: FDB only supports static addresses\n", dev
->name
);
2945 pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev
->name
);
2949 if (is_unicast_ether_addr(addr
) || is_link_local_ether_addr(addr
))
2950 err
= dev_uc_add_excl(dev
, addr
);
2951 else if (is_multicast_ether_addr(addr
))
2952 err
= dev_mc_add_excl(dev
, addr
);
2954 /* Only return duplicate errors if NLM_F_EXCL is set */
2955 if (err
== -EEXIST
&& !(flags
& NLM_F_EXCL
))
2960 EXPORT_SYMBOL(ndo_dflt_fdb_add
);
2962 static int fdb_vid_parse(struct nlattr
*vlan_attr
, u16
*p_vid
)
2967 if (nla_len(vlan_attr
) != sizeof(u16
)) {
2968 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan\n");
2972 vid
= nla_get_u16(vlan_attr
);
2974 if (!vid
|| vid
>= VLAN_VID_MASK
) {
2975 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan id %d\n",
2984 static int rtnl_fdb_add(struct sk_buff
*skb
, struct nlmsghdr
*nlh
,
2985 struct netlink_ext_ack
*extack
)
2987 struct net
*net
= sock_net(skb
->sk
);
2989 struct nlattr
*tb
[NDA_MAX
+1];
2990 struct net_device
*dev
;
2995 err
= nlmsg_parse(nlh
, sizeof(*ndm
), tb
, NDA_MAX
, NULL
, extack
);
2999 ndm
= nlmsg_data(nlh
);
3000 if (ndm
->ndm_ifindex
== 0) {
3001 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n");
3005 dev
= __dev_get_by_index(net
, ndm
->ndm_ifindex
);
3007 pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n");
3011 if (!tb
[NDA_LLADDR
] || nla_len(tb
[NDA_LLADDR
]) != ETH_ALEN
) {
3012 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n");
3016 addr
= nla_data(tb
[NDA_LLADDR
]);
3018 err
= fdb_vid_parse(tb
[NDA_VLAN
], &vid
);
3024 /* Support fdb on master device the net/bridge default case */
3025 if ((!ndm
->ndm_flags
|| ndm
->ndm_flags
& NTF_MASTER
) &&
3026 (dev
->priv_flags
& IFF_BRIDGE_PORT
)) {
3027 struct net_device
*br_dev
= netdev_master_upper_dev_get(dev
);
3028 const struct net_device_ops
*ops
= br_dev
->netdev_ops
;
3030 err
= ops
->ndo_fdb_add(ndm
, tb
, dev
, addr
, vid
,
3035 ndm
->ndm_flags
&= ~NTF_MASTER
;
3038 /* Embedded bridge, macvlan, and any other device support */
3039 if ((ndm
->ndm_flags
& NTF_SELF
)) {
3040 if (dev
->netdev_ops
->ndo_fdb_add
)
3041 err
= dev
->netdev_ops
->ndo_fdb_add(ndm
, tb
, dev
, addr
,
3045 err
= ndo_dflt_fdb_add(ndm
, tb
, dev
, addr
, vid
,
3049 rtnl_fdb_notify(dev
, addr
, vid
, RTM_NEWNEIGH
,
3051 ndm
->ndm_flags
&= ~NTF_SELF
;
3059 * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
3061 int ndo_dflt_fdb_del(struct ndmsg
*ndm
,
3062 struct nlattr
*tb
[],
3063 struct net_device
*dev
,
3064 const unsigned char *addr
, u16 vid
)
3068 /* If aging addresses are supported device will need to
3069 * implement its own handler for this.
3071 if (!(ndm
->ndm_state
& NUD_PERMANENT
)) {
3072 pr_info("%s: FDB only supports static addresses\n", dev
->name
);
3076 if (is_unicast_ether_addr(addr
) || is_link_local_ether_addr(addr
))
3077 err
= dev_uc_del(dev
, addr
);
3078 else if (is_multicast_ether_addr(addr
))
3079 err
= dev_mc_del(dev
, addr
);
3083 EXPORT_SYMBOL(ndo_dflt_fdb_del
);
3085 static int rtnl_fdb_del(struct sk_buff
*skb
, struct nlmsghdr
*nlh
,
3086 struct netlink_ext_ack
*extack
)
3088 struct net
*net
= sock_net(skb
->sk
);
3090 struct nlattr
*tb
[NDA_MAX
+1];
3091 struct net_device
*dev
;
3096 if (!netlink_capable(skb
, CAP_NET_ADMIN
))
3099 err
= nlmsg_parse(nlh
, sizeof(*ndm
), tb
, NDA_MAX
, NULL
, extack
);
3103 ndm
= nlmsg_data(nlh
);
3104 if (ndm
->ndm_ifindex
== 0) {
3105 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n");
3109 dev
= __dev_get_by_index(net
, ndm
->ndm_ifindex
);
3111 pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n");
3115 if (!tb
[NDA_LLADDR
] || nla_len(tb
[NDA_LLADDR
]) != ETH_ALEN
) {
3116 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid address\n");
3120 addr
= nla_data(tb
[NDA_LLADDR
]);
3122 err
= fdb_vid_parse(tb
[NDA_VLAN
], &vid
);
3128 /* Support fdb on master device the net/bridge default case */
3129 if ((!ndm
->ndm_flags
|| ndm
->ndm_flags
& NTF_MASTER
) &&
3130 (dev
->priv_flags
& IFF_BRIDGE_PORT
)) {
3131 struct net_device
*br_dev
= netdev_master_upper_dev_get(dev
);
3132 const struct net_device_ops
*ops
= br_dev
->netdev_ops
;
3134 if (ops
->ndo_fdb_del
)
3135 err
= ops
->ndo_fdb_del(ndm
, tb
, dev
, addr
, vid
);
3140 ndm
->ndm_flags
&= ~NTF_MASTER
;
3143 /* Embedded bridge, macvlan, and any other device support */
3144 if (ndm
->ndm_flags
& NTF_SELF
) {
3145 if (dev
->netdev_ops
->ndo_fdb_del
)
3146 err
= dev
->netdev_ops
->ndo_fdb_del(ndm
, tb
, dev
, addr
,
3149 err
= ndo_dflt_fdb_del(ndm
, tb
, dev
, addr
, vid
);
3152 rtnl_fdb_notify(dev
, addr
, vid
, RTM_DELNEIGH
,
3154 ndm
->ndm_flags
&= ~NTF_SELF
;
3161 static int nlmsg_populate_fdb(struct sk_buff
*skb
,
3162 struct netlink_callback
*cb
,
3163 struct net_device
*dev
,
3165 struct netdev_hw_addr_list
*list
)
3167 struct netdev_hw_addr
*ha
;
3171 portid
= NETLINK_CB(cb
->skb
).portid
;
3172 seq
= cb
->nlh
->nlmsg_seq
;
3174 list_for_each_entry(ha
, &list
->list
, list
) {
3175 if (*idx
< cb
->args
[2])
3178 err
= nlmsg_populate_fdb_fill(skb
, dev
, ha
->addr
, 0,
3180 RTM_NEWNEIGH
, NTF_SELF
,
3181 NLM_F_MULTI
, NUD_PERMANENT
);
3191 * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
3192 * @nlh: netlink message header
3195 * Default netdevice operation to dump the existing unicast address list.
3196 * Returns number of addresses from list put in skb.
3198 int ndo_dflt_fdb_dump(struct sk_buff
*skb
,
3199 struct netlink_callback
*cb
,
3200 struct net_device
*dev
,
3201 struct net_device
*filter_dev
,
3206 netif_addr_lock_bh(dev
);
3207 err
= nlmsg_populate_fdb(skb
, cb
, dev
, idx
, &dev
->uc
);
3210 err
= nlmsg_populate_fdb(skb
, cb
, dev
, idx
, &dev
->mc
);
3212 netif_addr_unlock_bh(dev
);
3215 EXPORT_SYMBOL(ndo_dflt_fdb_dump
);
3217 static int rtnl_fdb_dump(struct sk_buff
*skb
, struct netlink_callback
*cb
)
3219 struct net_device
*dev
;
3220 struct nlattr
*tb
[IFLA_MAX
+1];
3221 struct net_device
*br_dev
= NULL
;
3222 const struct net_device_ops
*ops
= NULL
;
3223 const struct net_device_ops
*cops
= NULL
;
3224 struct ifinfomsg
*ifm
= nlmsg_data(cb
->nlh
);
3225 struct net
*net
= sock_net(skb
->sk
);
3226 struct hlist_head
*head
;
3234 err
= nlmsg_parse(cb
->nlh
, sizeof(struct ifinfomsg
), tb
,
3235 IFLA_MAX
, ifla_policy
, NULL
);
3238 } else if (err
== 0) {
3239 if (tb
[IFLA_MASTER
])
3240 br_idx
= nla_get_u32(tb
[IFLA_MASTER
]);
3243 brport_idx
= ifm
->ifi_index
;
3246 br_dev
= __dev_get_by_index(net
, br_idx
);
3250 ops
= br_dev
->netdev_ops
;
3254 s_idx
= cb
->args
[1];
3256 for (h
= s_h
; h
< NETDEV_HASHENTRIES
; h
++, s_idx
= 0) {
3258 head
= &net
->dev_index_head
[h
];
3259 hlist_for_each_entry(dev
, head
, index_hlist
) {
3261 if (brport_idx
&& (dev
->ifindex
!= brport_idx
))
3264 if (!br_idx
) { /* user did not specify a specific bridge */
3265 if (dev
->priv_flags
& IFF_BRIDGE_PORT
) {
3266 br_dev
= netdev_master_upper_dev_get(dev
);
3267 cops
= br_dev
->netdev_ops
;
3270 if (dev
!= br_dev
&&
3271 !(dev
->priv_flags
& IFF_BRIDGE_PORT
))
3274 if (br_dev
!= netdev_master_upper_dev_get(dev
) &&
3275 !(dev
->priv_flags
& IFF_EBRIDGE
))
3283 if (dev
->priv_flags
& IFF_BRIDGE_PORT
) {
3284 if (cops
&& cops
->ndo_fdb_dump
) {
3285 err
= cops
->ndo_fdb_dump(skb
, cb
,
3288 if (err
== -EMSGSIZE
)
3293 if (dev
->netdev_ops
->ndo_fdb_dump
)
3294 err
= dev
->netdev_ops
->ndo_fdb_dump(skb
, cb
,
3298 err
= ndo_dflt_fdb_dump(skb
, cb
, dev
, NULL
,
3300 if (err
== -EMSGSIZE
)
3305 /* reset fdb offset to 0 for rest of the interfaces */
3321 static int brport_nla_put_flag(struct sk_buff
*skb
, u32 flags
, u32 mask
,
3322 unsigned int attrnum
, unsigned int flag
)
3325 return nla_put_u8(skb
, attrnum
, !!(flags
& flag
));
3329 int ndo_dflt_bridge_getlink(struct sk_buff
*skb
, u32 pid
, u32 seq
,
3330 struct net_device
*dev
, u16 mode
,
3331 u32 flags
, u32 mask
, int nlflags
,
3333 int (*vlan_fill
)(struct sk_buff
*skb
,
3334 struct net_device
*dev
,
3337 struct nlmsghdr
*nlh
;
3338 struct ifinfomsg
*ifm
;
3339 struct nlattr
*br_afspec
;
3340 struct nlattr
*protinfo
;
3341 u8 operstate
= netif_running(dev
) ? dev
->operstate
: IF_OPER_DOWN
;
3342 struct net_device
*br_dev
= netdev_master_upper_dev_get(dev
);
3345 nlh
= nlmsg_put(skb
, pid
, seq
, RTM_NEWLINK
, sizeof(*ifm
), nlflags
);
3349 ifm
= nlmsg_data(nlh
);
3350 ifm
->ifi_family
= AF_BRIDGE
;
3352 ifm
->ifi_type
= dev
->type
;
3353 ifm
->ifi_index
= dev
->ifindex
;
3354 ifm
->ifi_flags
= dev_get_flags(dev
);
3355 ifm
->ifi_change
= 0;
3358 if (nla_put_string(skb
, IFLA_IFNAME
, dev
->name
) ||
3359 nla_put_u32(skb
, IFLA_MTU
, dev
->mtu
) ||
3360 nla_put_u8(skb
, IFLA_OPERSTATE
, operstate
) ||
3362 nla_put_u32(skb
, IFLA_MASTER
, br_dev
->ifindex
)) ||
3364 nla_put(skb
, IFLA_ADDRESS
, dev
->addr_len
, dev
->dev_addr
)) ||
3365 (dev
->ifindex
!= dev_get_iflink(dev
) &&
3366 nla_put_u32(skb
, IFLA_LINK
, dev_get_iflink(dev
))))
3367 goto nla_put_failure
;
3369 br_afspec
= nla_nest_start(skb
, IFLA_AF_SPEC
);
3371 goto nla_put_failure
;
3373 if (nla_put_u16(skb
, IFLA_BRIDGE_FLAGS
, BRIDGE_FLAGS_SELF
)) {
3374 nla_nest_cancel(skb
, br_afspec
);
3375 goto nla_put_failure
;
3378 if (mode
!= BRIDGE_MODE_UNDEF
) {
3379 if (nla_put_u16(skb
, IFLA_BRIDGE_MODE
, mode
)) {
3380 nla_nest_cancel(skb
, br_afspec
);
3381 goto nla_put_failure
;
3385 err
= vlan_fill(skb
, dev
, filter_mask
);
3387 nla_nest_cancel(skb
, br_afspec
);
3388 goto nla_put_failure
;
3391 nla_nest_end(skb
, br_afspec
);
3393 protinfo
= nla_nest_start(skb
, IFLA_PROTINFO
| NLA_F_NESTED
);
3395 goto nla_put_failure
;
3397 if (brport_nla_put_flag(skb
, flags
, mask
,
3398 IFLA_BRPORT_MODE
, BR_HAIRPIN_MODE
) ||
3399 brport_nla_put_flag(skb
, flags
, mask
,
3400 IFLA_BRPORT_GUARD
, BR_BPDU_GUARD
) ||
3401 brport_nla_put_flag(skb
, flags
, mask
,
3402 IFLA_BRPORT_FAST_LEAVE
,
3403 BR_MULTICAST_FAST_LEAVE
) ||
3404 brport_nla_put_flag(skb
, flags
, mask
,
3405 IFLA_BRPORT_PROTECT
, BR_ROOT_BLOCK
) ||
3406 brport_nla_put_flag(skb
, flags
, mask
,
3407 IFLA_BRPORT_LEARNING
, BR_LEARNING
) ||
3408 brport_nla_put_flag(skb
, flags
, mask
,
3409 IFLA_BRPORT_LEARNING_SYNC
, BR_LEARNING_SYNC
) ||
3410 brport_nla_put_flag(skb
, flags
, mask
,
3411 IFLA_BRPORT_UNICAST_FLOOD
, BR_FLOOD
) ||
3412 brport_nla_put_flag(skb
, flags
, mask
,
3413 IFLA_BRPORT_PROXYARP
, BR_PROXYARP
)) {
3414 nla_nest_cancel(skb
, protinfo
);
3415 goto nla_put_failure
;
3418 nla_nest_end(skb
, protinfo
);
3420 nlmsg_end(skb
, nlh
);
3423 nlmsg_cancel(skb
, nlh
);
3424 return err
? err
: -EMSGSIZE
;
3426 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink
);
3428 static int rtnl_bridge_getlink(struct sk_buff
*skb
, struct netlink_callback
*cb
)
3430 struct net
*net
= sock_net(skb
->sk
);
3431 struct net_device
*dev
;
3433 u32 portid
= NETLINK_CB(cb
->skb
).portid
;
3434 u32 seq
= cb
->nlh
->nlmsg_seq
;
3435 u32 filter_mask
= 0;
3438 if (nlmsg_len(cb
->nlh
) > sizeof(struct ifinfomsg
)) {
3439 struct nlattr
*extfilt
;
3441 extfilt
= nlmsg_find_attr(cb
->nlh
, sizeof(struct ifinfomsg
),
3444 if (nla_len(extfilt
) < sizeof(filter_mask
))
3447 filter_mask
= nla_get_u32(extfilt
);
3452 for_each_netdev_rcu(net
, dev
) {
3453 const struct net_device_ops
*ops
= dev
->netdev_ops
;
3454 struct net_device
*br_dev
= netdev_master_upper_dev_get(dev
);
3456 if (br_dev
&& br_dev
->netdev_ops
->ndo_bridge_getlink
) {
3457 if (idx
>= cb
->args
[0]) {
3458 err
= br_dev
->netdev_ops
->ndo_bridge_getlink(
3459 skb
, portid
, seq
, dev
,
3460 filter_mask
, NLM_F_MULTI
);
3461 if (err
< 0 && err
!= -EOPNOTSUPP
) {
3462 if (likely(skb
->len
))
3471 if (ops
->ndo_bridge_getlink
) {
3472 if (idx
>= cb
->args
[0]) {
3473 err
= ops
->ndo_bridge_getlink(skb
, portid
,
3477 if (err
< 0 && err
!= -EOPNOTSUPP
) {
3478 if (likely(skb
->len
))
3495 static inline size_t bridge_nlmsg_size(void)
3497 return NLMSG_ALIGN(sizeof(struct ifinfomsg
))
3498 + nla_total_size(IFNAMSIZ
) /* IFLA_IFNAME */
3499 + nla_total_size(MAX_ADDR_LEN
) /* IFLA_ADDRESS */
3500 + nla_total_size(sizeof(u32
)) /* IFLA_MASTER */
3501 + nla_total_size(sizeof(u32
)) /* IFLA_MTU */
3502 + nla_total_size(sizeof(u32
)) /* IFLA_LINK */
3503 + nla_total_size(sizeof(u32
)) /* IFLA_OPERSTATE */
3504 + nla_total_size(sizeof(u8
)) /* IFLA_PROTINFO */
3505 + nla_total_size(sizeof(struct nlattr
)) /* IFLA_AF_SPEC */
3506 + nla_total_size(sizeof(u16
)) /* IFLA_BRIDGE_FLAGS */
3507 + nla_total_size(sizeof(u16
)); /* IFLA_BRIDGE_MODE */
3510 static int rtnl_bridge_notify(struct net_device
*dev
)
3512 struct net
*net
= dev_net(dev
);
3513 struct sk_buff
*skb
;
3514 int err
= -EOPNOTSUPP
;
3516 if (!dev
->netdev_ops
->ndo_bridge_getlink
)
3519 skb
= nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC
);
3525 err
= dev
->netdev_ops
->ndo_bridge_getlink(skb
, 0, 0, dev
, 0, 0);
3532 rtnl_notify(skb
, net
, 0, RTNLGRP_LINK
, NULL
, GFP_ATOMIC
);
3535 WARN_ON(err
== -EMSGSIZE
);
3538 rtnl_set_sk_err(net
, RTNLGRP_LINK
, err
);
3542 static int rtnl_bridge_setlink(struct sk_buff
*skb
, struct nlmsghdr
*nlh
,
3543 struct netlink_ext_ack
*extack
)
3545 struct net
*net
= sock_net(skb
->sk
);
3546 struct ifinfomsg
*ifm
;
3547 struct net_device
*dev
;
3548 struct nlattr
*br_spec
, *attr
= NULL
;
3549 int rem
, err
= -EOPNOTSUPP
;
3551 bool have_flags
= false;
3553 if (nlmsg_len(nlh
) < sizeof(*ifm
))
3556 ifm
= nlmsg_data(nlh
);
3557 if (ifm
->ifi_family
!= AF_BRIDGE
)
3558 return -EPFNOSUPPORT
;
3560 dev
= __dev_get_by_index(net
, ifm
->ifi_index
);
3562 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3566 br_spec
= nlmsg_find_attr(nlh
, sizeof(struct ifinfomsg
), IFLA_AF_SPEC
);
3568 nla_for_each_nested(attr
, br_spec
, rem
) {
3569 if (nla_type(attr
) == IFLA_BRIDGE_FLAGS
) {
3570 if (nla_len(attr
) < sizeof(flags
))
3574 flags
= nla_get_u16(attr
);
3580 if (!flags
|| (flags
& BRIDGE_FLAGS_MASTER
)) {
3581 struct net_device
*br_dev
= netdev_master_upper_dev_get(dev
);
3583 if (!br_dev
|| !br_dev
->netdev_ops
->ndo_bridge_setlink
) {
3588 err
= br_dev
->netdev_ops
->ndo_bridge_setlink(dev
, nlh
, flags
);
3592 flags
&= ~BRIDGE_FLAGS_MASTER
;
3595 if ((flags
& BRIDGE_FLAGS_SELF
)) {
3596 if (!dev
->netdev_ops
->ndo_bridge_setlink
)
3599 err
= dev
->netdev_ops
->ndo_bridge_setlink(dev
, nlh
,
3602 flags
&= ~BRIDGE_FLAGS_SELF
;
3604 /* Generate event to notify upper layer of bridge
3607 err
= rtnl_bridge_notify(dev
);
3612 memcpy(nla_data(attr
), &flags
, sizeof(flags
));
3617 static int rtnl_bridge_dellink(struct sk_buff
*skb
, struct nlmsghdr
*nlh
,
3618 struct netlink_ext_ack
*extack
)
3620 struct net
*net
= sock_net(skb
->sk
);
3621 struct ifinfomsg
*ifm
;
3622 struct net_device
*dev
;
3623 struct nlattr
*br_spec
, *attr
= NULL
;
3624 int rem
, err
= -EOPNOTSUPP
;
3626 bool have_flags
= false;
3628 if (nlmsg_len(nlh
) < sizeof(*ifm
))
3631 ifm
= nlmsg_data(nlh
);
3632 if (ifm
->ifi_family
!= AF_BRIDGE
)
3633 return -EPFNOSUPPORT
;
3635 dev
= __dev_get_by_index(net
, ifm
->ifi_index
);
3637 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3641 br_spec
= nlmsg_find_attr(nlh
, sizeof(struct ifinfomsg
), IFLA_AF_SPEC
);
3643 nla_for_each_nested(attr
, br_spec
, rem
) {
3644 if (nla_type(attr
) == IFLA_BRIDGE_FLAGS
) {
3645 if (nla_len(attr
) < sizeof(flags
))
3649 flags
= nla_get_u16(attr
);
3655 if (!flags
|| (flags
& BRIDGE_FLAGS_MASTER
)) {
3656 struct net_device
*br_dev
= netdev_master_upper_dev_get(dev
);
3658 if (!br_dev
|| !br_dev
->netdev_ops
->ndo_bridge_dellink
) {
3663 err
= br_dev
->netdev_ops
->ndo_bridge_dellink(dev
, nlh
, flags
);
3667 flags
&= ~BRIDGE_FLAGS_MASTER
;
3670 if ((flags
& BRIDGE_FLAGS_SELF
)) {
3671 if (!dev
->netdev_ops
->ndo_bridge_dellink
)
3674 err
= dev
->netdev_ops
->ndo_bridge_dellink(dev
, nlh
,
3678 flags
&= ~BRIDGE_FLAGS_SELF
;
3680 /* Generate event to notify upper layer of bridge
3683 err
= rtnl_bridge_notify(dev
);
3688 memcpy(nla_data(attr
), &flags
, sizeof(flags
));
3693 static bool stats_attr_valid(unsigned int mask
, int attrid
, int idxattr
)
3695 return (mask
& IFLA_STATS_FILTER_BIT(attrid
)) &&
3696 (!idxattr
|| idxattr
== attrid
);
3699 #define IFLA_OFFLOAD_XSTATS_FIRST (IFLA_OFFLOAD_XSTATS_UNSPEC + 1)
3700 static int rtnl_get_offload_stats_attr_size(int attr_id
)
3703 case IFLA_OFFLOAD_XSTATS_CPU_HIT
:
3704 return sizeof(struct rtnl_link_stats64
);
3710 static int rtnl_get_offload_stats(struct sk_buff
*skb
, struct net_device
*dev
,
3713 struct nlattr
*attr
= NULL
;
3718 if (!(dev
->netdev_ops
&& dev
->netdev_ops
->ndo_has_offload_stats
&&
3719 dev
->netdev_ops
->ndo_get_offload_stats
))
3722 for (attr_id
= IFLA_OFFLOAD_XSTATS_FIRST
;
3723 attr_id
<= IFLA_OFFLOAD_XSTATS_MAX
; attr_id
++) {
3724 if (attr_id
< *prividx
)
3727 size
= rtnl_get_offload_stats_attr_size(attr_id
);
3731 if (!dev
->netdev_ops
->ndo_has_offload_stats(dev
, attr_id
))
3734 attr
= nla_reserve_64bit(skb
, attr_id
, size
,
3735 IFLA_OFFLOAD_XSTATS_UNSPEC
);
3737 goto nla_put_failure
;
3739 attr_data
= nla_data(attr
);
3740 memset(attr_data
, 0, size
);
3741 err
= dev
->netdev_ops
->ndo_get_offload_stats(attr_id
, dev
,
3744 goto get_offload_stats_failure
;
3755 get_offload_stats_failure
:
3760 static int rtnl_get_offload_stats_size(const struct net_device
*dev
)
3766 if (!(dev
->netdev_ops
&& dev
->netdev_ops
->ndo_has_offload_stats
&&
3767 dev
->netdev_ops
->ndo_get_offload_stats
))
3770 for (attr_id
= IFLA_OFFLOAD_XSTATS_FIRST
;
3771 attr_id
<= IFLA_OFFLOAD_XSTATS_MAX
; attr_id
++) {
3772 if (!dev
->netdev_ops
->ndo_has_offload_stats(dev
, attr_id
))
3774 size
= rtnl_get_offload_stats_attr_size(attr_id
);
3775 nla_size
+= nla_total_size_64bit(size
);
3779 nla_size
+= nla_total_size(0);
3784 static int rtnl_fill_statsinfo(struct sk_buff
*skb
, struct net_device
*dev
,
3785 int type
, u32 pid
, u32 seq
, u32 change
,
3786 unsigned int flags
, unsigned int filter_mask
,
3787 int *idxattr
, int *prividx
)
3789 struct if_stats_msg
*ifsm
;
3790 struct nlmsghdr
*nlh
;
3791 struct nlattr
*attr
;
3792 int s_prividx
= *prividx
;
3797 nlh
= nlmsg_put(skb
, pid
, seq
, type
, sizeof(*ifsm
), flags
);
3801 ifsm
= nlmsg_data(nlh
);
3802 ifsm
->ifindex
= dev
->ifindex
;
3803 ifsm
->filter_mask
= filter_mask
;
3805 if (stats_attr_valid(filter_mask
, IFLA_STATS_LINK_64
, *idxattr
)) {
3806 struct rtnl_link_stats64
*sp
;
3808 attr
= nla_reserve_64bit(skb
, IFLA_STATS_LINK_64
,
3809 sizeof(struct rtnl_link_stats64
),
3812 goto nla_put_failure
;
3814 sp
= nla_data(attr
);
3815 dev_get_stats(dev
, sp
);
3818 if (stats_attr_valid(filter_mask
, IFLA_STATS_LINK_XSTATS
, *idxattr
)) {
3819 const struct rtnl_link_ops
*ops
= dev
->rtnl_link_ops
;
3821 if (ops
&& ops
->fill_linkxstats
) {
3822 *idxattr
= IFLA_STATS_LINK_XSTATS
;
3823 attr
= nla_nest_start(skb
,
3824 IFLA_STATS_LINK_XSTATS
);
3826 goto nla_put_failure
;
3828 err
= ops
->fill_linkxstats(skb
, dev
, prividx
, *idxattr
);
3829 nla_nest_end(skb
, attr
);
3831 goto nla_put_failure
;
3836 if (stats_attr_valid(filter_mask
, IFLA_STATS_LINK_XSTATS_SLAVE
,
3838 const struct rtnl_link_ops
*ops
= NULL
;
3839 const struct net_device
*master
;
3841 master
= netdev_master_upper_dev_get(dev
);
3843 ops
= master
->rtnl_link_ops
;
3844 if (ops
&& ops
->fill_linkxstats
) {
3845 *idxattr
= IFLA_STATS_LINK_XSTATS_SLAVE
;
3846 attr
= nla_nest_start(skb
,
3847 IFLA_STATS_LINK_XSTATS_SLAVE
);
3849 goto nla_put_failure
;
3851 err
= ops
->fill_linkxstats(skb
, dev
, prividx
, *idxattr
);
3852 nla_nest_end(skb
, attr
);
3854 goto nla_put_failure
;
3859 if (stats_attr_valid(filter_mask
, IFLA_STATS_LINK_OFFLOAD_XSTATS
,
3861 *idxattr
= IFLA_STATS_LINK_OFFLOAD_XSTATS
;
3862 attr
= nla_nest_start(skb
, IFLA_STATS_LINK_OFFLOAD_XSTATS
);
3864 goto nla_put_failure
;
3866 err
= rtnl_get_offload_stats(skb
, dev
, prividx
);
3867 if (err
== -ENODATA
)
3868 nla_nest_cancel(skb
, attr
);
3870 nla_nest_end(skb
, attr
);
3872 if (err
&& err
!= -ENODATA
)
3873 goto nla_put_failure
;
3877 if (stats_attr_valid(filter_mask
, IFLA_STATS_AF_SPEC
, *idxattr
)) {
3878 struct rtnl_af_ops
*af_ops
;
3880 *idxattr
= IFLA_STATS_AF_SPEC
;
3881 attr
= nla_nest_start(skb
, IFLA_STATS_AF_SPEC
);
3883 goto nla_put_failure
;
3885 list_for_each_entry(af_ops
, &rtnl_af_ops
, list
) {
3886 if (af_ops
->fill_stats_af
) {
3890 af
= nla_nest_start(skb
, af_ops
->family
);
3892 goto nla_put_failure
;
3894 err
= af_ops
->fill_stats_af(skb
, dev
);
3896 if (err
== -ENODATA
)
3897 nla_nest_cancel(skb
, af
);
3899 goto nla_put_failure
;
3901 nla_nest_end(skb
, af
);
3905 nla_nest_end(skb
, attr
);
3910 nlmsg_end(skb
, nlh
);
3915 /* not a multi message or no progress mean a real error */
3916 if (!(flags
& NLM_F_MULTI
) || s_prividx
== *prividx
)
3917 nlmsg_cancel(skb
, nlh
);
3919 nlmsg_end(skb
, nlh
);
3924 static size_t if_nlmsg_stats_size(const struct net_device
*dev
,
3929 if (stats_attr_valid(filter_mask
, IFLA_STATS_LINK_64
, 0))
3930 size
+= nla_total_size_64bit(sizeof(struct rtnl_link_stats64
));
3932 if (stats_attr_valid(filter_mask
, IFLA_STATS_LINK_XSTATS
, 0)) {
3933 const struct rtnl_link_ops
*ops
= dev
->rtnl_link_ops
;
3934 int attr
= IFLA_STATS_LINK_XSTATS
;
3936 if (ops
&& ops
->get_linkxstats_size
) {
3937 size
+= nla_total_size(ops
->get_linkxstats_size(dev
,
3939 /* for IFLA_STATS_LINK_XSTATS */
3940 size
+= nla_total_size(0);
3944 if (stats_attr_valid(filter_mask
, IFLA_STATS_LINK_XSTATS_SLAVE
, 0)) {
3945 struct net_device
*_dev
= (struct net_device
*)dev
;
3946 const struct rtnl_link_ops
*ops
= NULL
;
3947 const struct net_device
*master
;
3949 /* netdev_master_upper_dev_get can't take const */
3950 master
= netdev_master_upper_dev_get(_dev
);
3952 ops
= master
->rtnl_link_ops
;
3953 if (ops
&& ops
->get_linkxstats_size
) {
3954 int attr
= IFLA_STATS_LINK_XSTATS_SLAVE
;
3956 size
+= nla_total_size(ops
->get_linkxstats_size(dev
,
3958 /* for IFLA_STATS_LINK_XSTATS_SLAVE */
3959 size
+= nla_total_size(0);
3963 if (stats_attr_valid(filter_mask
, IFLA_STATS_LINK_OFFLOAD_XSTATS
, 0))
3964 size
+= rtnl_get_offload_stats_size(dev
);
3966 if (stats_attr_valid(filter_mask
, IFLA_STATS_AF_SPEC
, 0)) {
3967 struct rtnl_af_ops
*af_ops
;
3969 /* for IFLA_STATS_AF_SPEC */
3970 size
+= nla_total_size(0);
3972 list_for_each_entry(af_ops
, &rtnl_af_ops
, list
) {
3973 if (af_ops
->get_stats_af_size
) {
3974 size
+= nla_total_size(
3975 af_ops
->get_stats_af_size(dev
));
3978 size
+= nla_total_size(0);
3986 static int rtnl_stats_get(struct sk_buff
*skb
, struct nlmsghdr
*nlh
,
3987 struct netlink_ext_ack
*extack
)
3989 struct net
*net
= sock_net(skb
->sk
);
3990 struct net_device
*dev
= NULL
;
3991 int idxattr
= 0, prividx
= 0;
3992 struct if_stats_msg
*ifsm
;
3993 struct sk_buff
*nskb
;
3997 if (nlmsg_len(nlh
) < sizeof(*ifsm
))
4000 ifsm
= nlmsg_data(nlh
);
4001 if (ifsm
->ifindex
> 0)
4002 dev
= __dev_get_by_index(net
, ifsm
->ifindex
);
4009 filter_mask
= ifsm
->filter_mask
;
4013 nskb
= nlmsg_new(if_nlmsg_stats_size(dev
, filter_mask
), GFP_KERNEL
);
4017 err
= rtnl_fill_statsinfo(nskb
, dev
, RTM_NEWSTATS
,
4018 NETLINK_CB(skb
).portid
, nlh
->nlmsg_seq
, 0,
4019 0, filter_mask
, &idxattr
, &prividx
);
4021 /* -EMSGSIZE implies BUG in if_nlmsg_stats_size */
4022 WARN_ON(err
== -EMSGSIZE
);
4025 err
= rtnl_unicast(nskb
, net
, NETLINK_CB(skb
).portid
);
4031 static int rtnl_stats_dump(struct sk_buff
*skb
, struct netlink_callback
*cb
)
4033 int h
, s_h
, err
, s_idx
, s_idxattr
, s_prividx
;
4034 struct net
*net
= sock_net(skb
->sk
);
4035 unsigned int flags
= NLM_F_MULTI
;
4036 struct if_stats_msg
*ifsm
;
4037 struct hlist_head
*head
;
4038 struct net_device
*dev
;
4039 u32 filter_mask
= 0;
4043 s_idx
= cb
->args
[1];
4044 s_idxattr
= cb
->args
[2];
4045 s_prividx
= cb
->args
[3];
4047 cb
->seq
= net
->dev_base_seq
;
4049 if (nlmsg_len(cb
->nlh
) < sizeof(*ifsm
))
4052 ifsm
= nlmsg_data(cb
->nlh
);
4053 filter_mask
= ifsm
->filter_mask
;
4057 for (h
= s_h
; h
< NETDEV_HASHENTRIES
; h
++, s_idx
= 0) {
4059 head
= &net
->dev_index_head
[h
];
4060 hlist_for_each_entry(dev
, head
, index_hlist
) {
4063 err
= rtnl_fill_statsinfo(skb
, dev
, RTM_NEWSTATS
,
4064 NETLINK_CB(cb
->skb
).portid
,
4065 cb
->nlh
->nlmsg_seq
, 0,
4067 &s_idxattr
, &s_prividx
);
4068 /* If we ran out of room on the first message,
4071 WARN_ON((err
== -EMSGSIZE
) && (skb
->len
== 0));
4077 nl_dump_check_consistent(cb
, nlmsg_hdr(skb
));
4083 cb
->args
[3] = s_prividx
;
4084 cb
->args
[2] = s_idxattr
;
4091 /* Process one rtnetlink message. */
4093 static int rtnetlink_rcv_msg(struct sk_buff
*skb
, struct nlmsghdr
*nlh
,
4094 struct netlink_ext_ack
*extack
)
4096 struct net
*net
= sock_net(skb
->sk
);
4097 rtnl_doit_func doit
;
4103 type
= nlh
->nlmsg_type
;
4109 /* All the messages must have at least 1 byte length */
4110 if (nlmsg_len(nlh
) < sizeof(struct rtgenmsg
))
4113 family
= ((struct rtgenmsg
*)nlmsg_data(nlh
))->rtgen_family
;
4116 if (kind
!= 2 && !netlink_net_capable(skb
, CAP_NET_ADMIN
))
4119 if (kind
== 2 && nlh
->nlmsg_flags
&NLM_F_DUMP
) {
4121 rtnl_dumpit_func dumpit
;
4122 rtnl_calcit_func calcit
;
4123 u16 min_dump_alloc
= 0;
4125 dumpit
= rtnl_get_dumpit(family
, type
);
4128 calcit
= rtnl_get_calcit(family
, type
);
4130 min_dump_alloc
= calcit(skb
, nlh
);
4135 struct netlink_dump_control c
= {
4137 .min_dump_alloc
= min_dump_alloc
,
4139 err
= netlink_dump_start(rtnl
, skb
, nlh
, &c
);
4145 doit
= rtnl_get_doit(family
, type
);
4149 return doit(skb
, nlh
, extack
);
4152 static void rtnetlink_rcv(struct sk_buff
*skb
)
4155 netlink_rcv_skb(skb
, &rtnetlink_rcv_msg
);
4159 static int rtnetlink_event(struct notifier_block
*this, unsigned long event
, void *ptr
)
4161 struct net_device
*dev
= netdev_notifier_info_to_dev(ptr
);
4165 case NETDEV_CHANGENAME
:
4166 case NETDEV_FEAT_CHANGE
:
4167 case NETDEV_BONDING_FAILOVER
:
4168 case NETDEV_NOTIFY_PEERS
:
4169 case NETDEV_RESEND_IGMP
:
4170 case NETDEV_CHANGEINFODATA
:
4171 rtmsg_ifinfo(RTM_NEWLINK
, dev
, 0, GFP_KERNEL
);
4179 static struct notifier_block rtnetlink_dev_notifier
= {
4180 .notifier_call
= rtnetlink_event
,
4184 static int __net_init
rtnetlink_net_init(struct net
*net
)
4187 struct netlink_kernel_cfg cfg
= {
4188 .groups
= RTNLGRP_MAX
,
4189 .input
= rtnetlink_rcv
,
4190 .cb_mutex
= &rtnl_mutex
,
4191 .flags
= NL_CFG_F_NONROOT_RECV
,
4194 sk
= netlink_kernel_create(net
, NETLINK_ROUTE
, &cfg
);
4201 static void __net_exit
rtnetlink_net_exit(struct net
*net
)
4203 netlink_kernel_release(net
->rtnl
);
4207 static struct pernet_operations rtnetlink_net_ops
= {
4208 .init
= rtnetlink_net_init
,
4209 .exit
= rtnetlink_net_exit
,
4212 void __init
rtnetlink_init(void)
4214 if (register_pernet_subsys(&rtnetlink_net_ops
))
4215 panic("rtnetlink_init: cannot initialize rtnetlink\n");
4217 register_netdevice_notifier(&rtnetlink_dev_notifier
);
4219 rtnl_register(PF_UNSPEC
, RTM_GETLINK
, rtnl_getlink
,
4220 rtnl_dump_ifinfo
, rtnl_calcit
);
4221 rtnl_register(PF_UNSPEC
, RTM_SETLINK
, rtnl_setlink
, NULL
, NULL
);
4222 rtnl_register(PF_UNSPEC
, RTM_NEWLINK
, rtnl_newlink
, NULL
, NULL
);
4223 rtnl_register(PF_UNSPEC
, RTM_DELLINK
, rtnl_dellink
, NULL
, NULL
);
4225 rtnl_register(PF_UNSPEC
, RTM_GETADDR
, NULL
, rtnl_dump_all
, NULL
);
4226 rtnl_register(PF_UNSPEC
, RTM_GETROUTE
, NULL
, rtnl_dump_all
, NULL
);
4227 rtnl_register(PF_UNSPEC
, RTM_GETNETCONF
, NULL
, rtnl_dump_all
, NULL
);
4229 rtnl_register(PF_BRIDGE
, RTM_NEWNEIGH
, rtnl_fdb_add
, NULL
, NULL
);
4230 rtnl_register(PF_BRIDGE
, RTM_DELNEIGH
, rtnl_fdb_del
, NULL
, NULL
);
4231 rtnl_register(PF_BRIDGE
, RTM_GETNEIGH
, NULL
, rtnl_fdb_dump
, NULL
);
4233 rtnl_register(PF_BRIDGE
, RTM_GETLINK
, NULL
, rtnl_bridge_getlink
, NULL
);
4234 rtnl_register(PF_BRIDGE
, RTM_DELLINK
, rtnl_bridge_dellink
, NULL
, NULL
);
4235 rtnl_register(PF_BRIDGE
, RTM_SETLINK
, rtnl_bridge_setlink
, NULL
, NULL
);
4237 rtnl_register(PF_UNSPEC
, RTM_GETSTATS
, rtnl_stats_get
, rtnl_stats_dump
,