2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
6 * Routing netlink socket interface: protocol independent part.
8 * Authors: Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
10 * This program is free software; you can redistribute it and/or
11 * modify it under the terms of the GNU General Public License
12 * as published by the Free Software Foundation; either version
13 * 2 of the License, or (at your option) any later version.
16 * Vitaly E. Lavrov RTA_OK arithmetics was wrong.
19 #include <linux/errno.h>
20 #include <linux/module.h>
21 #include <linux/types.h>
22 #include <linux/socket.h>
23 #include <linux/kernel.h>
24 #include <linux/timer.h>
25 #include <linux/string.h>
26 #include <linux/sockios.h>
27 #include <linux/net.h>
28 #include <linux/fcntl.h>
30 #include <linux/slab.h>
31 #include <linux/interrupt.h>
32 #include <linux/capability.h>
33 #include <linux/skbuff.h>
34 #include <linux/init.h>
35 #include <linux/security.h>
36 #include <linux/mutex.h>
37 #include <linux/if_addr.h>
38 #include <linux/if_bridge.h>
39 #include <linux/pci.h>
40 #include <linux/etherdevice.h>
42 #include <asm/uaccess.h>
44 #include <linux/inet.h>
45 #include <linux/netdevice.h>
47 #include <net/protocol.h>
49 #include <net/route.h>
52 #include <net/pkt_sched.h>
53 #include <net/fib_rules.h>
54 #include <net/rtnetlink.h>
55 #include <net/net_namespace.h>
59 rtnl_dumpit_func dumpit
;
60 rtnl_calcit_func calcit
;
63 static DEFINE_MUTEX(rtnl_mutex
);
67 mutex_lock(&rtnl_mutex
);
69 EXPORT_SYMBOL(rtnl_lock
);
71 void __rtnl_unlock(void)
73 mutex_unlock(&rtnl_mutex
);
76 void rtnl_unlock(void)
78 /* This fellow will unlock it for us. */
81 EXPORT_SYMBOL(rtnl_unlock
);
83 int rtnl_trylock(void)
85 return mutex_trylock(&rtnl_mutex
);
87 EXPORT_SYMBOL(rtnl_trylock
);
89 int rtnl_is_locked(void)
91 return mutex_is_locked(&rtnl_mutex
);
93 EXPORT_SYMBOL(rtnl_is_locked
);
95 #ifdef CONFIG_PROVE_LOCKING
96 int lockdep_rtnl_is_held(void)
98 return lockdep_is_held(&rtnl_mutex
);
100 EXPORT_SYMBOL(lockdep_rtnl_is_held
);
101 #endif /* #ifdef CONFIG_PROVE_LOCKING */
103 static struct rtnl_link
*rtnl_msg_handlers
[RTNL_FAMILY_MAX
+ 1];
105 static inline int rtm_msgindex(int msgtype
)
107 int msgindex
= msgtype
- RTM_BASE
;
110 * msgindex < 0 implies someone tried to register a netlink
111 * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
112 * the message type has not been added to linux/rtnetlink.h
114 BUG_ON(msgindex
< 0 || msgindex
>= RTM_NR_MSGTYPES
);
119 static rtnl_doit_func
rtnl_get_doit(int protocol
, int msgindex
)
121 struct rtnl_link
*tab
;
123 if (protocol
<= RTNL_FAMILY_MAX
)
124 tab
= rtnl_msg_handlers
[protocol
];
128 if (tab
== NULL
|| tab
[msgindex
].doit
== NULL
)
129 tab
= rtnl_msg_handlers
[PF_UNSPEC
];
131 return tab
[msgindex
].doit
;
134 static rtnl_dumpit_func
rtnl_get_dumpit(int protocol
, int msgindex
)
136 struct rtnl_link
*tab
;
138 if (protocol
<= RTNL_FAMILY_MAX
)
139 tab
= rtnl_msg_handlers
[protocol
];
143 if (tab
== NULL
|| tab
[msgindex
].dumpit
== NULL
)
144 tab
= rtnl_msg_handlers
[PF_UNSPEC
];
146 return tab
[msgindex
].dumpit
;
149 static rtnl_calcit_func
rtnl_get_calcit(int protocol
, int msgindex
)
151 struct rtnl_link
*tab
;
153 if (protocol
<= RTNL_FAMILY_MAX
)
154 tab
= rtnl_msg_handlers
[protocol
];
158 if (tab
== NULL
|| tab
[msgindex
].calcit
== NULL
)
159 tab
= rtnl_msg_handlers
[PF_UNSPEC
];
161 return tab
[msgindex
].calcit
;
165 * __rtnl_register - Register a rtnetlink message type
166 * @protocol: Protocol family or PF_UNSPEC
167 * @msgtype: rtnetlink message type
168 * @doit: Function pointer called for each request message
169 * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
170 * @calcit: Function pointer to calc size of dump message
172 * Registers the specified function pointers (at least one of them has
173 * to be non-NULL) to be called whenever a request message for the
174 * specified protocol family and message type is received.
176 * The special protocol family PF_UNSPEC may be used to define fallback
177 * function pointers for the case when no entry for the specific protocol
180 * Returns 0 on success or a negative error code.
182 int __rtnl_register(int protocol
, int msgtype
,
183 rtnl_doit_func doit
, rtnl_dumpit_func dumpit
,
184 rtnl_calcit_func calcit
)
186 struct rtnl_link
*tab
;
189 BUG_ON(protocol
< 0 || protocol
> RTNL_FAMILY_MAX
);
190 msgindex
= rtm_msgindex(msgtype
);
192 tab
= rtnl_msg_handlers
[protocol
];
194 tab
= kcalloc(RTM_NR_MSGTYPES
, sizeof(*tab
), GFP_KERNEL
);
198 rtnl_msg_handlers
[protocol
] = tab
;
202 tab
[msgindex
].doit
= doit
;
205 tab
[msgindex
].dumpit
= dumpit
;
208 tab
[msgindex
].calcit
= calcit
;
212 EXPORT_SYMBOL_GPL(__rtnl_register
);
215 * rtnl_register - Register a rtnetlink message type
217 * Identical to __rtnl_register() but panics on failure. This is useful
218 * as failure of this function is very unlikely, it can only happen due
219 * to lack of memory when allocating the chain to store all message
220 * handlers for a protocol. Meant for use in init functions where lack
221 * of memory implies no sense in continuing.
223 void rtnl_register(int protocol
, int msgtype
,
224 rtnl_doit_func doit
, rtnl_dumpit_func dumpit
,
225 rtnl_calcit_func calcit
)
227 if (__rtnl_register(protocol
, msgtype
, doit
, dumpit
, calcit
) < 0)
228 panic("Unable to register rtnetlink message handler, "
229 "protocol = %d, message type = %d\n",
232 EXPORT_SYMBOL_GPL(rtnl_register
);
235 * rtnl_unregister - Unregister a rtnetlink message type
236 * @protocol: Protocol family or PF_UNSPEC
237 * @msgtype: rtnetlink message type
239 * Returns 0 on success or a negative error code.
241 int rtnl_unregister(int protocol
, int msgtype
)
245 BUG_ON(protocol
< 0 || protocol
> RTNL_FAMILY_MAX
);
246 msgindex
= rtm_msgindex(msgtype
);
248 if (rtnl_msg_handlers
[protocol
] == NULL
)
251 rtnl_msg_handlers
[protocol
][msgindex
].doit
= NULL
;
252 rtnl_msg_handlers
[protocol
][msgindex
].dumpit
= NULL
;
256 EXPORT_SYMBOL_GPL(rtnl_unregister
);
259 * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
260 * @protocol : Protocol family or PF_UNSPEC
262 * Identical to calling rtnl_unregster() for all registered message types
263 * of a certain protocol family.
265 void rtnl_unregister_all(int protocol
)
267 BUG_ON(protocol
< 0 || protocol
> RTNL_FAMILY_MAX
);
269 kfree(rtnl_msg_handlers
[protocol
]);
270 rtnl_msg_handlers
[protocol
] = NULL
;
272 EXPORT_SYMBOL_GPL(rtnl_unregister_all
);
274 static LIST_HEAD(link_ops
);
276 static const struct rtnl_link_ops
*rtnl_link_ops_get(const char *kind
)
278 const struct rtnl_link_ops
*ops
;
280 list_for_each_entry(ops
, &link_ops
, list
) {
281 if (!strcmp(ops
->kind
, kind
))
288 * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
289 * @ops: struct rtnl_link_ops * to register
291 * The caller must hold the rtnl_mutex. This function should be used
292 * by drivers that create devices during module initialization. It
293 * must be called before registering the devices.
295 * Returns 0 on success or a negative error code.
297 int __rtnl_link_register(struct rtnl_link_ops
*ops
)
299 if (rtnl_link_ops_get(ops
->kind
))
303 ops
->dellink
= unregister_netdevice_queue
;
305 list_add_tail(&ops
->list
, &link_ops
);
308 EXPORT_SYMBOL_GPL(__rtnl_link_register
);
311 * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
312 * @ops: struct rtnl_link_ops * to register
314 * Returns 0 on success or a negative error code.
316 int rtnl_link_register(struct rtnl_link_ops
*ops
)
321 err
= __rtnl_link_register(ops
);
325 EXPORT_SYMBOL_GPL(rtnl_link_register
);
327 static void __rtnl_kill_links(struct net
*net
, struct rtnl_link_ops
*ops
)
329 struct net_device
*dev
;
330 LIST_HEAD(list_kill
);
332 for_each_netdev(net
, dev
) {
333 if (dev
->rtnl_link_ops
== ops
)
334 ops
->dellink(dev
, &list_kill
);
336 unregister_netdevice_many(&list_kill
);
340 * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
341 * @ops: struct rtnl_link_ops * to unregister
343 * The caller must hold the rtnl_mutex.
345 void __rtnl_link_unregister(struct rtnl_link_ops
*ops
)
350 __rtnl_kill_links(net
, ops
);
352 list_del(&ops
->list
);
354 EXPORT_SYMBOL_GPL(__rtnl_link_unregister
);
357 * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
358 * @ops: struct rtnl_link_ops * to unregister
360 void rtnl_link_unregister(struct rtnl_link_ops
*ops
)
363 __rtnl_link_unregister(ops
);
366 EXPORT_SYMBOL_GPL(rtnl_link_unregister
);
368 static size_t rtnl_link_get_size(const struct net_device
*dev
)
370 const struct rtnl_link_ops
*ops
= dev
->rtnl_link_ops
;
376 size
= nla_total_size(sizeof(struct nlattr
)) + /* IFLA_LINKINFO */
377 nla_total_size(strlen(ops
->kind
) + 1); /* IFLA_INFO_KIND */
380 /* IFLA_INFO_DATA + nested data */
381 size
+= nla_total_size(sizeof(struct nlattr
)) +
384 if (ops
->get_xstats_size
)
385 /* IFLA_INFO_XSTATS */
386 size
+= nla_total_size(ops
->get_xstats_size(dev
));
391 static LIST_HEAD(rtnl_af_ops
);
393 static const struct rtnl_af_ops
*rtnl_af_lookup(const int family
)
395 const struct rtnl_af_ops
*ops
;
397 list_for_each_entry(ops
, &rtnl_af_ops
, list
) {
398 if (ops
->family
== family
)
406 * __rtnl_af_register - Register rtnl_af_ops with rtnetlink.
407 * @ops: struct rtnl_af_ops * to register
409 * The caller must hold the rtnl_mutex.
411 * Returns 0 on success or a negative error code.
413 int __rtnl_af_register(struct rtnl_af_ops
*ops
)
415 list_add_tail(&ops
->list
, &rtnl_af_ops
);
418 EXPORT_SYMBOL_GPL(__rtnl_af_register
);
421 * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
422 * @ops: struct rtnl_af_ops * to register
424 * Returns 0 on success or a negative error code.
426 int rtnl_af_register(struct rtnl_af_ops
*ops
)
431 err
= __rtnl_af_register(ops
);
435 EXPORT_SYMBOL_GPL(rtnl_af_register
);
438 * __rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
439 * @ops: struct rtnl_af_ops * to unregister
441 * The caller must hold the rtnl_mutex.
443 void __rtnl_af_unregister(struct rtnl_af_ops
*ops
)
445 list_del(&ops
->list
);
447 EXPORT_SYMBOL_GPL(__rtnl_af_unregister
);
450 * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
451 * @ops: struct rtnl_af_ops * to unregister
453 void rtnl_af_unregister(struct rtnl_af_ops
*ops
)
456 __rtnl_af_unregister(ops
);
459 EXPORT_SYMBOL_GPL(rtnl_af_unregister
);
461 static size_t rtnl_link_get_af_size(const struct net_device
*dev
)
463 struct rtnl_af_ops
*af_ops
;
467 size
= nla_total_size(sizeof(struct nlattr
));
469 list_for_each_entry(af_ops
, &rtnl_af_ops
, list
) {
470 if (af_ops
->get_link_af_size
) {
471 /* AF_* + nested data */
472 size
+= nla_total_size(sizeof(struct nlattr
)) +
473 af_ops
->get_link_af_size(dev
);
480 static int rtnl_link_fill(struct sk_buff
*skb
, const struct net_device
*dev
)
482 const struct rtnl_link_ops
*ops
= dev
->rtnl_link_ops
;
483 struct nlattr
*linkinfo
, *data
;
486 linkinfo
= nla_nest_start(skb
, IFLA_LINKINFO
);
487 if (linkinfo
== NULL
)
490 if (nla_put_string(skb
, IFLA_INFO_KIND
, ops
->kind
) < 0)
491 goto err_cancel_link
;
492 if (ops
->fill_xstats
) {
493 err
= ops
->fill_xstats(skb
, dev
);
495 goto err_cancel_link
;
497 if (ops
->fill_info
) {
498 data
= nla_nest_start(skb
, IFLA_INFO_DATA
);
500 goto err_cancel_link
;
501 err
= ops
->fill_info(skb
, dev
);
503 goto err_cancel_data
;
504 nla_nest_end(skb
, data
);
507 nla_nest_end(skb
, linkinfo
);
511 nla_nest_cancel(skb
, data
);
513 nla_nest_cancel(skb
, linkinfo
);
518 static const int rtm_min
[RTM_NR_FAMILIES
] =
520 [RTM_FAM(RTM_NEWLINK
)] = NLMSG_LENGTH(sizeof(struct ifinfomsg
)),
521 [RTM_FAM(RTM_NEWADDR
)] = NLMSG_LENGTH(sizeof(struct ifaddrmsg
)),
522 [RTM_FAM(RTM_NEWROUTE
)] = NLMSG_LENGTH(sizeof(struct rtmsg
)),
523 [RTM_FAM(RTM_NEWRULE
)] = NLMSG_LENGTH(sizeof(struct fib_rule_hdr
)),
524 [RTM_FAM(RTM_NEWQDISC
)] = NLMSG_LENGTH(sizeof(struct tcmsg
)),
525 [RTM_FAM(RTM_NEWTCLASS
)] = NLMSG_LENGTH(sizeof(struct tcmsg
)),
526 [RTM_FAM(RTM_NEWTFILTER
)] = NLMSG_LENGTH(sizeof(struct tcmsg
)),
527 [RTM_FAM(RTM_NEWACTION
)] = NLMSG_LENGTH(sizeof(struct tcamsg
)),
528 [RTM_FAM(RTM_GETMULTICAST
)] = NLMSG_LENGTH(sizeof(struct rtgenmsg
)),
529 [RTM_FAM(RTM_GETANYCAST
)] = NLMSG_LENGTH(sizeof(struct rtgenmsg
)),
532 static const int rta_max
[RTM_NR_FAMILIES
] =
534 [RTM_FAM(RTM_NEWLINK
)] = IFLA_MAX
,
535 [RTM_FAM(RTM_NEWADDR
)] = IFA_MAX
,
536 [RTM_FAM(RTM_NEWROUTE
)] = RTA_MAX
,
537 [RTM_FAM(RTM_NEWRULE
)] = FRA_MAX
,
538 [RTM_FAM(RTM_NEWQDISC
)] = TCA_MAX
,
539 [RTM_FAM(RTM_NEWTCLASS
)] = TCA_MAX
,
540 [RTM_FAM(RTM_NEWTFILTER
)] = TCA_MAX
,
541 [RTM_FAM(RTM_NEWACTION
)] = TCAA_MAX
,
544 int rtnetlink_send(struct sk_buff
*skb
, struct net
*net
, u32 pid
, unsigned int group
, int echo
)
546 struct sock
*rtnl
= net
->rtnl
;
549 NETLINK_CB(skb
).dst_group
= group
;
551 atomic_inc(&skb
->users
);
552 netlink_broadcast(rtnl
, skb
, pid
, group
, GFP_KERNEL
);
554 err
= netlink_unicast(rtnl
, skb
, pid
, MSG_DONTWAIT
);
558 int rtnl_unicast(struct sk_buff
*skb
, struct net
*net
, u32 pid
)
560 struct sock
*rtnl
= net
->rtnl
;
562 return nlmsg_unicast(rtnl
, skb
, pid
);
564 EXPORT_SYMBOL(rtnl_unicast
);
566 void rtnl_notify(struct sk_buff
*skb
, struct net
*net
, u32 pid
, u32 group
,
567 struct nlmsghdr
*nlh
, gfp_t flags
)
569 struct sock
*rtnl
= net
->rtnl
;
573 report
= nlmsg_report(nlh
);
575 nlmsg_notify(rtnl
, skb
, pid
, group
, report
, flags
);
577 EXPORT_SYMBOL(rtnl_notify
);
579 void rtnl_set_sk_err(struct net
*net
, u32 group
, int error
)
581 struct sock
*rtnl
= net
->rtnl
;
583 netlink_set_err(rtnl
, 0, group
, error
);
585 EXPORT_SYMBOL(rtnl_set_sk_err
);
587 int rtnetlink_put_metrics(struct sk_buff
*skb
, u32
*metrics
)
592 mx
= nla_nest_start(skb
, RTA_METRICS
);
596 for (i
= 0; i
< RTAX_MAX
; i
++) {
599 if (nla_put_u32(skb
, i
+1, metrics
[i
]))
600 goto nla_put_failure
;
605 nla_nest_cancel(skb
, mx
);
609 return nla_nest_end(skb
, mx
);
612 nla_nest_cancel(skb
, mx
);
615 EXPORT_SYMBOL(rtnetlink_put_metrics
);
617 int rtnl_put_cacheinfo(struct sk_buff
*skb
, struct dst_entry
*dst
, u32 id
,
618 long expires
, u32 error
)
620 struct rta_cacheinfo ci
= {
621 .rta_lastuse
= jiffies_delta_to_clock_t(jiffies
- dst
->lastuse
),
622 .rta_used
= dst
->__use
,
623 .rta_clntref
= atomic_read(&(dst
->__refcnt
)),
631 clock
= jiffies_to_clock_t(abs(expires
));
632 clock
= min_t(unsigned long, clock
, INT_MAX
);
633 ci
.rta_expires
= (expires
> 0) ? clock
: -clock
;
635 return nla_put(skb
, RTA_CACHEINFO
, sizeof(ci
), &ci
);
637 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo
);
639 static void set_operstate(struct net_device
*dev
, unsigned char transition
)
641 unsigned char operstate
= dev
->operstate
;
643 switch (transition
) {
645 if ((operstate
== IF_OPER_DORMANT
||
646 operstate
== IF_OPER_UNKNOWN
) &&
648 operstate
= IF_OPER_UP
;
651 case IF_OPER_DORMANT
:
652 if (operstate
== IF_OPER_UP
||
653 operstate
== IF_OPER_UNKNOWN
)
654 operstate
= IF_OPER_DORMANT
;
658 if (dev
->operstate
!= operstate
) {
659 write_lock_bh(&dev_base_lock
);
660 dev
->operstate
= operstate
;
661 write_unlock_bh(&dev_base_lock
);
662 netdev_state_change(dev
);
666 static unsigned int rtnl_dev_get_flags(const struct net_device
*dev
)
668 return (dev
->flags
& ~(IFF_PROMISC
| IFF_ALLMULTI
)) |
669 (dev
->gflags
& (IFF_PROMISC
| IFF_ALLMULTI
));
672 static unsigned int rtnl_dev_combine_flags(const struct net_device
*dev
,
673 const struct ifinfomsg
*ifm
)
675 unsigned int flags
= ifm
->ifi_flags
;
677 /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
679 flags
= (flags
& ifm
->ifi_change
) |
680 (rtnl_dev_get_flags(dev
) & ~ifm
->ifi_change
);
685 static void copy_rtnl_link_stats(struct rtnl_link_stats
*a
,
686 const struct rtnl_link_stats64
*b
)
688 a
->rx_packets
= b
->rx_packets
;
689 a
->tx_packets
= b
->tx_packets
;
690 a
->rx_bytes
= b
->rx_bytes
;
691 a
->tx_bytes
= b
->tx_bytes
;
692 a
->rx_errors
= b
->rx_errors
;
693 a
->tx_errors
= b
->tx_errors
;
694 a
->rx_dropped
= b
->rx_dropped
;
695 a
->tx_dropped
= b
->tx_dropped
;
697 a
->multicast
= b
->multicast
;
698 a
->collisions
= b
->collisions
;
700 a
->rx_length_errors
= b
->rx_length_errors
;
701 a
->rx_over_errors
= b
->rx_over_errors
;
702 a
->rx_crc_errors
= b
->rx_crc_errors
;
703 a
->rx_frame_errors
= b
->rx_frame_errors
;
704 a
->rx_fifo_errors
= b
->rx_fifo_errors
;
705 a
->rx_missed_errors
= b
->rx_missed_errors
;
707 a
->tx_aborted_errors
= b
->tx_aborted_errors
;
708 a
->tx_carrier_errors
= b
->tx_carrier_errors
;
709 a
->tx_fifo_errors
= b
->tx_fifo_errors
;
710 a
->tx_heartbeat_errors
= b
->tx_heartbeat_errors
;
711 a
->tx_window_errors
= b
->tx_window_errors
;
713 a
->rx_compressed
= b
->rx_compressed
;
714 a
->tx_compressed
= b
->tx_compressed
;
717 static void copy_rtnl_link_stats64(void *v
, const struct rtnl_link_stats64
*b
)
719 memcpy(v
, b
, sizeof(*b
));
723 static inline int rtnl_vfinfo_size(const struct net_device
*dev
,
726 if (dev
->dev
.parent
&& dev_is_pci(dev
->dev
.parent
) &&
727 (ext_filter_mask
& RTEXT_FILTER_VF
)) {
728 int num_vfs
= dev_num_vf(dev
->dev
.parent
);
729 size_t size
= nla_total_size(sizeof(struct nlattr
));
730 size
+= nla_total_size(num_vfs
* sizeof(struct nlattr
));
732 (nla_total_size(sizeof(struct ifla_vf_mac
)) +
733 nla_total_size(sizeof(struct ifla_vf_vlan
)) +
734 nla_total_size(sizeof(struct ifla_vf_tx_rate
)) +
735 nla_total_size(sizeof(struct ifla_vf_spoofchk
)));
741 static size_t rtnl_port_size(const struct net_device
*dev
)
743 size_t port_size
= nla_total_size(4) /* PORT_VF */
744 + nla_total_size(PORT_PROFILE_MAX
) /* PORT_PROFILE */
745 + nla_total_size(sizeof(struct ifla_port_vsi
))
747 + nla_total_size(PORT_UUID_MAX
) /* PORT_INSTANCE_UUID */
748 + nla_total_size(PORT_UUID_MAX
) /* PORT_HOST_UUID */
749 + nla_total_size(1) /* PROT_VDP_REQUEST */
750 + nla_total_size(2); /* PORT_VDP_RESPONSE */
751 size_t vf_ports_size
= nla_total_size(sizeof(struct nlattr
));
752 size_t vf_port_size
= nla_total_size(sizeof(struct nlattr
))
754 size_t port_self_size
= nla_total_size(sizeof(struct nlattr
))
757 if (!dev
->netdev_ops
->ndo_get_vf_port
|| !dev
->dev
.parent
)
759 if (dev_num_vf(dev
->dev
.parent
))
760 return port_self_size
+ vf_ports_size
+
761 vf_port_size
* dev_num_vf(dev
->dev
.parent
);
763 return port_self_size
;
766 static noinline
size_t if_nlmsg_size(const struct net_device
*dev
,
769 return NLMSG_ALIGN(sizeof(struct ifinfomsg
))
770 + nla_total_size(IFNAMSIZ
) /* IFLA_IFNAME */
771 + nla_total_size(IFALIASZ
) /* IFLA_IFALIAS */
772 + nla_total_size(IFNAMSIZ
) /* IFLA_QDISC */
773 + nla_total_size(sizeof(struct rtnl_link_ifmap
))
774 + nla_total_size(sizeof(struct rtnl_link_stats
))
775 + nla_total_size(sizeof(struct rtnl_link_stats64
))
776 + nla_total_size(MAX_ADDR_LEN
) /* IFLA_ADDRESS */
777 + nla_total_size(MAX_ADDR_LEN
) /* IFLA_BROADCAST */
778 + nla_total_size(4) /* IFLA_TXQLEN */
779 + nla_total_size(4) /* IFLA_WEIGHT */
780 + nla_total_size(4) /* IFLA_MTU */
781 + nla_total_size(4) /* IFLA_LINK */
782 + nla_total_size(4) /* IFLA_MASTER */
783 + nla_total_size(4) /* IFLA_PROMISCUITY */
784 + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
785 + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
786 + nla_total_size(1) /* IFLA_OPERSTATE */
787 + nla_total_size(1) /* IFLA_LINKMODE */
788 + nla_total_size(ext_filter_mask
789 & RTEXT_FILTER_VF
? 4 : 0) /* IFLA_NUM_VF */
790 + rtnl_vfinfo_size(dev
, ext_filter_mask
) /* IFLA_VFINFO_LIST */
791 + rtnl_port_size(dev
) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
792 + rtnl_link_get_size(dev
) /* IFLA_LINKINFO */
793 + rtnl_link_get_af_size(dev
); /* IFLA_AF_SPEC */
796 static int rtnl_vf_ports_fill(struct sk_buff
*skb
, struct net_device
*dev
)
798 struct nlattr
*vf_ports
;
799 struct nlattr
*vf_port
;
803 vf_ports
= nla_nest_start(skb
, IFLA_VF_PORTS
);
807 for (vf
= 0; vf
< dev_num_vf(dev
->dev
.parent
); vf
++) {
808 vf_port
= nla_nest_start(skb
, IFLA_VF_PORT
);
810 goto nla_put_failure
;
811 if (nla_put_u32(skb
, IFLA_PORT_VF
, vf
))
812 goto nla_put_failure
;
813 err
= dev
->netdev_ops
->ndo_get_vf_port(dev
, vf
, skb
);
814 if (err
== -EMSGSIZE
)
815 goto nla_put_failure
;
817 nla_nest_cancel(skb
, vf_port
);
820 nla_nest_end(skb
, vf_port
);
823 nla_nest_end(skb
, vf_ports
);
828 nla_nest_cancel(skb
, vf_ports
);
832 static int rtnl_port_self_fill(struct sk_buff
*skb
, struct net_device
*dev
)
834 struct nlattr
*port_self
;
837 port_self
= nla_nest_start(skb
, IFLA_PORT_SELF
);
841 err
= dev
->netdev_ops
->ndo_get_vf_port(dev
, PORT_SELF_VF
, skb
);
843 nla_nest_cancel(skb
, port_self
);
844 return (err
== -EMSGSIZE
) ? err
: 0;
847 nla_nest_end(skb
, port_self
);
852 static int rtnl_port_fill(struct sk_buff
*skb
, struct net_device
*dev
)
856 if (!dev
->netdev_ops
->ndo_get_vf_port
|| !dev
->dev
.parent
)
859 err
= rtnl_port_self_fill(skb
, dev
);
863 if (dev_num_vf(dev
->dev
.parent
)) {
864 err
= rtnl_vf_ports_fill(skb
, dev
);
872 static int rtnl_fill_ifinfo(struct sk_buff
*skb
, struct net_device
*dev
,
873 int type
, u32 pid
, u32 seq
, u32 change
,
874 unsigned int flags
, u32 ext_filter_mask
)
876 struct ifinfomsg
*ifm
;
877 struct nlmsghdr
*nlh
;
878 struct rtnl_link_stats64 temp
;
879 const struct rtnl_link_stats64
*stats
;
880 struct nlattr
*attr
, *af_spec
;
881 struct rtnl_af_ops
*af_ops
;
884 nlh
= nlmsg_put(skb
, pid
, seq
, type
, sizeof(*ifm
), flags
);
888 ifm
= nlmsg_data(nlh
);
889 ifm
->ifi_family
= AF_UNSPEC
;
891 ifm
->ifi_type
= dev
->type
;
892 ifm
->ifi_index
= dev
->ifindex
;
893 ifm
->ifi_flags
= dev_get_flags(dev
);
894 ifm
->ifi_change
= change
;
896 if (nla_put_string(skb
, IFLA_IFNAME
, dev
->name
) ||
897 nla_put_u32(skb
, IFLA_TXQLEN
, dev
->tx_queue_len
) ||
898 nla_put_u8(skb
, IFLA_OPERSTATE
,
899 netif_running(dev
) ? dev
->operstate
: IF_OPER_DOWN
) ||
900 nla_put_u8(skb
, IFLA_LINKMODE
, dev
->link_mode
) ||
901 nla_put_u32(skb
, IFLA_MTU
, dev
->mtu
) ||
902 nla_put_u32(skb
, IFLA_GROUP
, dev
->group
) ||
903 nla_put_u32(skb
, IFLA_PROMISCUITY
, dev
->promiscuity
) ||
904 nla_put_u32(skb
, IFLA_NUM_TX_QUEUES
, dev
->num_tx_queues
) ||
906 nla_put_u32(skb
, IFLA_NUM_RX_QUEUES
, dev
->num_rx_queues
) ||
908 (dev
->ifindex
!= dev
->iflink
&&
909 nla_put_u32(skb
, IFLA_LINK
, dev
->iflink
)) ||
911 nla_put_u32(skb
, IFLA_MASTER
, dev
->master
->ifindex
)) ||
913 nla_put_string(skb
, IFLA_QDISC
, dev
->qdisc
->ops
->id
)) ||
915 nla_put_string(skb
, IFLA_IFALIAS
, dev
->ifalias
)))
916 goto nla_put_failure
;
919 struct rtnl_link_ifmap map
= {
920 .mem_start
= dev
->mem_start
,
921 .mem_end
= dev
->mem_end
,
922 .base_addr
= dev
->base_addr
,
925 .port
= dev
->if_port
,
927 if (nla_put(skb
, IFLA_MAP
, sizeof(map
), &map
))
928 goto nla_put_failure
;
932 if (nla_put(skb
, IFLA_ADDRESS
, dev
->addr_len
, dev
->dev_addr
) ||
933 nla_put(skb
, IFLA_BROADCAST
, dev
->addr_len
, dev
->broadcast
))
934 goto nla_put_failure
;
937 attr
= nla_reserve(skb
, IFLA_STATS
,
938 sizeof(struct rtnl_link_stats
));
940 goto nla_put_failure
;
942 stats
= dev_get_stats(dev
, &temp
);
943 copy_rtnl_link_stats(nla_data(attr
), stats
);
945 attr
= nla_reserve(skb
, IFLA_STATS64
,
946 sizeof(struct rtnl_link_stats64
));
948 goto nla_put_failure
;
949 copy_rtnl_link_stats64(nla_data(attr
), stats
);
951 if (dev
->dev
.parent
&& (ext_filter_mask
& RTEXT_FILTER_VF
) &&
952 nla_put_u32(skb
, IFLA_NUM_VF
, dev_num_vf(dev
->dev
.parent
)))
953 goto nla_put_failure
;
955 if (dev
->netdev_ops
->ndo_get_vf_config
&& dev
->dev
.parent
956 && (ext_filter_mask
& RTEXT_FILTER_VF
)) {
959 struct nlattr
*vfinfo
, *vf
;
960 int num_vfs
= dev_num_vf(dev
->dev
.parent
);
962 vfinfo
= nla_nest_start(skb
, IFLA_VFINFO_LIST
);
964 goto nla_put_failure
;
965 for (i
= 0; i
< num_vfs
; i
++) {
966 struct ifla_vf_info ivi
;
967 struct ifla_vf_mac vf_mac
;
968 struct ifla_vf_vlan vf_vlan
;
969 struct ifla_vf_tx_rate vf_tx_rate
;
970 struct ifla_vf_spoofchk vf_spoofchk
;
973 * Not all SR-IOV capable drivers support the
974 * spoofcheck query. Preset to -1 so the user
975 * space tool can detect that the driver didn't
979 if (dev
->netdev_ops
->ndo_get_vf_config(dev
, i
, &ivi
))
984 vf_spoofchk
.vf
= ivi
.vf
;
986 memcpy(vf_mac
.mac
, ivi
.mac
, sizeof(ivi
.mac
));
987 vf_vlan
.vlan
= ivi
.vlan
;
988 vf_vlan
.qos
= ivi
.qos
;
989 vf_tx_rate
.rate
= ivi
.tx_rate
;
990 vf_spoofchk
.setting
= ivi
.spoofchk
;
991 vf
= nla_nest_start(skb
, IFLA_VF_INFO
);
993 nla_nest_cancel(skb
, vfinfo
);
994 goto nla_put_failure
;
996 if (nla_put(skb
, IFLA_VF_MAC
, sizeof(vf_mac
), &vf_mac
) ||
997 nla_put(skb
, IFLA_VF_VLAN
, sizeof(vf_vlan
), &vf_vlan
) ||
998 nla_put(skb
, IFLA_VF_TX_RATE
, sizeof(vf_tx_rate
),
1000 nla_put(skb
, IFLA_VF_SPOOFCHK
, sizeof(vf_spoofchk
),
1002 goto nla_put_failure
;
1003 nla_nest_end(skb
, vf
);
1005 nla_nest_end(skb
, vfinfo
);
1008 if (rtnl_port_fill(skb
, dev
))
1009 goto nla_put_failure
;
1011 if (dev
->rtnl_link_ops
) {
1012 if (rtnl_link_fill(skb
, dev
) < 0)
1013 goto nla_put_failure
;
1016 if (!(af_spec
= nla_nest_start(skb
, IFLA_AF_SPEC
)))
1017 goto nla_put_failure
;
1019 list_for_each_entry(af_ops
, &rtnl_af_ops
, list
) {
1020 if (af_ops
->fill_link_af
) {
1024 if (!(af
= nla_nest_start(skb
, af_ops
->family
)))
1025 goto nla_put_failure
;
1027 err
= af_ops
->fill_link_af(skb
, dev
);
1030 * Caller may return ENODATA to indicate that there
1031 * was no data to be dumped. This is not an error, it
1032 * means we should trim the attribute header and
1035 if (err
== -ENODATA
)
1036 nla_nest_cancel(skb
, af
);
1038 goto nla_put_failure
;
1040 nla_nest_end(skb
, af
);
1044 nla_nest_end(skb
, af_spec
);
1046 return nlmsg_end(skb
, nlh
);
1049 nlmsg_cancel(skb
, nlh
);
1053 static int rtnl_dump_ifinfo(struct sk_buff
*skb
, struct netlink_callback
*cb
)
1055 struct net
*net
= sock_net(skb
->sk
);
1058 struct net_device
*dev
;
1059 struct hlist_head
*head
;
1060 struct hlist_node
*node
;
1061 struct nlattr
*tb
[IFLA_MAX
+1];
1062 u32 ext_filter_mask
= 0;
1065 s_idx
= cb
->args
[1];
1068 cb
->seq
= net
->dev_base_seq
;
1070 if (nlmsg_parse(cb
->nlh
, sizeof(struct rtgenmsg
), tb
, IFLA_MAX
,
1071 ifla_policy
) >= 0) {
1073 if (tb
[IFLA_EXT_MASK
])
1074 ext_filter_mask
= nla_get_u32(tb
[IFLA_EXT_MASK
]);
1077 for (h
= s_h
; h
< NETDEV_HASHENTRIES
; h
++, s_idx
= 0) {
1079 head
= &net
->dev_index_head
[h
];
1080 hlist_for_each_entry_rcu(dev
, node
, head
, index_hlist
) {
1083 if (rtnl_fill_ifinfo(skb
, dev
, RTM_NEWLINK
,
1084 NETLINK_CB(cb
->skb
).portid
,
1085 cb
->nlh
->nlmsg_seq
, 0,
1087 ext_filter_mask
) <= 0)
1090 nl_dump_check_consistent(cb
, nlmsg_hdr(skb
));
1103 const struct nla_policy ifla_policy
[IFLA_MAX
+1] = {
1104 [IFLA_IFNAME
] = { .type
= NLA_STRING
, .len
= IFNAMSIZ
-1 },
1105 [IFLA_ADDRESS
] = { .type
= NLA_BINARY
, .len
= MAX_ADDR_LEN
},
1106 [IFLA_BROADCAST
] = { .type
= NLA_BINARY
, .len
= MAX_ADDR_LEN
},
1107 [IFLA_MAP
] = { .len
= sizeof(struct rtnl_link_ifmap
) },
1108 [IFLA_MTU
] = { .type
= NLA_U32
},
1109 [IFLA_LINK
] = { .type
= NLA_U32
},
1110 [IFLA_MASTER
] = { .type
= NLA_U32
},
1111 [IFLA_TXQLEN
] = { .type
= NLA_U32
},
1112 [IFLA_WEIGHT
] = { .type
= NLA_U32
},
1113 [IFLA_OPERSTATE
] = { .type
= NLA_U8
},
1114 [IFLA_LINKMODE
] = { .type
= NLA_U8
},
1115 [IFLA_LINKINFO
] = { .type
= NLA_NESTED
},
1116 [IFLA_NET_NS_PID
] = { .type
= NLA_U32
},
1117 [IFLA_NET_NS_FD
] = { .type
= NLA_U32
},
1118 [IFLA_IFALIAS
] = { .type
= NLA_STRING
, .len
= IFALIASZ
-1 },
1119 [IFLA_VFINFO_LIST
] = {. type
= NLA_NESTED
},
1120 [IFLA_VF_PORTS
] = { .type
= NLA_NESTED
},
1121 [IFLA_PORT_SELF
] = { .type
= NLA_NESTED
},
1122 [IFLA_AF_SPEC
] = { .type
= NLA_NESTED
},
1123 [IFLA_EXT_MASK
] = { .type
= NLA_U32
},
1124 [IFLA_PROMISCUITY
] = { .type
= NLA_U32
},
1125 [IFLA_NUM_TX_QUEUES
] = { .type
= NLA_U32
},
1126 [IFLA_NUM_RX_QUEUES
] = { .type
= NLA_U32
},
1128 EXPORT_SYMBOL(ifla_policy
);
1130 static const struct nla_policy ifla_info_policy
[IFLA_INFO_MAX
+1] = {
1131 [IFLA_INFO_KIND
] = { .type
= NLA_STRING
},
1132 [IFLA_INFO_DATA
] = { .type
= NLA_NESTED
},
1135 static const struct nla_policy ifla_vfinfo_policy
[IFLA_VF_INFO_MAX
+1] = {
1136 [IFLA_VF_INFO
] = { .type
= NLA_NESTED
},
1139 static const struct nla_policy ifla_vf_policy
[IFLA_VF_MAX
+1] = {
1140 [IFLA_VF_MAC
] = { .type
= NLA_BINARY
,
1141 .len
= sizeof(struct ifla_vf_mac
) },
1142 [IFLA_VF_VLAN
] = { .type
= NLA_BINARY
,
1143 .len
= sizeof(struct ifla_vf_vlan
) },
1144 [IFLA_VF_TX_RATE
] = { .type
= NLA_BINARY
,
1145 .len
= sizeof(struct ifla_vf_tx_rate
) },
1146 [IFLA_VF_SPOOFCHK
] = { .type
= NLA_BINARY
,
1147 .len
= sizeof(struct ifla_vf_spoofchk
) },
1150 static const struct nla_policy ifla_port_policy
[IFLA_PORT_MAX
+1] = {
1151 [IFLA_PORT_VF
] = { .type
= NLA_U32
},
1152 [IFLA_PORT_PROFILE
] = { .type
= NLA_STRING
,
1153 .len
= PORT_PROFILE_MAX
},
1154 [IFLA_PORT_VSI_TYPE
] = { .type
= NLA_BINARY
,
1155 .len
= sizeof(struct ifla_port_vsi
)},
1156 [IFLA_PORT_INSTANCE_UUID
] = { .type
= NLA_BINARY
,
1157 .len
= PORT_UUID_MAX
},
1158 [IFLA_PORT_HOST_UUID
] = { .type
= NLA_STRING
,
1159 .len
= PORT_UUID_MAX
},
1160 [IFLA_PORT_REQUEST
] = { .type
= NLA_U8
, },
1161 [IFLA_PORT_RESPONSE
] = { .type
= NLA_U16
, },
1164 struct net
*rtnl_link_get_net(struct net
*src_net
, struct nlattr
*tb
[])
1167 /* Examine the link attributes and figure out which
1168 * network namespace we are talking about.
1170 if (tb
[IFLA_NET_NS_PID
])
1171 net
= get_net_ns_by_pid(nla_get_u32(tb
[IFLA_NET_NS_PID
]));
1172 else if (tb
[IFLA_NET_NS_FD
])
1173 net
= get_net_ns_by_fd(nla_get_u32(tb
[IFLA_NET_NS_FD
]));
1175 net
= get_net(src_net
);
1178 EXPORT_SYMBOL(rtnl_link_get_net
);
1180 static int validate_linkmsg(struct net_device
*dev
, struct nlattr
*tb
[])
1183 if (tb
[IFLA_ADDRESS
] &&
1184 nla_len(tb
[IFLA_ADDRESS
]) < dev
->addr_len
)
1187 if (tb
[IFLA_BROADCAST
] &&
1188 nla_len(tb
[IFLA_BROADCAST
]) < dev
->addr_len
)
1192 if (tb
[IFLA_AF_SPEC
]) {
1196 nla_for_each_nested(af
, tb
[IFLA_AF_SPEC
], rem
) {
1197 const struct rtnl_af_ops
*af_ops
;
1199 if (!(af_ops
= rtnl_af_lookup(nla_type(af
))))
1200 return -EAFNOSUPPORT
;
1202 if (!af_ops
->set_link_af
)
1205 if (af_ops
->validate_link_af
) {
1206 err
= af_ops
->validate_link_af(dev
, af
);
1216 static int do_setvfinfo(struct net_device
*dev
, struct nlattr
*attr
)
1218 int rem
, err
= -EINVAL
;
1220 const struct net_device_ops
*ops
= dev
->netdev_ops
;
1222 nla_for_each_nested(vf
, attr
, rem
) {
1223 switch (nla_type(vf
)) {
1225 struct ifla_vf_mac
*ivm
;
1228 if (ops
->ndo_set_vf_mac
)
1229 err
= ops
->ndo_set_vf_mac(dev
, ivm
->vf
,
1233 case IFLA_VF_VLAN
: {
1234 struct ifla_vf_vlan
*ivv
;
1237 if (ops
->ndo_set_vf_vlan
)
1238 err
= ops
->ndo_set_vf_vlan(dev
, ivv
->vf
,
1243 case IFLA_VF_TX_RATE
: {
1244 struct ifla_vf_tx_rate
*ivt
;
1247 if (ops
->ndo_set_vf_tx_rate
)
1248 err
= ops
->ndo_set_vf_tx_rate(dev
, ivt
->vf
,
1252 case IFLA_VF_SPOOFCHK
: {
1253 struct ifla_vf_spoofchk
*ivs
;
1256 if (ops
->ndo_set_vf_spoofchk
)
1257 err
= ops
->ndo_set_vf_spoofchk(dev
, ivs
->vf
,
1271 static int do_set_master(struct net_device
*dev
, int ifindex
)
1273 struct net_device
*master_dev
;
1274 const struct net_device_ops
*ops
;
1278 if (dev
->master
->ifindex
== ifindex
)
1280 ops
= dev
->master
->netdev_ops
;
1281 if (ops
->ndo_del_slave
) {
1282 err
= ops
->ndo_del_slave(dev
->master
, dev
);
1291 master_dev
= __dev_get_by_index(dev_net(dev
), ifindex
);
1294 ops
= master_dev
->netdev_ops
;
1295 if (ops
->ndo_add_slave
) {
1296 err
= ops
->ndo_add_slave(master_dev
, dev
);
1306 static int do_setlink(struct net_device
*dev
, struct ifinfomsg
*ifm
,
1307 struct nlattr
**tb
, char *ifname
, int modified
)
1309 const struct net_device_ops
*ops
= dev
->netdev_ops
;
1310 int send_addr_notify
= 0;
1313 if (tb
[IFLA_NET_NS_PID
] || tb
[IFLA_NET_NS_FD
]) {
1314 struct net
*net
= rtnl_link_get_net(dev_net(dev
), tb
);
1319 err
= dev_change_net_namespace(dev
, net
, ifname
);
1327 struct rtnl_link_ifmap
*u_map
;
1330 if (!ops
->ndo_set_config
) {
1335 if (!netif_device_present(dev
)) {
1340 u_map
= nla_data(tb
[IFLA_MAP
]);
1341 k_map
.mem_start
= (unsigned long) u_map
->mem_start
;
1342 k_map
.mem_end
= (unsigned long) u_map
->mem_end
;
1343 k_map
.base_addr
= (unsigned short) u_map
->base_addr
;
1344 k_map
.irq
= (unsigned char) u_map
->irq
;
1345 k_map
.dma
= (unsigned char) u_map
->dma
;
1346 k_map
.port
= (unsigned char) u_map
->port
;
1348 err
= ops
->ndo_set_config(dev
, &k_map
);
1355 if (tb
[IFLA_ADDRESS
]) {
1356 struct sockaddr
*sa
;
1359 if (!ops
->ndo_set_mac_address
) {
1364 if (!netif_device_present(dev
)) {
1369 len
= sizeof(sa_family_t
) + dev
->addr_len
;
1370 sa
= kmalloc(len
, GFP_KERNEL
);
1375 sa
->sa_family
= dev
->type
;
1376 memcpy(sa
->sa_data
, nla_data(tb
[IFLA_ADDRESS
]),
1378 err
= ops
->ndo_set_mac_address(dev
, sa
);
1382 send_addr_notify
= 1;
1384 add_device_randomness(dev
->dev_addr
, dev
->addr_len
);
1388 err
= dev_set_mtu(dev
, nla_get_u32(tb
[IFLA_MTU
]));
1394 if (tb
[IFLA_GROUP
]) {
1395 dev_set_group(dev
, nla_get_u32(tb
[IFLA_GROUP
]));
1400 * Interface selected by interface index but interface
1401 * name provided implies that a name change has been
1404 if (ifm
->ifi_index
> 0 && ifname
[0]) {
1405 err
= dev_change_name(dev
, ifname
);
1411 if (tb
[IFLA_IFALIAS
]) {
1412 err
= dev_set_alias(dev
, nla_data(tb
[IFLA_IFALIAS
]),
1413 nla_len(tb
[IFLA_IFALIAS
]));
1419 if (tb
[IFLA_BROADCAST
]) {
1420 nla_memcpy(dev
->broadcast
, tb
[IFLA_BROADCAST
], dev
->addr_len
);
1421 send_addr_notify
= 1;
1424 if (ifm
->ifi_flags
|| ifm
->ifi_change
) {
1425 err
= dev_change_flags(dev
, rtnl_dev_combine_flags(dev
, ifm
));
1430 if (tb
[IFLA_MASTER
]) {
1431 err
= do_set_master(dev
, nla_get_u32(tb
[IFLA_MASTER
]));
1437 if (tb
[IFLA_TXQLEN
])
1438 dev
->tx_queue_len
= nla_get_u32(tb
[IFLA_TXQLEN
]);
1440 if (tb
[IFLA_OPERSTATE
])
1441 set_operstate(dev
, nla_get_u8(tb
[IFLA_OPERSTATE
]));
1443 if (tb
[IFLA_LINKMODE
]) {
1444 write_lock_bh(&dev_base_lock
);
1445 dev
->link_mode
= nla_get_u8(tb
[IFLA_LINKMODE
]);
1446 write_unlock_bh(&dev_base_lock
);
1449 if (tb
[IFLA_VFINFO_LIST
]) {
1450 struct nlattr
*attr
;
1452 nla_for_each_nested(attr
, tb
[IFLA_VFINFO_LIST
], rem
) {
1453 if (nla_type(attr
) != IFLA_VF_INFO
) {
1457 err
= do_setvfinfo(dev
, attr
);
1465 if (tb
[IFLA_VF_PORTS
]) {
1466 struct nlattr
*port
[IFLA_PORT_MAX
+1];
1467 struct nlattr
*attr
;
1472 if (!ops
->ndo_set_vf_port
)
1475 nla_for_each_nested(attr
, tb
[IFLA_VF_PORTS
], rem
) {
1476 if (nla_type(attr
) != IFLA_VF_PORT
)
1478 err
= nla_parse_nested(port
, IFLA_PORT_MAX
,
1479 attr
, ifla_port_policy
);
1482 if (!port
[IFLA_PORT_VF
]) {
1486 vf
= nla_get_u32(port
[IFLA_PORT_VF
]);
1487 err
= ops
->ndo_set_vf_port(dev
, vf
, port
);
1495 if (tb
[IFLA_PORT_SELF
]) {
1496 struct nlattr
*port
[IFLA_PORT_MAX
+1];
1498 err
= nla_parse_nested(port
, IFLA_PORT_MAX
,
1499 tb
[IFLA_PORT_SELF
], ifla_port_policy
);
1504 if (ops
->ndo_set_vf_port
)
1505 err
= ops
->ndo_set_vf_port(dev
, PORT_SELF_VF
, port
);
1511 if (tb
[IFLA_AF_SPEC
]) {
1515 nla_for_each_nested(af
, tb
[IFLA_AF_SPEC
], rem
) {
1516 const struct rtnl_af_ops
*af_ops
;
1518 if (!(af_ops
= rtnl_af_lookup(nla_type(af
))))
1521 err
= af_ops
->set_link_af(dev
, af
);
1531 if (err
< 0 && modified
)
1532 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",
1535 if (send_addr_notify
)
1536 call_netdevice_notifiers(NETDEV_CHANGEADDR
, dev
);
1541 static int rtnl_setlink(struct sk_buff
*skb
, struct nlmsghdr
*nlh
, void *arg
)
1543 struct net
*net
= sock_net(skb
->sk
);
1544 struct ifinfomsg
*ifm
;
1545 struct net_device
*dev
;
1547 struct nlattr
*tb
[IFLA_MAX
+1];
1548 char ifname
[IFNAMSIZ
];
1550 err
= nlmsg_parse(nlh
, sizeof(*ifm
), tb
, IFLA_MAX
, ifla_policy
);
1554 if (tb
[IFLA_IFNAME
])
1555 nla_strlcpy(ifname
, tb
[IFLA_IFNAME
], IFNAMSIZ
);
1560 ifm
= nlmsg_data(nlh
);
1561 if (ifm
->ifi_index
> 0)
1562 dev
= __dev_get_by_index(net
, ifm
->ifi_index
);
1563 else if (tb
[IFLA_IFNAME
])
1564 dev
= __dev_get_by_name(net
, ifname
);
1573 err
= validate_linkmsg(dev
, tb
);
1577 err
= do_setlink(dev
, ifm
, tb
, ifname
, 0);
1582 static int rtnl_dellink(struct sk_buff
*skb
, struct nlmsghdr
*nlh
, void *arg
)
1584 struct net
*net
= sock_net(skb
->sk
);
1585 const struct rtnl_link_ops
*ops
;
1586 struct net_device
*dev
;
1587 struct ifinfomsg
*ifm
;
1588 char ifname
[IFNAMSIZ
];
1589 struct nlattr
*tb
[IFLA_MAX
+1];
1591 LIST_HEAD(list_kill
);
1593 err
= nlmsg_parse(nlh
, sizeof(*ifm
), tb
, IFLA_MAX
, ifla_policy
);
1597 if (tb
[IFLA_IFNAME
])
1598 nla_strlcpy(ifname
, tb
[IFLA_IFNAME
], IFNAMSIZ
);
1600 ifm
= nlmsg_data(nlh
);
1601 if (ifm
->ifi_index
> 0)
1602 dev
= __dev_get_by_index(net
, ifm
->ifi_index
);
1603 else if (tb
[IFLA_IFNAME
])
1604 dev
= __dev_get_by_name(net
, ifname
);
1611 ops
= dev
->rtnl_link_ops
;
1615 ops
->dellink(dev
, &list_kill
);
1616 unregister_netdevice_many(&list_kill
);
1617 list_del(&list_kill
);
1621 int rtnl_configure_link(struct net_device
*dev
, const struct ifinfomsg
*ifm
)
1623 unsigned int old_flags
;
1626 old_flags
= dev
->flags
;
1627 if (ifm
&& (ifm
->ifi_flags
|| ifm
->ifi_change
)) {
1628 err
= __dev_change_flags(dev
, rtnl_dev_combine_flags(dev
, ifm
));
1633 dev
->rtnl_link_state
= RTNL_LINK_INITIALIZED
;
1634 rtmsg_ifinfo(RTM_NEWLINK
, dev
, ~0U);
1636 __dev_notify_flags(dev
, old_flags
);
1639 EXPORT_SYMBOL(rtnl_configure_link
);
1641 struct net_device
*rtnl_create_link(struct net
*src_net
, struct net
*net
,
1642 char *ifname
, const struct rtnl_link_ops
*ops
, struct nlattr
*tb
[])
1645 struct net_device
*dev
;
1646 unsigned int num_tx_queues
= 1;
1647 unsigned int num_rx_queues
= 1;
1649 if (tb
[IFLA_NUM_TX_QUEUES
])
1650 num_tx_queues
= nla_get_u32(tb
[IFLA_NUM_TX_QUEUES
]);
1651 else if (ops
->get_num_tx_queues
)
1652 num_tx_queues
= ops
->get_num_tx_queues();
1654 if (tb
[IFLA_NUM_RX_QUEUES
])
1655 num_rx_queues
= nla_get_u32(tb
[IFLA_NUM_RX_QUEUES
]);
1656 else if (ops
->get_num_rx_queues
)
1657 num_rx_queues
= ops
->get_num_rx_queues();
1660 dev
= alloc_netdev_mqs(ops
->priv_size
, ifname
, ops
->setup
,
1661 num_tx_queues
, num_rx_queues
);
1665 dev_net_set(dev
, net
);
1666 dev
->rtnl_link_ops
= ops
;
1667 dev
->rtnl_link_state
= RTNL_LINK_INITIALIZING
;
1670 dev
->mtu
= nla_get_u32(tb
[IFLA_MTU
]);
1671 if (tb
[IFLA_ADDRESS
])
1672 memcpy(dev
->dev_addr
, nla_data(tb
[IFLA_ADDRESS
]),
1673 nla_len(tb
[IFLA_ADDRESS
]));
1674 if (tb
[IFLA_BROADCAST
])
1675 memcpy(dev
->broadcast
, nla_data(tb
[IFLA_BROADCAST
]),
1676 nla_len(tb
[IFLA_BROADCAST
]));
1677 if (tb
[IFLA_TXQLEN
])
1678 dev
->tx_queue_len
= nla_get_u32(tb
[IFLA_TXQLEN
]);
1679 if (tb
[IFLA_OPERSTATE
])
1680 set_operstate(dev
, nla_get_u8(tb
[IFLA_OPERSTATE
]));
1681 if (tb
[IFLA_LINKMODE
])
1682 dev
->link_mode
= nla_get_u8(tb
[IFLA_LINKMODE
]);
1684 dev_set_group(dev
, nla_get_u32(tb
[IFLA_GROUP
]));
1689 return ERR_PTR(err
);
1691 EXPORT_SYMBOL(rtnl_create_link
);
1693 static int rtnl_group_changelink(struct net
*net
, int group
,
1694 struct ifinfomsg
*ifm
,
1697 struct net_device
*dev
;
1700 for_each_netdev(net
, dev
) {
1701 if (dev
->group
== group
) {
1702 err
= do_setlink(dev
, ifm
, tb
, NULL
, 0);
1711 static int rtnl_newlink(struct sk_buff
*skb
, struct nlmsghdr
*nlh
, void *arg
)
1713 struct net
*net
= sock_net(skb
->sk
);
1714 const struct rtnl_link_ops
*ops
;
1715 struct net_device
*dev
;
1716 struct ifinfomsg
*ifm
;
1717 char kind
[MODULE_NAME_LEN
];
1718 char ifname
[IFNAMSIZ
];
1719 struct nlattr
*tb
[IFLA_MAX
+1];
1720 struct nlattr
*linkinfo
[IFLA_INFO_MAX
+1];
1723 #ifdef CONFIG_MODULES
1726 err
= nlmsg_parse(nlh
, sizeof(*ifm
), tb
, IFLA_MAX
, ifla_policy
);
1730 if (tb
[IFLA_IFNAME
])
1731 nla_strlcpy(ifname
, tb
[IFLA_IFNAME
], IFNAMSIZ
);
1735 ifm
= nlmsg_data(nlh
);
1736 if (ifm
->ifi_index
> 0)
1737 dev
= __dev_get_by_index(net
, ifm
->ifi_index
);
1740 dev
= __dev_get_by_name(net
, ifname
);
1745 err
= validate_linkmsg(dev
, tb
);
1749 if (tb
[IFLA_LINKINFO
]) {
1750 err
= nla_parse_nested(linkinfo
, IFLA_INFO_MAX
,
1751 tb
[IFLA_LINKINFO
], ifla_info_policy
);
1755 memset(linkinfo
, 0, sizeof(linkinfo
));
1757 if (linkinfo
[IFLA_INFO_KIND
]) {
1758 nla_strlcpy(kind
, linkinfo
[IFLA_INFO_KIND
], sizeof(kind
));
1759 ops
= rtnl_link_ops_get(kind
);
1766 struct nlattr
*attr
[ops
? ops
->maxtype
+ 1 : 0], **data
= NULL
;
1767 struct net
*dest_net
;
1770 if (ops
->maxtype
&& linkinfo
[IFLA_INFO_DATA
]) {
1771 err
= nla_parse_nested(attr
, ops
->maxtype
,
1772 linkinfo
[IFLA_INFO_DATA
],
1778 if (ops
->validate
) {
1779 err
= ops
->validate(tb
, data
);
1788 if (nlh
->nlmsg_flags
& NLM_F_EXCL
)
1790 if (nlh
->nlmsg_flags
& NLM_F_REPLACE
)
1793 if (linkinfo
[IFLA_INFO_DATA
]) {
1794 if (!ops
|| ops
!= dev
->rtnl_link_ops
||
1798 err
= ops
->changelink(dev
, tb
, data
);
1804 return do_setlink(dev
, ifm
, tb
, ifname
, modified
);
1807 if (!(nlh
->nlmsg_flags
& NLM_F_CREATE
)) {
1808 if (ifm
->ifi_index
== 0 && tb
[IFLA_GROUP
])
1809 return rtnl_group_changelink(net
,
1810 nla_get_u32(tb
[IFLA_GROUP
]),
1815 if (tb
[IFLA_MAP
] || tb
[IFLA_MASTER
] || tb
[IFLA_PROTINFO
])
1819 #ifdef CONFIG_MODULES
1822 request_module("rtnl-link-%s", kind
);
1824 ops
= rtnl_link_ops_get(kind
);
1833 snprintf(ifname
, IFNAMSIZ
, "%s%%d", ops
->kind
);
1835 dest_net
= rtnl_link_get_net(net
, tb
);
1836 if (IS_ERR(dest_net
))
1837 return PTR_ERR(dest_net
);
1839 dev
= rtnl_create_link(net
, dest_net
, ifname
, ops
, tb
);
1845 dev
->ifindex
= ifm
->ifi_index
;
1848 err
= ops
->newlink(net
, dev
, tb
, data
);
1850 err
= register_netdevice(dev
);
1852 if (err
< 0 && !IS_ERR(dev
))
1857 err
= rtnl_configure_link(dev
, ifm
);
1859 unregister_netdevice(dev
);
1866 static int rtnl_getlink(struct sk_buff
*skb
, struct nlmsghdr
* nlh
, void *arg
)
1868 struct net
*net
= sock_net(skb
->sk
);
1869 struct ifinfomsg
*ifm
;
1870 char ifname
[IFNAMSIZ
];
1871 struct nlattr
*tb
[IFLA_MAX
+1];
1872 struct net_device
*dev
= NULL
;
1873 struct sk_buff
*nskb
;
1875 u32 ext_filter_mask
= 0;
1877 err
= nlmsg_parse(nlh
, sizeof(*ifm
), tb
, IFLA_MAX
, ifla_policy
);
1881 if (tb
[IFLA_IFNAME
])
1882 nla_strlcpy(ifname
, tb
[IFLA_IFNAME
], IFNAMSIZ
);
1884 if (tb
[IFLA_EXT_MASK
])
1885 ext_filter_mask
= nla_get_u32(tb
[IFLA_EXT_MASK
]);
1887 ifm
= nlmsg_data(nlh
);
1888 if (ifm
->ifi_index
> 0)
1889 dev
= __dev_get_by_index(net
, ifm
->ifi_index
);
1890 else if (tb
[IFLA_IFNAME
])
1891 dev
= __dev_get_by_name(net
, ifname
);
1898 nskb
= nlmsg_new(if_nlmsg_size(dev
, ext_filter_mask
), GFP_KERNEL
);
1902 err
= rtnl_fill_ifinfo(nskb
, dev
, RTM_NEWLINK
, NETLINK_CB(skb
).portid
,
1903 nlh
->nlmsg_seq
, 0, 0, ext_filter_mask
);
1905 /* -EMSGSIZE implies BUG in if_nlmsg_size */
1906 WARN_ON(err
== -EMSGSIZE
);
1909 err
= rtnl_unicast(nskb
, net
, NETLINK_CB(skb
).portid
);
1914 static u16
rtnl_calcit(struct sk_buff
*skb
, struct nlmsghdr
*nlh
)
1916 struct net
*net
= sock_net(skb
->sk
);
1917 struct net_device
*dev
;
1918 struct nlattr
*tb
[IFLA_MAX
+1];
1919 u32 ext_filter_mask
= 0;
1920 u16 min_ifinfo_dump_size
= 0;
1922 if (nlmsg_parse(nlh
, sizeof(struct rtgenmsg
), tb
, IFLA_MAX
,
1923 ifla_policy
) >= 0) {
1924 if (tb
[IFLA_EXT_MASK
])
1925 ext_filter_mask
= nla_get_u32(tb
[IFLA_EXT_MASK
]);
1928 if (!ext_filter_mask
)
1929 return NLMSG_GOODSIZE
;
1931 * traverse the list of net devices and compute the minimum
1932 * buffer size based upon the filter mask.
1934 list_for_each_entry(dev
, &net
->dev_base_head
, dev_list
) {
1935 min_ifinfo_dump_size
= max_t(u16
, min_ifinfo_dump_size
,
1940 return min_ifinfo_dump_size
;
1943 static int rtnl_dump_all(struct sk_buff
*skb
, struct netlink_callback
*cb
)
1946 int s_idx
= cb
->family
;
1950 for (idx
= 1; idx
<= RTNL_FAMILY_MAX
; idx
++) {
1951 int type
= cb
->nlh
->nlmsg_type
-RTM_BASE
;
1952 if (idx
< s_idx
|| idx
== PF_PACKET
)
1954 if (rtnl_msg_handlers
[idx
] == NULL
||
1955 rtnl_msg_handlers
[idx
][type
].dumpit
== NULL
)
1958 memset(&cb
->args
[0], 0, sizeof(cb
->args
));
1959 if (rtnl_msg_handlers
[idx
][type
].dumpit(skb
, cb
))
1967 void rtmsg_ifinfo(int type
, struct net_device
*dev
, unsigned int change
)
1969 struct net
*net
= dev_net(dev
);
1970 struct sk_buff
*skb
;
1972 size_t if_info_size
;
1974 skb
= nlmsg_new((if_info_size
= if_nlmsg_size(dev
, 0)), GFP_KERNEL
);
1978 err
= rtnl_fill_ifinfo(skb
, dev
, type
, 0, 0, change
, 0, 0);
1980 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
1981 WARN_ON(err
== -EMSGSIZE
);
1985 rtnl_notify(skb
, net
, 0, RTNLGRP_LINK
, NULL
, GFP_KERNEL
);
1989 rtnl_set_sk_err(net
, RTNLGRP_LINK
, err
);
1992 static int nlmsg_populate_fdb_fill(struct sk_buff
*skb
,
1993 struct net_device
*dev
,
1994 u8
*addr
, u32 pid
, u32 seq
,
1995 int type
, unsigned int flags
)
1997 struct nlmsghdr
*nlh
;
2000 nlh
= nlmsg_put(skb
, pid
, seq
, type
, sizeof(*ndm
), NLM_F_MULTI
);
2004 ndm
= nlmsg_data(nlh
);
2005 ndm
->ndm_family
= AF_BRIDGE
;
2008 ndm
->ndm_flags
= flags
;
2010 ndm
->ndm_ifindex
= dev
->ifindex
;
2011 ndm
->ndm_state
= NUD_PERMANENT
;
2013 if (nla_put(skb
, NDA_LLADDR
, ETH_ALEN
, addr
))
2014 goto nla_put_failure
;
2016 return nlmsg_end(skb
, nlh
);
2019 nlmsg_cancel(skb
, nlh
);
2023 static inline size_t rtnl_fdb_nlmsg_size(void)
2025 return NLMSG_ALIGN(sizeof(struct ndmsg
)) + nla_total_size(ETH_ALEN
);
2028 static void rtnl_fdb_notify(struct net_device
*dev
, u8
*addr
, int type
)
2030 struct net
*net
= dev_net(dev
);
2031 struct sk_buff
*skb
;
2034 skb
= nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC
);
2038 err
= nlmsg_populate_fdb_fill(skb
, dev
, addr
, 0, 0, type
, NTF_SELF
);
2044 rtnl_notify(skb
, net
, 0, RTNLGRP_NEIGH
, NULL
, GFP_ATOMIC
);
2047 rtnl_set_sk_err(net
, RTNLGRP_NEIGH
, err
);
2050 static int rtnl_fdb_add(struct sk_buff
*skb
, struct nlmsghdr
*nlh
, void *arg
)
2052 struct net
*net
= sock_net(skb
->sk
);
2053 struct net_device
*master
= NULL
;
2055 struct nlattr
*tb
[NDA_MAX
+1];
2056 struct net_device
*dev
;
2060 err
= nlmsg_parse(nlh
, sizeof(*ndm
), tb
, NDA_MAX
, NULL
);
2064 ndm
= nlmsg_data(nlh
);
2065 if (ndm
->ndm_ifindex
== 0) {
2066 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n");
2070 dev
= __dev_get_by_index(net
, ndm
->ndm_ifindex
);
2072 pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n");
2076 if (!tb
[NDA_LLADDR
] || nla_len(tb
[NDA_LLADDR
]) != ETH_ALEN
) {
2077 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n");
2081 addr
= nla_data(tb
[NDA_LLADDR
]);
2082 if (!is_valid_ether_addr(addr
)) {
2083 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ether address\n");
2089 /* Support fdb on master device the net/bridge default case */
2090 if ((!ndm
->ndm_flags
|| ndm
->ndm_flags
& NTF_MASTER
) &&
2091 (dev
->priv_flags
& IFF_BRIDGE_PORT
)) {
2092 master
= dev
->master
;
2093 err
= master
->netdev_ops
->ndo_fdb_add(ndm
, tb
,
2099 ndm
->ndm_flags
&= ~NTF_MASTER
;
2102 /* Embedded bridge, macvlan, and any other device support */
2103 if ((ndm
->ndm_flags
& NTF_SELF
) && dev
->netdev_ops
->ndo_fdb_add
) {
2104 err
= dev
->netdev_ops
->ndo_fdb_add(ndm
, tb
,
2109 rtnl_fdb_notify(dev
, addr
, RTM_NEWNEIGH
);
2110 ndm
->ndm_flags
&= ~NTF_SELF
;
2117 static int rtnl_fdb_del(struct sk_buff
*skb
, struct nlmsghdr
*nlh
, void *arg
)
2119 struct net
*net
= sock_net(skb
->sk
);
2121 struct nlattr
*llattr
;
2122 struct net_device
*dev
;
2126 if (nlmsg_len(nlh
) < sizeof(*ndm
))
2129 ndm
= nlmsg_data(nlh
);
2130 if (ndm
->ndm_ifindex
== 0) {
2131 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n");
2135 dev
= __dev_get_by_index(net
, ndm
->ndm_ifindex
);
2137 pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n");
2141 llattr
= nlmsg_find_attr(nlh
, sizeof(*ndm
), NDA_LLADDR
);
2142 if (llattr
== NULL
|| nla_len(llattr
) != ETH_ALEN
) {
2143 pr_info("PF_BRIGDE: RTM_DELNEIGH with invalid address\n");
2147 addr
= nla_data(llattr
);
2150 /* Support fdb on master device the net/bridge default case */
2151 if ((!ndm
->ndm_flags
|| ndm
->ndm_flags
& NTF_MASTER
) &&
2152 (dev
->priv_flags
& IFF_BRIDGE_PORT
)) {
2153 struct net_device
*master
= dev
->master
;
2155 if (master
->netdev_ops
->ndo_fdb_del
)
2156 err
= master
->netdev_ops
->ndo_fdb_del(ndm
, dev
, addr
);
2161 ndm
->ndm_flags
&= ~NTF_MASTER
;
2164 /* Embedded bridge, macvlan, and any other device support */
2165 if ((ndm
->ndm_flags
& NTF_SELF
) && dev
->netdev_ops
->ndo_fdb_del
) {
2166 err
= dev
->netdev_ops
->ndo_fdb_del(ndm
, dev
, addr
);
2169 rtnl_fdb_notify(dev
, addr
, RTM_DELNEIGH
);
2170 ndm
->ndm_flags
&= ~NTF_SELF
;
2177 static int nlmsg_populate_fdb(struct sk_buff
*skb
,
2178 struct netlink_callback
*cb
,
2179 struct net_device
*dev
,
2181 struct netdev_hw_addr_list
*list
)
2183 struct netdev_hw_addr
*ha
;
2187 portid
= NETLINK_CB(cb
->skb
).portid
;
2188 seq
= cb
->nlh
->nlmsg_seq
;
2190 list_for_each_entry(ha
, &list
->list
, list
) {
2191 if (*idx
< cb
->args
[0])
2194 err
= nlmsg_populate_fdb_fill(skb
, dev
, ha
->addr
,
2196 RTM_NEWNEIGH
, NTF_SELF
);
2206 * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
2207 * @nlh: netlink message header
2210 * Default netdevice operation to dump the existing unicast address list.
2211 * Returns zero on success.
2213 int ndo_dflt_fdb_dump(struct sk_buff
*skb
,
2214 struct netlink_callback
*cb
,
2215 struct net_device
*dev
,
2220 netif_addr_lock_bh(dev
);
2221 err
= nlmsg_populate_fdb(skb
, cb
, dev
, &idx
, &dev
->uc
);
2224 nlmsg_populate_fdb(skb
, cb
, dev
, &idx
, &dev
->mc
);
2226 netif_addr_unlock_bh(dev
);
2229 EXPORT_SYMBOL(ndo_dflt_fdb_dump
);
2231 static int rtnl_fdb_dump(struct sk_buff
*skb
, struct netlink_callback
*cb
)
2234 struct net
*net
= sock_net(skb
->sk
);
2235 struct net_device
*dev
;
2238 for_each_netdev_rcu(net
, dev
) {
2239 if (dev
->priv_flags
& IFF_BRIDGE_PORT
) {
2240 struct net_device
*master
= dev
->master
;
2241 const struct net_device_ops
*ops
= master
->netdev_ops
;
2243 if (ops
->ndo_fdb_dump
)
2244 idx
= ops
->ndo_fdb_dump(skb
, cb
, dev
, idx
);
2247 if (dev
->netdev_ops
->ndo_fdb_dump
)
2248 idx
= dev
->netdev_ops
->ndo_fdb_dump(skb
, cb
, dev
, idx
);
2256 int ndo_dflt_bridge_getlink(struct sk_buff
*skb
, u32 pid
, u32 seq
,
2257 struct net_device
*dev
, u16 mode
)
2259 struct nlmsghdr
*nlh
;
2260 struct ifinfomsg
*ifm
;
2261 struct nlattr
*br_afspec
;
2262 u8 operstate
= netif_running(dev
) ? dev
->operstate
: IF_OPER_DOWN
;
2264 nlh
= nlmsg_put(skb
, pid
, seq
, RTM_NEWLINK
, sizeof(*ifm
), NLM_F_MULTI
);
2268 ifm
= nlmsg_data(nlh
);
2269 ifm
->ifi_family
= AF_BRIDGE
;
2271 ifm
->ifi_type
= dev
->type
;
2272 ifm
->ifi_index
= dev
->ifindex
;
2273 ifm
->ifi_flags
= dev_get_flags(dev
);
2274 ifm
->ifi_change
= 0;
2277 if (nla_put_string(skb
, IFLA_IFNAME
, dev
->name
) ||
2278 nla_put_u32(skb
, IFLA_MTU
, dev
->mtu
) ||
2279 nla_put_u8(skb
, IFLA_OPERSTATE
, operstate
) ||
2281 nla_put_u32(skb
, IFLA_MASTER
, dev
->master
->ifindex
)) ||
2283 nla_put(skb
, IFLA_ADDRESS
, dev
->addr_len
, dev
->dev_addr
)) ||
2284 (dev
->ifindex
!= dev
->iflink
&&
2285 nla_put_u32(skb
, IFLA_LINK
, dev
->iflink
)))
2286 goto nla_put_failure
;
2288 br_afspec
= nla_nest_start(skb
, IFLA_AF_SPEC
);
2290 goto nla_put_failure
;
2292 if (nla_put_u16(skb
, IFLA_BRIDGE_FLAGS
, BRIDGE_FLAGS_SELF
) ||
2293 nla_put_u16(skb
, IFLA_BRIDGE_MODE
, mode
)) {
2294 nla_nest_cancel(skb
, br_afspec
);
2295 goto nla_put_failure
;
2297 nla_nest_end(skb
, br_afspec
);
2299 return nlmsg_end(skb
, nlh
);
2301 nlmsg_cancel(skb
, nlh
);
2304 EXPORT_SYMBOL(ndo_dflt_bridge_getlink
);
2306 static int rtnl_bridge_getlink(struct sk_buff
*skb
, struct netlink_callback
*cb
)
2308 struct net
*net
= sock_net(skb
->sk
);
2309 struct net_device
*dev
;
2311 u32 portid
= NETLINK_CB(cb
->skb
).portid
;
2312 u32 seq
= cb
->nlh
->nlmsg_seq
;
2315 for_each_netdev_rcu(net
, dev
) {
2316 const struct net_device_ops
*ops
= dev
->netdev_ops
;
2317 struct net_device
*master
= dev
->master
;
2319 if (master
&& master
->netdev_ops
->ndo_bridge_getlink
) {
2320 if (idx
>= cb
->args
[0] &&
2321 master
->netdev_ops
->ndo_bridge_getlink(
2322 skb
, portid
, seq
, dev
) < 0)
2327 if (ops
->ndo_bridge_getlink
) {
2328 if (idx
>= cb
->args
[0] &&
2329 ops
->ndo_bridge_getlink(skb
, portid
, seq
, dev
) < 0)
2340 static inline size_t bridge_nlmsg_size(void)
2342 return NLMSG_ALIGN(sizeof(struct ifinfomsg
))
2343 + nla_total_size(IFNAMSIZ
) /* IFLA_IFNAME */
2344 + nla_total_size(MAX_ADDR_LEN
) /* IFLA_ADDRESS */
2345 + nla_total_size(sizeof(u32
)) /* IFLA_MASTER */
2346 + nla_total_size(sizeof(u32
)) /* IFLA_MTU */
2347 + nla_total_size(sizeof(u32
)) /* IFLA_LINK */
2348 + nla_total_size(sizeof(u32
)) /* IFLA_OPERSTATE */
2349 + nla_total_size(sizeof(u8
)) /* IFLA_PROTINFO */
2350 + nla_total_size(sizeof(struct nlattr
)) /* IFLA_AF_SPEC */
2351 + nla_total_size(sizeof(u16
)) /* IFLA_BRIDGE_FLAGS */
2352 + nla_total_size(sizeof(u16
)); /* IFLA_BRIDGE_MODE */
2355 static int rtnl_bridge_notify(struct net_device
*dev
, u16 flags
)
2357 struct net
*net
= dev_net(dev
);
2358 struct net_device
*master
= dev
->master
;
2359 struct sk_buff
*skb
;
2360 int err
= -EOPNOTSUPP
;
2362 skb
= nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC
);
2368 if ((!flags
|| (flags
& BRIDGE_FLAGS_MASTER
)) &&
2369 master
&& master
->netdev_ops
->ndo_bridge_getlink
) {
2370 err
= master
->netdev_ops
->ndo_bridge_getlink(skb
, 0, 0, dev
);
2375 if ((flags
& BRIDGE_FLAGS_SELF
) &&
2376 dev
->netdev_ops
->ndo_bridge_getlink
) {
2377 err
= dev
->netdev_ops
->ndo_bridge_getlink(skb
, 0, 0, dev
);
2382 rtnl_notify(skb
, net
, 0, RTNLGRP_LINK
, NULL
, GFP_ATOMIC
);
2385 WARN_ON(err
== -EMSGSIZE
);
2387 rtnl_set_sk_err(net
, RTNLGRP_LINK
, err
);
2391 static int rtnl_bridge_setlink(struct sk_buff
*skb
, struct nlmsghdr
*nlh
,
2394 struct net
*net
= sock_net(skb
->sk
);
2395 struct ifinfomsg
*ifm
;
2396 struct net_device
*dev
;
2397 struct nlattr
*br_spec
, *attr
= NULL
;
2398 int rem
, err
= -EOPNOTSUPP
;
2399 u16 oflags
, flags
= 0;
2400 bool have_flags
= false;
2402 if (nlmsg_len(nlh
) < sizeof(*ifm
))
2405 ifm
= nlmsg_data(nlh
);
2406 if (ifm
->ifi_family
!= AF_BRIDGE
)
2407 return -EPFNOSUPPORT
;
2409 dev
= __dev_get_by_index(net
, ifm
->ifi_index
);
2411 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
2415 br_spec
= nlmsg_find_attr(nlh
, sizeof(struct ifinfomsg
), IFLA_AF_SPEC
);
2417 nla_for_each_nested(attr
, br_spec
, rem
) {
2418 if (nla_type(attr
) == IFLA_BRIDGE_FLAGS
) {
2420 flags
= nla_get_u16(attr
);
2428 if (!flags
|| (flags
& BRIDGE_FLAGS_MASTER
)) {
2430 !dev
->master
->netdev_ops
->ndo_bridge_setlink
) {
2435 err
= dev
->master
->netdev_ops
->ndo_bridge_setlink(dev
, nlh
);
2439 flags
&= ~BRIDGE_FLAGS_MASTER
;
2442 if ((flags
& BRIDGE_FLAGS_SELF
)) {
2443 if (!dev
->netdev_ops
->ndo_bridge_setlink
)
2446 err
= dev
->netdev_ops
->ndo_bridge_setlink(dev
, nlh
);
2449 flags
&= ~BRIDGE_FLAGS_SELF
;
2453 memcpy(nla_data(attr
), &flags
, sizeof(flags
));
2454 /* Generate event to notify upper layer of bridge change */
2456 err
= rtnl_bridge_notify(dev
, oflags
);
2461 /* Protected by RTNL sempahore. */
2462 static struct rtattr
**rta_buf
;
2463 static int rtattr_max
;
2465 /* Process one rtnetlink message. */
2467 static int rtnetlink_rcv_msg(struct sk_buff
*skb
, struct nlmsghdr
*nlh
)
2469 struct net
*net
= sock_net(skb
->sk
);
2470 rtnl_doit_func doit
;
2477 type
= nlh
->nlmsg_type
;
2483 /* All the messages must have at least 1 byte length */
2484 if (nlh
->nlmsg_len
< NLMSG_LENGTH(sizeof(struct rtgenmsg
)))
2487 family
= ((struct rtgenmsg
*)NLMSG_DATA(nlh
))->rtgen_family
;
2491 if (kind
!= 2 && !capable(CAP_NET_ADMIN
))
2494 if (kind
== 2 && nlh
->nlmsg_flags
&NLM_F_DUMP
) {
2496 rtnl_dumpit_func dumpit
;
2497 rtnl_calcit_func calcit
;
2498 u16 min_dump_alloc
= 0;
2500 dumpit
= rtnl_get_dumpit(family
, type
);
2503 calcit
= rtnl_get_calcit(family
, type
);
2505 min_dump_alloc
= calcit(skb
, nlh
);
2510 struct netlink_dump_control c
= {
2512 .min_dump_alloc
= min_dump_alloc
,
2514 err
= netlink_dump_start(rtnl
, skb
, nlh
, &c
);
2520 memset(rta_buf
, 0, (rtattr_max
* sizeof(struct rtattr
*)));
2522 min_len
= rtm_min
[sz_idx
];
2523 if (nlh
->nlmsg_len
< min_len
)
2526 if (nlh
->nlmsg_len
> min_len
) {
2527 int attrlen
= nlh
->nlmsg_len
- NLMSG_ALIGN(min_len
);
2528 struct rtattr
*attr
= (void *)nlh
+ NLMSG_ALIGN(min_len
);
2530 while (RTA_OK(attr
, attrlen
)) {
2531 unsigned int flavor
= attr
->rta_type
;
2533 if (flavor
> rta_max
[sz_idx
])
2535 rta_buf
[flavor
-1] = attr
;
2537 attr
= RTA_NEXT(attr
, attrlen
);
2541 doit
= rtnl_get_doit(family
, type
);
2545 return doit(skb
, nlh
, (void *)&rta_buf
[0]);
2548 static void rtnetlink_rcv(struct sk_buff
*skb
)
2551 netlink_rcv_skb(skb
, &rtnetlink_rcv_msg
);
2555 static int rtnetlink_event(struct notifier_block
*this, unsigned long event
, void *ptr
)
2557 struct net_device
*dev
= ptr
;
2563 case NETDEV_POST_INIT
:
2564 case NETDEV_REGISTER
:
2566 case NETDEV_PRE_TYPE_CHANGE
:
2567 case NETDEV_GOING_DOWN
:
2568 case NETDEV_UNREGISTER
:
2569 case NETDEV_UNREGISTER_FINAL
:
2570 case NETDEV_RELEASE
:
2574 rtmsg_ifinfo(RTM_NEWLINK
, dev
, 0);
2580 static struct notifier_block rtnetlink_dev_notifier
= {
2581 .notifier_call
= rtnetlink_event
,
2585 static int __net_init
rtnetlink_net_init(struct net
*net
)
2588 struct netlink_kernel_cfg cfg
= {
2589 .groups
= RTNLGRP_MAX
,
2590 .input
= rtnetlink_rcv
,
2591 .cb_mutex
= &rtnl_mutex
,
2592 .flags
= NL_CFG_F_NONROOT_RECV
,
2595 sk
= netlink_kernel_create(net
, NETLINK_ROUTE
, &cfg
);
2602 static void __net_exit
rtnetlink_net_exit(struct net
*net
)
2604 netlink_kernel_release(net
->rtnl
);
2608 static struct pernet_operations rtnetlink_net_ops
= {
2609 .init
= rtnetlink_net_init
,
2610 .exit
= rtnetlink_net_exit
,
2613 void __init
rtnetlink_init(void)
2618 for (i
= 0; i
< ARRAY_SIZE(rta_max
); i
++)
2619 if (rta_max
[i
] > rtattr_max
)
2620 rtattr_max
= rta_max
[i
];
2621 rta_buf
= kmalloc(rtattr_max
* sizeof(struct rtattr
*), GFP_KERNEL
);
2623 panic("rtnetlink_init: cannot allocate rta_buf\n");
2625 if (register_pernet_subsys(&rtnetlink_net_ops
))
2626 panic("rtnetlink_init: cannot initialize rtnetlink\n");
2628 register_netdevice_notifier(&rtnetlink_dev_notifier
);
2630 rtnl_register(PF_UNSPEC
, RTM_GETLINK
, rtnl_getlink
,
2631 rtnl_dump_ifinfo
, rtnl_calcit
);
2632 rtnl_register(PF_UNSPEC
, RTM_SETLINK
, rtnl_setlink
, NULL
, NULL
);
2633 rtnl_register(PF_UNSPEC
, RTM_NEWLINK
, rtnl_newlink
, NULL
, NULL
);
2634 rtnl_register(PF_UNSPEC
, RTM_DELLINK
, rtnl_dellink
, NULL
, NULL
);
2636 rtnl_register(PF_UNSPEC
, RTM_GETADDR
, NULL
, rtnl_dump_all
, NULL
);
2637 rtnl_register(PF_UNSPEC
, RTM_GETROUTE
, NULL
, rtnl_dump_all
, NULL
);
2639 rtnl_register(PF_BRIDGE
, RTM_NEWNEIGH
, rtnl_fdb_add
, NULL
, NULL
);
2640 rtnl_register(PF_BRIDGE
, RTM_DELNEIGH
, rtnl_fdb_del
, NULL
, NULL
);
2641 rtnl_register(PF_BRIDGE
, RTM_GETNEIGH
, NULL
, rtnl_fdb_dump
, NULL
);
2643 rtnl_register(PF_BRIDGE
, RTM_GETLINK
, NULL
, rtnl_bridge_getlink
, NULL
);
2644 rtnl_register(PF_BRIDGE
, RTM_SETLINK
, rtnl_bridge_setlink
, NULL
, NULL
);