Merge tag 'iommu-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/arm64/linux
[linux/fpc-iii.git] / net / core / rtnetlink.c
blobbb0596c41b3efb8b86de22797891c0e88fc9cbb4
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3 * INET An implementation of the TCP/IP protocol suite for the LINUX
4 * operating system. INET is implemented using the BSD Socket
5 * interface as the means of communication with the user level.
7 * Routing netlink socket interface: protocol independent part.
9 * Authors: Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
11 * Fixes:
12 * Vitaly E. Lavrov RTA_OK arithmetics was wrong.
15 #include <linux/bitops.h>
16 #include <linux/errno.h>
17 #include <linux/module.h>
18 #include <linux/types.h>
19 #include <linux/socket.h>
20 #include <linux/kernel.h>
21 #include <linux/timer.h>
22 #include <linux/string.h>
23 #include <linux/sockios.h>
24 #include <linux/net.h>
25 #include <linux/fcntl.h>
26 #include <linux/mm.h>
27 #include <linux/slab.h>
28 #include <linux/interrupt.h>
29 #include <linux/capability.h>
30 #include <linux/skbuff.h>
31 #include <linux/init.h>
32 #include <linux/security.h>
33 #include <linux/mutex.h>
34 #include <linux/if_addr.h>
35 #include <linux/if_bridge.h>
36 #include <linux/if_vlan.h>
37 #include <linux/pci.h>
38 #include <linux/etherdevice.h>
39 #include <linux/bpf.h>
41 #include <linux/uaccess.h>
43 #include <linux/inet.h>
44 #include <linux/netdevice.h>
45 #include <net/ip.h>
46 #include <net/protocol.h>
47 #include <net/arp.h>
48 #include <net/route.h>
49 #include <net/udp.h>
50 #include <net/tcp.h>
51 #include <net/sock.h>
52 #include <net/pkt_sched.h>
53 #include <net/fib_rules.h>
54 #include <net/rtnetlink.h>
55 #include <net/net_namespace.h>
57 #define RTNL_MAX_TYPE 50
58 #define RTNL_SLAVE_MAX_TYPE 36
60 struct rtnl_link {
61 rtnl_doit_func doit;
62 rtnl_dumpit_func dumpit;
63 struct module *owner;
64 unsigned int flags;
65 struct rcu_head rcu;
68 static DEFINE_MUTEX(rtnl_mutex);
70 void rtnl_lock(void)
72 mutex_lock(&rtnl_mutex);
74 EXPORT_SYMBOL(rtnl_lock);
76 int rtnl_lock_killable(void)
78 return mutex_lock_killable(&rtnl_mutex);
80 EXPORT_SYMBOL(rtnl_lock_killable);
82 static struct sk_buff *defer_kfree_skb_list;
83 void rtnl_kfree_skbs(struct sk_buff *head, struct sk_buff *tail)
85 if (head && tail) {
86 tail->next = defer_kfree_skb_list;
87 defer_kfree_skb_list = head;
90 EXPORT_SYMBOL(rtnl_kfree_skbs);
92 void __rtnl_unlock(void)
94 struct sk_buff *head = defer_kfree_skb_list;
96 defer_kfree_skb_list = NULL;
98 mutex_unlock(&rtnl_mutex);
100 while (head) {
101 struct sk_buff *next = head->next;
103 kfree_skb(head);
104 cond_resched();
105 head = next;
109 void rtnl_unlock(void)
111 /* This fellow will unlock it for us. */
112 netdev_run_todo();
114 EXPORT_SYMBOL(rtnl_unlock);
116 int rtnl_trylock(void)
118 return mutex_trylock(&rtnl_mutex);
120 EXPORT_SYMBOL(rtnl_trylock);
122 int rtnl_is_locked(void)
124 return mutex_is_locked(&rtnl_mutex);
126 EXPORT_SYMBOL(rtnl_is_locked);
128 bool refcount_dec_and_rtnl_lock(refcount_t *r)
130 return refcount_dec_and_mutex_lock(r, &rtnl_mutex);
132 EXPORT_SYMBOL(refcount_dec_and_rtnl_lock);
134 #ifdef CONFIG_PROVE_LOCKING
135 bool lockdep_rtnl_is_held(void)
137 return lockdep_is_held(&rtnl_mutex);
139 EXPORT_SYMBOL(lockdep_rtnl_is_held);
140 #endif /* #ifdef CONFIG_PROVE_LOCKING */
142 static struct rtnl_link __rcu *__rcu *rtnl_msg_handlers[RTNL_FAMILY_MAX + 1];
144 static inline int rtm_msgindex(int msgtype)
146 int msgindex = msgtype - RTM_BASE;
149 * msgindex < 0 implies someone tried to register a netlink
150 * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
151 * the message type has not been added to linux/rtnetlink.h
153 BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
155 return msgindex;
158 static struct rtnl_link *rtnl_get_link(int protocol, int msgtype)
160 struct rtnl_link __rcu **tab;
162 if (protocol >= ARRAY_SIZE(rtnl_msg_handlers))
163 protocol = PF_UNSPEC;
165 tab = rcu_dereference_rtnl(rtnl_msg_handlers[protocol]);
166 if (!tab)
167 tab = rcu_dereference_rtnl(rtnl_msg_handlers[PF_UNSPEC]);
169 return rcu_dereference_rtnl(tab[msgtype]);
172 static int rtnl_register_internal(struct module *owner,
173 int protocol, int msgtype,
174 rtnl_doit_func doit, rtnl_dumpit_func dumpit,
175 unsigned int flags)
177 struct rtnl_link *link, *old;
178 struct rtnl_link __rcu **tab;
179 int msgindex;
180 int ret = -ENOBUFS;
182 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
183 msgindex = rtm_msgindex(msgtype);
185 rtnl_lock();
186 tab = rtnl_dereference(rtnl_msg_handlers[protocol]);
187 if (tab == NULL) {
188 tab = kcalloc(RTM_NR_MSGTYPES, sizeof(void *), GFP_KERNEL);
189 if (!tab)
190 goto unlock;
192 /* ensures we see the 0 stores */
193 rcu_assign_pointer(rtnl_msg_handlers[protocol], tab);
196 old = rtnl_dereference(tab[msgindex]);
197 if (old) {
198 link = kmemdup(old, sizeof(*old), GFP_KERNEL);
199 if (!link)
200 goto unlock;
201 } else {
202 link = kzalloc(sizeof(*link), GFP_KERNEL);
203 if (!link)
204 goto unlock;
207 WARN_ON(link->owner && link->owner != owner);
208 link->owner = owner;
210 WARN_ON(doit && link->doit && link->doit != doit);
211 if (doit)
212 link->doit = doit;
213 WARN_ON(dumpit && link->dumpit && link->dumpit != dumpit);
214 if (dumpit)
215 link->dumpit = dumpit;
217 link->flags |= flags;
219 /* publish protocol:msgtype */
220 rcu_assign_pointer(tab[msgindex], link);
221 ret = 0;
222 if (old)
223 kfree_rcu(old, rcu);
224 unlock:
225 rtnl_unlock();
226 return ret;
230 * rtnl_register_module - Register a rtnetlink message type
232 * @owner: module registering the hook (THIS_MODULE)
233 * @protocol: Protocol family or PF_UNSPEC
234 * @msgtype: rtnetlink message type
235 * @doit: Function pointer called for each request message
236 * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
237 * @flags: rtnl_link_flags to modifiy behaviour of doit/dumpit functions
239 * Like rtnl_register, but for use by removable modules.
241 int rtnl_register_module(struct module *owner,
242 int protocol, int msgtype,
243 rtnl_doit_func doit, rtnl_dumpit_func dumpit,
244 unsigned int flags)
246 return rtnl_register_internal(owner, protocol, msgtype,
247 doit, dumpit, flags);
249 EXPORT_SYMBOL_GPL(rtnl_register_module);
252 * rtnl_register - Register a rtnetlink message type
253 * @protocol: Protocol family or PF_UNSPEC
254 * @msgtype: rtnetlink message type
255 * @doit: Function pointer called for each request message
256 * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
257 * @flags: rtnl_link_flags to modifiy behaviour of doit/dumpit functions
259 * Registers the specified function pointers (at least one of them has
260 * to be non-NULL) to be called whenever a request message for the
261 * specified protocol family and message type is received.
263 * The special protocol family PF_UNSPEC may be used to define fallback
264 * function pointers for the case when no entry for the specific protocol
265 * family exists.
267 void rtnl_register(int protocol, int msgtype,
268 rtnl_doit_func doit, rtnl_dumpit_func dumpit,
269 unsigned int flags)
271 int err;
273 err = rtnl_register_internal(NULL, protocol, msgtype, doit, dumpit,
274 flags);
275 if (err)
276 pr_err("Unable to register rtnetlink message handler, "
277 "protocol = %d, message type = %d\n", protocol, msgtype);
281 * rtnl_unregister - Unregister a rtnetlink message type
282 * @protocol: Protocol family or PF_UNSPEC
283 * @msgtype: rtnetlink message type
285 * Returns 0 on success or a negative error code.
287 int rtnl_unregister(int protocol, int msgtype)
289 struct rtnl_link __rcu **tab;
290 struct rtnl_link *link;
291 int msgindex;
293 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
294 msgindex = rtm_msgindex(msgtype);
296 rtnl_lock();
297 tab = rtnl_dereference(rtnl_msg_handlers[protocol]);
298 if (!tab) {
299 rtnl_unlock();
300 return -ENOENT;
303 link = rtnl_dereference(tab[msgindex]);
304 rcu_assign_pointer(tab[msgindex], NULL);
305 rtnl_unlock();
307 kfree_rcu(link, rcu);
309 return 0;
311 EXPORT_SYMBOL_GPL(rtnl_unregister);
314 * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
315 * @protocol : Protocol family or PF_UNSPEC
317 * Identical to calling rtnl_unregster() for all registered message types
318 * of a certain protocol family.
320 void rtnl_unregister_all(int protocol)
322 struct rtnl_link __rcu **tab;
323 struct rtnl_link *link;
324 int msgindex;
326 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
328 rtnl_lock();
329 tab = rtnl_dereference(rtnl_msg_handlers[protocol]);
330 if (!tab) {
331 rtnl_unlock();
332 return;
334 RCU_INIT_POINTER(rtnl_msg_handlers[protocol], NULL);
335 for (msgindex = 0; msgindex < RTM_NR_MSGTYPES; msgindex++) {
336 link = rtnl_dereference(tab[msgindex]);
337 if (!link)
338 continue;
340 rcu_assign_pointer(tab[msgindex], NULL);
341 kfree_rcu(link, rcu);
343 rtnl_unlock();
345 synchronize_net();
347 kfree(tab);
349 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
351 static LIST_HEAD(link_ops);
353 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
355 const struct rtnl_link_ops *ops;
357 list_for_each_entry(ops, &link_ops, list) {
358 if (!strcmp(ops->kind, kind))
359 return ops;
361 return NULL;
365 * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
366 * @ops: struct rtnl_link_ops * to register
368 * The caller must hold the rtnl_mutex. This function should be used
369 * by drivers that create devices during module initialization. It
370 * must be called before registering the devices.
372 * Returns 0 on success or a negative error code.
374 int __rtnl_link_register(struct rtnl_link_ops *ops)
376 if (rtnl_link_ops_get(ops->kind))
377 return -EEXIST;
379 /* The check for setup is here because if ops
380 * does not have that filled up, it is not possible
381 * to use the ops for creating device. So do not
382 * fill up dellink as well. That disables rtnl_dellink.
384 if (ops->setup && !ops->dellink)
385 ops->dellink = unregister_netdevice_queue;
387 list_add_tail(&ops->list, &link_ops);
388 return 0;
390 EXPORT_SYMBOL_GPL(__rtnl_link_register);
393 * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
394 * @ops: struct rtnl_link_ops * to register
396 * Returns 0 on success or a negative error code.
398 int rtnl_link_register(struct rtnl_link_ops *ops)
400 int err;
402 /* Sanity-check max sizes to avoid stack buffer overflow. */
403 if (WARN_ON(ops->maxtype > RTNL_MAX_TYPE ||
404 ops->slave_maxtype > RTNL_SLAVE_MAX_TYPE))
405 return -EINVAL;
407 rtnl_lock();
408 err = __rtnl_link_register(ops);
409 rtnl_unlock();
410 return err;
412 EXPORT_SYMBOL_GPL(rtnl_link_register);
414 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
416 struct net_device *dev;
417 LIST_HEAD(list_kill);
419 for_each_netdev(net, dev) {
420 if (dev->rtnl_link_ops == ops)
421 ops->dellink(dev, &list_kill);
423 unregister_netdevice_many(&list_kill);
427 * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
428 * @ops: struct rtnl_link_ops * to unregister
430 * The caller must hold the rtnl_mutex and guarantee net_namespace_list
431 * integrity (hold pernet_ops_rwsem for writing to close the race
432 * with setup_net() and cleanup_net()).
434 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
436 struct net *net;
438 for_each_net(net) {
439 __rtnl_kill_links(net, ops);
441 list_del(&ops->list);
443 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
445 /* Return with the rtnl_lock held when there are no network
446 * devices unregistering in any network namespace.
448 static void rtnl_lock_unregistering_all(void)
450 struct net *net;
451 bool unregistering;
452 DEFINE_WAIT_FUNC(wait, woken_wake_function);
454 add_wait_queue(&netdev_unregistering_wq, &wait);
455 for (;;) {
456 unregistering = false;
457 rtnl_lock();
458 /* We held write locked pernet_ops_rwsem, and parallel
459 * setup_net() and cleanup_net() are not possible.
461 for_each_net(net) {
462 if (net->dev_unreg_count > 0) {
463 unregistering = true;
464 break;
467 if (!unregistering)
468 break;
469 __rtnl_unlock();
471 wait_woken(&wait, TASK_UNINTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
473 remove_wait_queue(&netdev_unregistering_wq, &wait);
477 * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
478 * @ops: struct rtnl_link_ops * to unregister
480 void rtnl_link_unregister(struct rtnl_link_ops *ops)
482 /* Close the race with setup_net() and cleanup_net() */
483 down_write(&pernet_ops_rwsem);
484 rtnl_lock_unregistering_all();
485 __rtnl_link_unregister(ops);
486 rtnl_unlock();
487 up_write(&pernet_ops_rwsem);
489 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
491 static size_t rtnl_link_get_slave_info_data_size(const struct net_device *dev)
493 struct net_device *master_dev;
494 const struct rtnl_link_ops *ops;
495 size_t size = 0;
497 rcu_read_lock();
499 master_dev = netdev_master_upper_dev_get_rcu((struct net_device *)dev);
500 if (!master_dev)
501 goto out;
503 ops = master_dev->rtnl_link_ops;
504 if (!ops || !ops->get_slave_size)
505 goto out;
506 /* IFLA_INFO_SLAVE_DATA + nested data */
507 size = nla_total_size(sizeof(struct nlattr)) +
508 ops->get_slave_size(master_dev, dev);
510 out:
511 rcu_read_unlock();
512 return size;
515 static size_t rtnl_link_get_size(const struct net_device *dev)
517 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
518 size_t size;
520 if (!ops)
521 return 0;
523 size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
524 nla_total_size(strlen(ops->kind) + 1); /* IFLA_INFO_KIND */
526 if (ops->get_size)
527 /* IFLA_INFO_DATA + nested data */
528 size += nla_total_size(sizeof(struct nlattr)) +
529 ops->get_size(dev);
531 if (ops->get_xstats_size)
532 /* IFLA_INFO_XSTATS */
533 size += nla_total_size(ops->get_xstats_size(dev));
535 size += rtnl_link_get_slave_info_data_size(dev);
537 return size;
540 static LIST_HEAD(rtnl_af_ops);
542 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
544 const struct rtnl_af_ops *ops;
546 list_for_each_entry_rcu(ops, &rtnl_af_ops, list) {
547 if (ops->family == family)
548 return ops;
551 return NULL;
555 * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
556 * @ops: struct rtnl_af_ops * to register
558 * Returns 0 on success or a negative error code.
560 void rtnl_af_register(struct rtnl_af_ops *ops)
562 rtnl_lock();
563 list_add_tail_rcu(&ops->list, &rtnl_af_ops);
564 rtnl_unlock();
566 EXPORT_SYMBOL_GPL(rtnl_af_register);
569 * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
570 * @ops: struct rtnl_af_ops * to unregister
572 void rtnl_af_unregister(struct rtnl_af_ops *ops)
574 rtnl_lock();
575 list_del_rcu(&ops->list);
576 rtnl_unlock();
578 synchronize_rcu();
580 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
582 static size_t rtnl_link_get_af_size(const struct net_device *dev,
583 u32 ext_filter_mask)
585 struct rtnl_af_ops *af_ops;
586 size_t size;
588 /* IFLA_AF_SPEC */
589 size = nla_total_size(sizeof(struct nlattr));
591 rcu_read_lock();
592 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
593 if (af_ops->get_link_af_size) {
594 /* AF_* + nested data */
595 size += nla_total_size(sizeof(struct nlattr)) +
596 af_ops->get_link_af_size(dev, ext_filter_mask);
599 rcu_read_unlock();
601 return size;
604 static bool rtnl_have_link_slave_info(const struct net_device *dev)
606 struct net_device *master_dev;
607 bool ret = false;
609 rcu_read_lock();
611 master_dev = netdev_master_upper_dev_get_rcu((struct net_device *)dev);
612 if (master_dev && master_dev->rtnl_link_ops)
613 ret = true;
614 rcu_read_unlock();
615 return ret;
618 static int rtnl_link_slave_info_fill(struct sk_buff *skb,
619 const struct net_device *dev)
621 struct net_device *master_dev;
622 const struct rtnl_link_ops *ops;
623 struct nlattr *slave_data;
624 int err;
626 master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
627 if (!master_dev)
628 return 0;
629 ops = master_dev->rtnl_link_ops;
630 if (!ops)
631 return 0;
632 if (nla_put_string(skb, IFLA_INFO_SLAVE_KIND, ops->kind) < 0)
633 return -EMSGSIZE;
634 if (ops->fill_slave_info) {
635 slave_data = nla_nest_start_noflag(skb, IFLA_INFO_SLAVE_DATA);
636 if (!slave_data)
637 return -EMSGSIZE;
638 err = ops->fill_slave_info(skb, master_dev, dev);
639 if (err < 0)
640 goto err_cancel_slave_data;
641 nla_nest_end(skb, slave_data);
643 return 0;
645 err_cancel_slave_data:
646 nla_nest_cancel(skb, slave_data);
647 return err;
650 static int rtnl_link_info_fill(struct sk_buff *skb,
651 const struct net_device *dev)
653 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
654 struct nlattr *data;
655 int err;
657 if (!ops)
658 return 0;
659 if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
660 return -EMSGSIZE;
661 if (ops->fill_xstats) {
662 err = ops->fill_xstats(skb, dev);
663 if (err < 0)
664 return err;
666 if (ops->fill_info) {
667 data = nla_nest_start_noflag(skb, IFLA_INFO_DATA);
668 if (data == NULL)
669 return -EMSGSIZE;
670 err = ops->fill_info(skb, dev);
671 if (err < 0)
672 goto err_cancel_data;
673 nla_nest_end(skb, data);
675 return 0;
677 err_cancel_data:
678 nla_nest_cancel(skb, data);
679 return err;
682 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
684 struct nlattr *linkinfo;
685 int err = -EMSGSIZE;
687 linkinfo = nla_nest_start_noflag(skb, IFLA_LINKINFO);
688 if (linkinfo == NULL)
689 goto out;
691 err = rtnl_link_info_fill(skb, dev);
692 if (err < 0)
693 goto err_cancel_link;
695 err = rtnl_link_slave_info_fill(skb, dev);
696 if (err < 0)
697 goto err_cancel_link;
699 nla_nest_end(skb, linkinfo);
700 return 0;
702 err_cancel_link:
703 nla_nest_cancel(skb, linkinfo);
704 out:
705 return err;
708 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
710 struct sock *rtnl = net->rtnl;
711 int err = 0;
713 NETLINK_CB(skb).dst_group = group;
714 if (echo)
715 refcount_inc(&skb->users);
716 netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
717 if (echo)
718 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
719 return err;
722 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
724 struct sock *rtnl = net->rtnl;
726 return nlmsg_unicast(rtnl, skb, pid);
728 EXPORT_SYMBOL(rtnl_unicast);
730 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
731 struct nlmsghdr *nlh, gfp_t flags)
733 struct sock *rtnl = net->rtnl;
734 int report = 0;
736 if (nlh)
737 report = nlmsg_report(nlh);
739 nlmsg_notify(rtnl, skb, pid, group, report, flags);
741 EXPORT_SYMBOL(rtnl_notify);
743 void rtnl_set_sk_err(struct net *net, u32 group, int error)
745 struct sock *rtnl = net->rtnl;
747 netlink_set_err(rtnl, 0, group, error);
749 EXPORT_SYMBOL(rtnl_set_sk_err);
751 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
753 struct nlattr *mx;
754 int i, valid = 0;
756 /* nothing is dumped for dst_default_metrics, so just skip the loop */
757 if (metrics == dst_default_metrics.metrics)
758 return 0;
760 mx = nla_nest_start_noflag(skb, RTA_METRICS);
761 if (mx == NULL)
762 return -ENOBUFS;
764 for (i = 0; i < RTAX_MAX; i++) {
765 if (metrics[i]) {
766 if (i == RTAX_CC_ALGO - 1) {
767 char tmp[TCP_CA_NAME_MAX], *name;
769 name = tcp_ca_get_name_by_key(metrics[i], tmp);
770 if (!name)
771 continue;
772 if (nla_put_string(skb, i + 1, name))
773 goto nla_put_failure;
774 } else if (i == RTAX_FEATURES - 1) {
775 u32 user_features = metrics[i] & RTAX_FEATURE_MASK;
777 if (!user_features)
778 continue;
779 BUILD_BUG_ON(RTAX_FEATURE_MASK & DST_FEATURE_MASK);
780 if (nla_put_u32(skb, i + 1, user_features))
781 goto nla_put_failure;
782 } else {
783 if (nla_put_u32(skb, i + 1, metrics[i]))
784 goto nla_put_failure;
786 valid++;
790 if (!valid) {
791 nla_nest_cancel(skb, mx);
792 return 0;
795 return nla_nest_end(skb, mx);
797 nla_put_failure:
798 nla_nest_cancel(skb, mx);
799 return -EMSGSIZE;
801 EXPORT_SYMBOL(rtnetlink_put_metrics);
803 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
804 long expires, u32 error)
806 struct rta_cacheinfo ci = {
807 .rta_error = error,
808 .rta_id = id,
811 if (dst) {
812 ci.rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse);
813 ci.rta_used = dst->__use;
814 ci.rta_clntref = atomic_read(&dst->__refcnt);
816 if (expires) {
817 unsigned long clock;
819 clock = jiffies_to_clock_t(abs(expires));
820 clock = min_t(unsigned long, clock, INT_MAX);
821 ci.rta_expires = (expires > 0) ? clock : -clock;
823 return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
825 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
827 static void set_operstate(struct net_device *dev, unsigned char transition)
829 unsigned char operstate = dev->operstate;
831 switch (transition) {
832 case IF_OPER_UP:
833 if ((operstate == IF_OPER_DORMANT ||
834 operstate == IF_OPER_TESTING ||
835 operstate == IF_OPER_UNKNOWN) &&
836 !netif_dormant(dev) && !netif_testing(dev))
837 operstate = IF_OPER_UP;
838 break;
840 case IF_OPER_TESTING:
841 if (operstate == IF_OPER_UP ||
842 operstate == IF_OPER_UNKNOWN)
843 operstate = IF_OPER_TESTING;
844 break;
846 case IF_OPER_DORMANT:
847 if (operstate == IF_OPER_UP ||
848 operstate == IF_OPER_UNKNOWN)
849 operstate = IF_OPER_DORMANT;
850 break;
853 if (dev->operstate != operstate) {
854 write_lock_bh(&dev_base_lock);
855 dev->operstate = operstate;
856 write_unlock_bh(&dev_base_lock);
857 netdev_state_change(dev);
861 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
863 return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
864 (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
867 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
868 const struct ifinfomsg *ifm)
870 unsigned int flags = ifm->ifi_flags;
872 /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
873 if (ifm->ifi_change)
874 flags = (flags & ifm->ifi_change) |
875 (rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
877 return flags;
880 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
881 const struct rtnl_link_stats64 *b)
883 a->rx_packets = b->rx_packets;
884 a->tx_packets = b->tx_packets;
885 a->rx_bytes = b->rx_bytes;
886 a->tx_bytes = b->tx_bytes;
887 a->rx_errors = b->rx_errors;
888 a->tx_errors = b->tx_errors;
889 a->rx_dropped = b->rx_dropped;
890 a->tx_dropped = b->tx_dropped;
892 a->multicast = b->multicast;
893 a->collisions = b->collisions;
895 a->rx_length_errors = b->rx_length_errors;
896 a->rx_over_errors = b->rx_over_errors;
897 a->rx_crc_errors = b->rx_crc_errors;
898 a->rx_frame_errors = b->rx_frame_errors;
899 a->rx_fifo_errors = b->rx_fifo_errors;
900 a->rx_missed_errors = b->rx_missed_errors;
902 a->tx_aborted_errors = b->tx_aborted_errors;
903 a->tx_carrier_errors = b->tx_carrier_errors;
904 a->tx_fifo_errors = b->tx_fifo_errors;
905 a->tx_heartbeat_errors = b->tx_heartbeat_errors;
906 a->tx_window_errors = b->tx_window_errors;
908 a->rx_compressed = b->rx_compressed;
909 a->tx_compressed = b->tx_compressed;
911 a->rx_nohandler = b->rx_nohandler;
914 /* All VF info */
915 static inline int rtnl_vfinfo_size(const struct net_device *dev,
916 u32 ext_filter_mask)
918 if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF)) {
919 int num_vfs = dev_num_vf(dev->dev.parent);
920 size_t size = nla_total_size(0);
921 size += num_vfs *
922 (nla_total_size(0) +
923 nla_total_size(sizeof(struct ifla_vf_mac)) +
924 nla_total_size(sizeof(struct ifla_vf_broadcast)) +
925 nla_total_size(sizeof(struct ifla_vf_vlan)) +
926 nla_total_size(0) + /* nest IFLA_VF_VLAN_LIST */
927 nla_total_size(MAX_VLAN_LIST_LEN *
928 sizeof(struct ifla_vf_vlan_info)) +
929 nla_total_size(sizeof(struct ifla_vf_spoofchk)) +
930 nla_total_size(sizeof(struct ifla_vf_tx_rate)) +
931 nla_total_size(sizeof(struct ifla_vf_rate)) +
932 nla_total_size(sizeof(struct ifla_vf_link_state)) +
933 nla_total_size(sizeof(struct ifla_vf_rss_query_en)) +
934 nla_total_size(0) + /* nest IFLA_VF_STATS */
935 /* IFLA_VF_STATS_RX_PACKETS */
936 nla_total_size_64bit(sizeof(__u64)) +
937 /* IFLA_VF_STATS_TX_PACKETS */
938 nla_total_size_64bit(sizeof(__u64)) +
939 /* IFLA_VF_STATS_RX_BYTES */
940 nla_total_size_64bit(sizeof(__u64)) +
941 /* IFLA_VF_STATS_TX_BYTES */
942 nla_total_size_64bit(sizeof(__u64)) +
943 /* IFLA_VF_STATS_BROADCAST */
944 nla_total_size_64bit(sizeof(__u64)) +
945 /* IFLA_VF_STATS_MULTICAST */
946 nla_total_size_64bit(sizeof(__u64)) +
947 /* IFLA_VF_STATS_RX_DROPPED */
948 nla_total_size_64bit(sizeof(__u64)) +
949 /* IFLA_VF_STATS_TX_DROPPED */
950 nla_total_size_64bit(sizeof(__u64)) +
951 nla_total_size(sizeof(struct ifla_vf_trust)));
952 return size;
953 } else
954 return 0;
957 static size_t rtnl_port_size(const struct net_device *dev,
958 u32 ext_filter_mask)
960 size_t port_size = nla_total_size(4) /* PORT_VF */
961 + nla_total_size(PORT_PROFILE_MAX) /* PORT_PROFILE */
962 + nla_total_size(PORT_UUID_MAX) /* PORT_INSTANCE_UUID */
963 + nla_total_size(PORT_UUID_MAX) /* PORT_HOST_UUID */
964 + nla_total_size(1) /* PROT_VDP_REQUEST */
965 + nla_total_size(2); /* PORT_VDP_RESPONSE */
966 size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
967 size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
968 + port_size;
969 size_t port_self_size = nla_total_size(sizeof(struct nlattr))
970 + port_size;
972 if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
973 !(ext_filter_mask & RTEXT_FILTER_VF))
974 return 0;
975 if (dev_num_vf(dev->dev.parent))
976 return port_self_size + vf_ports_size +
977 vf_port_size * dev_num_vf(dev->dev.parent);
978 else
979 return port_self_size;
982 static size_t rtnl_xdp_size(void)
984 size_t xdp_size = nla_total_size(0) + /* nest IFLA_XDP */
985 nla_total_size(1) + /* XDP_ATTACHED */
986 nla_total_size(4) + /* XDP_PROG_ID (or 1st mode) */
987 nla_total_size(4); /* XDP_<mode>_PROG_ID */
989 return xdp_size;
992 static size_t rtnl_prop_list_size(const struct net_device *dev)
994 struct netdev_name_node *name_node;
995 size_t size;
997 if (list_empty(&dev->name_node->list))
998 return 0;
999 size = nla_total_size(0);
1000 list_for_each_entry(name_node, &dev->name_node->list, list)
1001 size += nla_total_size(ALTIFNAMSIZ);
1002 return size;
1005 static size_t rtnl_proto_down_size(const struct net_device *dev)
1007 size_t size = nla_total_size(1);
1009 if (dev->proto_down_reason)
1010 size += nla_total_size(0) + nla_total_size(4);
1012 return size;
1015 static noinline size_t if_nlmsg_size(const struct net_device *dev,
1016 u32 ext_filter_mask)
1018 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
1019 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
1020 + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
1021 + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
1022 + nla_total_size_64bit(sizeof(struct rtnl_link_ifmap))
1023 + nla_total_size(sizeof(struct rtnl_link_stats))
1024 + nla_total_size_64bit(sizeof(struct rtnl_link_stats64))
1025 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
1026 + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
1027 + nla_total_size(4) /* IFLA_TXQLEN */
1028 + nla_total_size(4) /* IFLA_WEIGHT */
1029 + nla_total_size(4) /* IFLA_MTU */
1030 + nla_total_size(4) /* IFLA_LINK */
1031 + nla_total_size(4) /* IFLA_MASTER */
1032 + nla_total_size(1) /* IFLA_CARRIER */
1033 + nla_total_size(4) /* IFLA_PROMISCUITY */
1034 + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
1035 + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
1036 + nla_total_size(4) /* IFLA_GSO_MAX_SEGS */
1037 + nla_total_size(4) /* IFLA_GSO_MAX_SIZE */
1038 + nla_total_size(1) /* IFLA_OPERSTATE */
1039 + nla_total_size(1) /* IFLA_LINKMODE */
1040 + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
1041 + nla_total_size(4) /* IFLA_LINK_NETNSID */
1042 + nla_total_size(4) /* IFLA_GROUP */
1043 + nla_total_size(ext_filter_mask
1044 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
1045 + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
1046 + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
1047 + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
1048 + rtnl_link_get_af_size(dev, ext_filter_mask) /* IFLA_AF_SPEC */
1049 + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */
1050 + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_SWITCH_ID */
1051 + nla_total_size(IFNAMSIZ) /* IFLA_PHYS_PORT_NAME */
1052 + rtnl_xdp_size() /* IFLA_XDP */
1053 + nla_total_size(4) /* IFLA_EVENT */
1054 + nla_total_size(4) /* IFLA_NEW_NETNSID */
1055 + nla_total_size(4) /* IFLA_NEW_IFINDEX */
1056 + rtnl_proto_down_size(dev) /* proto down */
1057 + nla_total_size(4) /* IFLA_TARGET_NETNSID */
1058 + nla_total_size(4) /* IFLA_CARRIER_UP_COUNT */
1059 + nla_total_size(4) /* IFLA_CARRIER_DOWN_COUNT */
1060 + nla_total_size(4) /* IFLA_MIN_MTU */
1061 + nla_total_size(4) /* IFLA_MAX_MTU */
1062 + rtnl_prop_list_size(dev)
1063 + nla_total_size(MAX_ADDR_LEN) /* IFLA_PERM_ADDRESS */
1064 + 0;
1067 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
1069 struct nlattr *vf_ports;
1070 struct nlattr *vf_port;
1071 int vf;
1072 int err;
1074 vf_ports = nla_nest_start_noflag(skb, IFLA_VF_PORTS);
1075 if (!vf_ports)
1076 return -EMSGSIZE;
1078 for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
1079 vf_port = nla_nest_start_noflag(skb, IFLA_VF_PORT);
1080 if (!vf_port)
1081 goto nla_put_failure;
1082 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
1083 goto nla_put_failure;
1084 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
1085 if (err == -EMSGSIZE)
1086 goto nla_put_failure;
1087 if (err) {
1088 nla_nest_cancel(skb, vf_port);
1089 continue;
1091 nla_nest_end(skb, vf_port);
1094 nla_nest_end(skb, vf_ports);
1096 return 0;
1098 nla_put_failure:
1099 nla_nest_cancel(skb, vf_ports);
1100 return -EMSGSIZE;
1103 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
1105 struct nlattr *port_self;
1106 int err;
1108 port_self = nla_nest_start_noflag(skb, IFLA_PORT_SELF);
1109 if (!port_self)
1110 return -EMSGSIZE;
1112 err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
1113 if (err) {
1114 nla_nest_cancel(skb, port_self);
1115 return (err == -EMSGSIZE) ? err : 0;
1118 nla_nest_end(skb, port_self);
1120 return 0;
1123 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
1124 u32 ext_filter_mask)
1126 int err;
1128 if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
1129 !(ext_filter_mask & RTEXT_FILTER_VF))
1130 return 0;
1132 err = rtnl_port_self_fill(skb, dev);
1133 if (err)
1134 return err;
1136 if (dev_num_vf(dev->dev.parent)) {
1137 err = rtnl_vf_ports_fill(skb, dev);
1138 if (err)
1139 return err;
1142 return 0;
1145 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
1147 int err;
1148 struct netdev_phys_item_id ppid;
1150 err = dev_get_phys_port_id(dev, &ppid);
1151 if (err) {
1152 if (err == -EOPNOTSUPP)
1153 return 0;
1154 return err;
1157 if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
1158 return -EMSGSIZE;
1160 return 0;
1163 static int rtnl_phys_port_name_fill(struct sk_buff *skb, struct net_device *dev)
1165 char name[IFNAMSIZ];
1166 int err;
1168 err = dev_get_phys_port_name(dev, name, sizeof(name));
1169 if (err) {
1170 if (err == -EOPNOTSUPP)
1171 return 0;
1172 return err;
1175 if (nla_put_string(skb, IFLA_PHYS_PORT_NAME, name))
1176 return -EMSGSIZE;
1178 return 0;
1181 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev)
1183 struct netdev_phys_item_id ppid = { };
1184 int err;
1186 err = dev_get_port_parent_id(dev, &ppid, false);
1187 if (err) {
1188 if (err == -EOPNOTSUPP)
1189 return 0;
1190 return err;
1193 if (nla_put(skb, IFLA_PHYS_SWITCH_ID, ppid.id_len, ppid.id))
1194 return -EMSGSIZE;
1196 return 0;
1199 static noinline_for_stack int rtnl_fill_stats(struct sk_buff *skb,
1200 struct net_device *dev)
1202 struct rtnl_link_stats64 *sp;
1203 struct nlattr *attr;
1205 attr = nla_reserve_64bit(skb, IFLA_STATS64,
1206 sizeof(struct rtnl_link_stats64), IFLA_PAD);
1207 if (!attr)
1208 return -EMSGSIZE;
1210 sp = nla_data(attr);
1211 dev_get_stats(dev, sp);
1213 attr = nla_reserve(skb, IFLA_STATS,
1214 sizeof(struct rtnl_link_stats));
1215 if (!attr)
1216 return -EMSGSIZE;
1218 copy_rtnl_link_stats(nla_data(attr), sp);
1220 return 0;
1223 static noinline_for_stack int rtnl_fill_vfinfo(struct sk_buff *skb,
1224 struct net_device *dev,
1225 int vfs_num,
1226 struct nlattr *vfinfo)
1228 struct ifla_vf_rss_query_en vf_rss_query_en;
1229 struct nlattr *vf, *vfstats, *vfvlanlist;
1230 struct ifla_vf_link_state vf_linkstate;
1231 struct ifla_vf_vlan_info vf_vlan_info;
1232 struct ifla_vf_spoofchk vf_spoofchk;
1233 struct ifla_vf_tx_rate vf_tx_rate;
1234 struct ifla_vf_stats vf_stats;
1235 struct ifla_vf_trust vf_trust;
1236 struct ifla_vf_vlan vf_vlan;
1237 struct ifla_vf_rate vf_rate;
1238 struct ifla_vf_mac vf_mac;
1239 struct ifla_vf_broadcast vf_broadcast;
1240 struct ifla_vf_info ivi;
1241 struct ifla_vf_guid node_guid;
1242 struct ifla_vf_guid port_guid;
1244 memset(&ivi, 0, sizeof(ivi));
1246 /* Not all SR-IOV capable drivers support the
1247 * spoofcheck and "RSS query enable" query. Preset to
1248 * -1 so the user space tool can detect that the driver
1249 * didn't report anything.
1251 ivi.spoofchk = -1;
1252 ivi.rss_query_en = -1;
1253 ivi.trusted = -1;
1254 /* The default value for VF link state is "auto"
1255 * IFLA_VF_LINK_STATE_AUTO which equals zero
1257 ivi.linkstate = 0;
1258 /* VLAN Protocol by default is 802.1Q */
1259 ivi.vlan_proto = htons(ETH_P_8021Q);
1260 if (dev->netdev_ops->ndo_get_vf_config(dev, vfs_num, &ivi))
1261 return 0;
1263 memset(&vf_vlan_info, 0, sizeof(vf_vlan_info));
1264 memset(&node_guid, 0, sizeof(node_guid));
1265 memset(&port_guid, 0, sizeof(port_guid));
1267 vf_mac.vf =
1268 vf_vlan.vf =
1269 vf_vlan_info.vf =
1270 vf_rate.vf =
1271 vf_tx_rate.vf =
1272 vf_spoofchk.vf =
1273 vf_linkstate.vf =
1274 vf_rss_query_en.vf =
1275 vf_trust.vf =
1276 node_guid.vf =
1277 port_guid.vf = ivi.vf;
1279 memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1280 memcpy(vf_broadcast.broadcast, dev->broadcast, dev->addr_len);
1281 vf_vlan.vlan = ivi.vlan;
1282 vf_vlan.qos = ivi.qos;
1283 vf_vlan_info.vlan = ivi.vlan;
1284 vf_vlan_info.qos = ivi.qos;
1285 vf_vlan_info.vlan_proto = ivi.vlan_proto;
1286 vf_tx_rate.rate = ivi.max_tx_rate;
1287 vf_rate.min_tx_rate = ivi.min_tx_rate;
1288 vf_rate.max_tx_rate = ivi.max_tx_rate;
1289 vf_spoofchk.setting = ivi.spoofchk;
1290 vf_linkstate.link_state = ivi.linkstate;
1291 vf_rss_query_en.setting = ivi.rss_query_en;
1292 vf_trust.setting = ivi.trusted;
1293 vf = nla_nest_start_noflag(skb, IFLA_VF_INFO);
1294 if (!vf)
1295 goto nla_put_vfinfo_failure;
1296 if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1297 nla_put(skb, IFLA_VF_BROADCAST, sizeof(vf_broadcast), &vf_broadcast) ||
1298 nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1299 nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1300 &vf_rate) ||
1301 nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1302 &vf_tx_rate) ||
1303 nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1304 &vf_spoofchk) ||
1305 nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1306 &vf_linkstate) ||
1307 nla_put(skb, IFLA_VF_RSS_QUERY_EN,
1308 sizeof(vf_rss_query_en),
1309 &vf_rss_query_en) ||
1310 nla_put(skb, IFLA_VF_TRUST,
1311 sizeof(vf_trust), &vf_trust))
1312 goto nla_put_vf_failure;
1314 if (dev->netdev_ops->ndo_get_vf_guid &&
1315 !dev->netdev_ops->ndo_get_vf_guid(dev, vfs_num, &node_guid,
1316 &port_guid)) {
1317 if (nla_put(skb, IFLA_VF_IB_NODE_GUID, sizeof(node_guid),
1318 &node_guid) ||
1319 nla_put(skb, IFLA_VF_IB_PORT_GUID, sizeof(port_guid),
1320 &port_guid))
1321 goto nla_put_vf_failure;
1323 vfvlanlist = nla_nest_start_noflag(skb, IFLA_VF_VLAN_LIST);
1324 if (!vfvlanlist)
1325 goto nla_put_vf_failure;
1326 if (nla_put(skb, IFLA_VF_VLAN_INFO, sizeof(vf_vlan_info),
1327 &vf_vlan_info)) {
1328 nla_nest_cancel(skb, vfvlanlist);
1329 goto nla_put_vf_failure;
1331 nla_nest_end(skb, vfvlanlist);
1332 memset(&vf_stats, 0, sizeof(vf_stats));
1333 if (dev->netdev_ops->ndo_get_vf_stats)
1334 dev->netdev_ops->ndo_get_vf_stats(dev, vfs_num,
1335 &vf_stats);
1336 vfstats = nla_nest_start_noflag(skb, IFLA_VF_STATS);
1337 if (!vfstats)
1338 goto nla_put_vf_failure;
1339 if (nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_PACKETS,
1340 vf_stats.rx_packets, IFLA_VF_STATS_PAD) ||
1341 nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_PACKETS,
1342 vf_stats.tx_packets, IFLA_VF_STATS_PAD) ||
1343 nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_BYTES,
1344 vf_stats.rx_bytes, IFLA_VF_STATS_PAD) ||
1345 nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_BYTES,
1346 vf_stats.tx_bytes, IFLA_VF_STATS_PAD) ||
1347 nla_put_u64_64bit(skb, IFLA_VF_STATS_BROADCAST,
1348 vf_stats.broadcast, IFLA_VF_STATS_PAD) ||
1349 nla_put_u64_64bit(skb, IFLA_VF_STATS_MULTICAST,
1350 vf_stats.multicast, IFLA_VF_STATS_PAD) ||
1351 nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_DROPPED,
1352 vf_stats.rx_dropped, IFLA_VF_STATS_PAD) ||
1353 nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_DROPPED,
1354 vf_stats.tx_dropped, IFLA_VF_STATS_PAD)) {
1355 nla_nest_cancel(skb, vfstats);
1356 goto nla_put_vf_failure;
1358 nla_nest_end(skb, vfstats);
1359 nla_nest_end(skb, vf);
1360 return 0;
1362 nla_put_vf_failure:
1363 nla_nest_cancel(skb, vf);
1364 nla_put_vfinfo_failure:
1365 nla_nest_cancel(skb, vfinfo);
1366 return -EMSGSIZE;
1369 static noinline_for_stack int rtnl_fill_vf(struct sk_buff *skb,
1370 struct net_device *dev,
1371 u32 ext_filter_mask)
1373 struct nlattr *vfinfo;
1374 int i, num_vfs;
1376 if (!dev->dev.parent || ((ext_filter_mask & RTEXT_FILTER_VF) == 0))
1377 return 0;
1379 num_vfs = dev_num_vf(dev->dev.parent);
1380 if (nla_put_u32(skb, IFLA_NUM_VF, num_vfs))
1381 return -EMSGSIZE;
1383 if (!dev->netdev_ops->ndo_get_vf_config)
1384 return 0;
1386 vfinfo = nla_nest_start_noflag(skb, IFLA_VFINFO_LIST);
1387 if (!vfinfo)
1388 return -EMSGSIZE;
1390 for (i = 0; i < num_vfs; i++) {
1391 if (rtnl_fill_vfinfo(skb, dev, i, vfinfo))
1392 return -EMSGSIZE;
1395 nla_nest_end(skb, vfinfo);
1396 return 0;
1399 static int rtnl_fill_link_ifmap(struct sk_buff *skb, struct net_device *dev)
1401 struct rtnl_link_ifmap map;
1403 memset(&map, 0, sizeof(map));
1404 map.mem_start = dev->mem_start;
1405 map.mem_end = dev->mem_end;
1406 map.base_addr = dev->base_addr;
1407 map.irq = dev->irq;
1408 map.dma = dev->dma;
1409 map.port = dev->if_port;
1411 if (nla_put_64bit(skb, IFLA_MAP, sizeof(map), &map, IFLA_PAD))
1412 return -EMSGSIZE;
1414 return 0;
1417 static u32 rtnl_xdp_prog_skb(struct net_device *dev)
1419 const struct bpf_prog *generic_xdp_prog;
1421 ASSERT_RTNL();
1423 generic_xdp_prog = rtnl_dereference(dev->xdp_prog);
1424 if (!generic_xdp_prog)
1425 return 0;
1426 return generic_xdp_prog->aux->id;
1429 static u32 rtnl_xdp_prog_drv(struct net_device *dev)
1431 return dev_xdp_prog_id(dev, XDP_MODE_DRV);
1434 static u32 rtnl_xdp_prog_hw(struct net_device *dev)
1436 return dev_xdp_prog_id(dev, XDP_MODE_HW);
1439 static int rtnl_xdp_report_one(struct sk_buff *skb, struct net_device *dev,
1440 u32 *prog_id, u8 *mode, u8 tgt_mode, u32 attr,
1441 u32 (*get_prog_id)(struct net_device *dev))
1443 u32 curr_id;
1444 int err;
1446 curr_id = get_prog_id(dev);
1447 if (!curr_id)
1448 return 0;
1450 *prog_id = curr_id;
1451 err = nla_put_u32(skb, attr, curr_id);
1452 if (err)
1453 return err;
1455 if (*mode != XDP_ATTACHED_NONE)
1456 *mode = XDP_ATTACHED_MULTI;
1457 else
1458 *mode = tgt_mode;
1460 return 0;
1463 static int rtnl_xdp_fill(struct sk_buff *skb, struct net_device *dev)
1465 struct nlattr *xdp;
1466 u32 prog_id;
1467 int err;
1468 u8 mode;
1470 xdp = nla_nest_start_noflag(skb, IFLA_XDP);
1471 if (!xdp)
1472 return -EMSGSIZE;
1474 prog_id = 0;
1475 mode = XDP_ATTACHED_NONE;
1476 err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_SKB,
1477 IFLA_XDP_SKB_PROG_ID, rtnl_xdp_prog_skb);
1478 if (err)
1479 goto err_cancel;
1480 err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_DRV,
1481 IFLA_XDP_DRV_PROG_ID, rtnl_xdp_prog_drv);
1482 if (err)
1483 goto err_cancel;
1484 err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_HW,
1485 IFLA_XDP_HW_PROG_ID, rtnl_xdp_prog_hw);
1486 if (err)
1487 goto err_cancel;
1489 err = nla_put_u8(skb, IFLA_XDP_ATTACHED, mode);
1490 if (err)
1491 goto err_cancel;
1493 if (prog_id && mode != XDP_ATTACHED_MULTI) {
1494 err = nla_put_u32(skb, IFLA_XDP_PROG_ID, prog_id);
1495 if (err)
1496 goto err_cancel;
1499 nla_nest_end(skb, xdp);
1500 return 0;
1502 err_cancel:
1503 nla_nest_cancel(skb, xdp);
1504 return err;
1507 static u32 rtnl_get_event(unsigned long event)
1509 u32 rtnl_event_type = IFLA_EVENT_NONE;
1511 switch (event) {
1512 case NETDEV_REBOOT:
1513 rtnl_event_type = IFLA_EVENT_REBOOT;
1514 break;
1515 case NETDEV_FEAT_CHANGE:
1516 rtnl_event_type = IFLA_EVENT_FEATURES;
1517 break;
1518 case NETDEV_BONDING_FAILOVER:
1519 rtnl_event_type = IFLA_EVENT_BONDING_FAILOVER;
1520 break;
1521 case NETDEV_NOTIFY_PEERS:
1522 rtnl_event_type = IFLA_EVENT_NOTIFY_PEERS;
1523 break;
1524 case NETDEV_RESEND_IGMP:
1525 rtnl_event_type = IFLA_EVENT_IGMP_RESEND;
1526 break;
1527 case NETDEV_CHANGEINFODATA:
1528 rtnl_event_type = IFLA_EVENT_BONDING_OPTIONS;
1529 break;
1530 default:
1531 break;
1534 return rtnl_event_type;
1537 static int put_master_ifindex(struct sk_buff *skb, struct net_device *dev)
1539 const struct net_device *upper_dev;
1540 int ret = 0;
1542 rcu_read_lock();
1544 upper_dev = netdev_master_upper_dev_get_rcu(dev);
1545 if (upper_dev)
1546 ret = nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex);
1548 rcu_read_unlock();
1549 return ret;
1552 static int nla_put_iflink(struct sk_buff *skb, const struct net_device *dev,
1553 bool force)
1555 int ifindex = dev_get_iflink(dev);
1557 if (force || dev->ifindex != ifindex)
1558 return nla_put_u32(skb, IFLA_LINK, ifindex);
1560 return 0;
1563 static noinline_for_stack int nla_put_ifalias(struct sk_buff *skb,
1564 struct net_device *dev)
1566 char buf[IFALIASZ];
1567 int ret;
1569 ret = dev_get_alias(dev, buf, sizeof(buf));
1570 return ret > 0 ? nla_put_string(skb, IFLA_IFALIAS, buf) : 0;
1573 static int rtnl_fill_link_netnsid(struct sk_buff *skb,
1574 const struct net_device *dev,
1575 struct net *src_net, gfp_t gfp)
1577 bool put_iflink = false;
1579 if (dev->rtnl_link_ops && dev->rtnl_link_ops->get_link_net) {
1580 struct net *link_net = dev->rtnl_link_ops->get_link_net(dev);
1582 if (!net_eq(dev_net(dev), link_net)) {
1583 int id = peernet2id_alloc(src_net, link_net, gfp);
1585 if (nla_put_s32(skb, IFLA_LINK_NETNSID, id))
1586 return -EMSGSIZE;
1588 put_iflink = true;
1592 return nla_put_iflink(skb, dev, put_iflink);
1595 static int rtnl_fill_link_af(struct sk_buff *skb,
1596 const struct net_device *dev,
1597 u32 ext_filter_mask)
1599 const struct rtnl_af_ops *af_ops;
1600 struct nlattr *af_spec;
1602 af_spec = nla_nest_start_noflag(skb, IFLA_AF_SPEC);
1603 if (!af_spec)
1604 return -EMSGSIZE;
1606 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
1607 struct nlattr *af;
1608 int err;
1610 if (!af_ops->fill_link_af)
1611 continue;
1613 af = nla_nest_start_noflag(skb, af_ops->family);
1614 if (!af)
1615 return -EMSGSIZE;
1617 err = af_ops->fill_link_af(skb, dev, ext_filter_mask);
1619 * Caller may return ENODATA to indicate that there
1620 * was no data to be dumped. This is not an error, it
1621 * means we should trim the attribute header and
1622 * continue.
1624 if (err == -ENODATA)
1625 nla_nest_cancel(skb, af);
1626 else if (err < 0)
1627 return -EMSGSIZE;
1629 nla_nest_end(skb, af);
1632 nla_nest_end(skb, af_spec);
1633 return 0;
1636 static int rtnl_fill_alt_ifnames(struct sk_buff *skb,
1637 const struct net_device *dev)
1639 struct netdev_name_node *name_node;
1640 int count = 0;
1642 list_for_each_entry(name_node, &dev->name_node->list, list) {
1643 if (nla_put_string(skb, IFLA_ALT_IFNAME, name_node->name))
1644 return -EMSGSIZE;
1645 count++;
1647 return count;
1650 static int rtnl_fill_prop_list(struct sk_buff *skb,
1651 const struct net_device *dev)
1653 struct nlattr *prop_list;
1654 int ret;
1656 prop_list = nla_nest_start(skb, IFLA_PROP_LIST);
1657 if (!prop_list)
1658 return -EMSGSIZE;
1660 ret = rtnl_fill_alt_ifnames(skb, dev);
1661 if (ret <= 0)
1662 goto nest_cancel;
1664 nla_nest_end(skb, prop_list);
1665 return 0;
1667 nest_cancel:
1668 nla_nest_cancel(skb, prop_list);
1669 return ret;
1672 static int rtnl_fill_proto_down(struct sk_buff *skb,
1673 const struct net_device *dev)
1675 struct nlattr *pr;
1676 u32 preason;
1678 if (nla_put_u8(skb, IFLA_PROTO_DOWN, dev->proto_down))
1679 goto nla_put_failure;
1681 preason = dev->proto_down_reason;
1682 if (!preason)
1683 return 0;
1685 pr = nla_nest_start(skb, IFLA_PROTO_DOWN_REASON);
1686 if (!pr)
1687 return -EMSGSIZE;
1689 if (nla_put_u32(skb, IFLA_PROTO_DOWN_REASON_VALUE, preason)) {
1690 nla_nest_cancel(skb, pr);
1691 goto nla_put_failure;
1694 nla_nest_end(skb, pr);
1695 return 0;
1697 nla_put_failure:
1698 return -EMSGSIZE;
1701 static int rtnl_fill_ifinfo(struct sk_buff *skb,
1702 struct net_device *dev, struct net *src_net,
1703 int type, u32 pid, u32 seq, u32 change,
1704 unsigned int flags, u32 ext_filter_mask,
1705 u32 event, int *new_nsid, int new_ifindex,
1706 int tgt_netnsid, gfp_t gfp)
1708 struct ifinfomsg *ifm;
1709 struct nlmsghdr *nlh;
1711 ASSERT_RTNL();
1712 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
1713 if (nlh == NULL)
1714 return -EMSGSIZE;
1716 ifm = nlmsg_data(nlh);
1717 ifm->ifi_family = AF_UNSPEC;
1718 ifm->__ifi_pad = 0;
1719 ifm->ifi_type = dev->type;
1720 ifm->ifi_index = dev->ifindex;
1721 ifm->ifi_flags = dev_get_flags(dev);
1722 ifm->ifi_change = change;
1724 if (tgt_netnsid >= 0 && nla_put_s32(skb, IFLA_TARGET_NETNSID, tgt_netnsid))
1725 goto nla_put_failure;
1727 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
1728 nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
1729 nla_put_u8(skb, IFLA_OPERSTATE,
1730 netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
1731 nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
1732 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
1733 nla_put_u32(skb, IFLA_MIN_MTU, dev->min_mtu) ||
1734 nla_put_u32(skb, IFLA_MAX_MTU, dev->max_mtu) ||
1735 nla_put_u32(skb, IFLA_GROUP, dev->group) ||
1736 nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
1737 nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
1738 nla_put_u32(skb, IFLA_GSO_MAX_SEGS, dev->gso_max_segs) ||
1739 nla_put_u32(skb, IFLA_GSO_MAX_SIZE, dev->gso_max_size) ||
1740 #ifdef CONFIG_RPS
1741 nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1742 #endif
1743 put_master_ifindex(skb, dev) ||
1744 nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1745 (dev->qdisc &&
1746 nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
1747 nla_put_ifalias(skb, dev) ||
1748 nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1749 atomic_read(&dev->carrier_up_count) +
1750 atomic_read(&dev->carrier_down_count)) ||
1751 nla_put_u32(skb, IFLA_CARRIER_UP_COUNT,
1752 atomic_read(&dev->carrier_up_count)) ||
1753 nla_put_u32(skb, IFLA_CARRIER_DOWN_COUNT,
1754 atomic_read(&dev->carrier_down_count)))
1755 goto nla_put_failure;
1757 if (rtnl_fill_proto_down(skb, dev))
1758 goto nla_put_failure;
1760 if (event != IFLA_EVENT_NONE) {
1761 if (nla_put_u32(skb, IFLA_EVENT, event))
1762 goto nla_put_failure;
1765 if (rtnl_fill_link_ifmap(skb, dev))
1766 goto nla_put_failure;
1768 if (dev->addr_len) {
1769 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1770 nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1771 goto nla_put_failure;
1774 if (rtnl_phys_port_id_fill(skb, dev))
1775 goto nla_put_failure;
1777 if (rtnl_phys_port_name_fill(skb, dev))
1778 goto nla_put_failure;
1780 if (rtnl_phys_switch_id_fill(skb, dev))
1781 goto nla_put_failure;
1783 if (rtnl_fill_stats(skb, dev))
1784 goto nla_put_failure;
1786 if (rtnl_fill_vf(skb, dev, ext_filter_mask))
1787 goto nla_put_failure;
1789 if (rtnl_port_fill(skb, dev, ext_filter_mask))
1790 goto nla_put_failure;
1792 if (rtnl_xdp_fill(skb, dev))
1793 goto nla_put_failure;
1795 if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1796 if (rtnl_link_fill(skb, dev) < 0)
1797 goto nla_put_failure;
1800 if (rtnl_fill_link_netnsid(skb, dev, src_net, gfp))
1801 goto nla_put_failure;
1803 if (new_nsid &&
1804 nla_put_s32(skb, IFLA_NEW_NETNSID, *new_nsid) < 0)
1805 goto nla_put_failure;
1806 if (new_ifindex &&
1807 nla_put_s32(skb, IFLA_NEW_IFINDEX, new_ifindex) < 0)
1808 goto nla_put_failure;
1810 if (memchr_inv(dev->perm_addr, '\0', dev->addr_len) &&
1811 nla_put(skb, IFLA_PERM_ADDRESS, dev->addr_len, dev->perm_addr))
1812 goto nla_put_failure;
1814 rcu_read_lock();
1815 if (rtnl_fill_link_af(skb, dev, ext_filter_mask))
1816 goto nla_put_failure_rcu;
1817 rcu_read_unlock();
1819 if (rtnl_fill_prop_list(skb, dev))
1820 goto nla_put_failure;
1822 nlmsg_end(skb, nlh);
1823 return 0;
1825 nla_put_failure_rcu:
1826 rcu_read_unlock();
1827 nla_put_failure:
1828 nlmsg_cancel(skb, nlh);
1829 return -EMSGSIZE;
1832 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1833 [IFLA_IFNAME] = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1834 [IFLA_ADDRESS] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1835 [IFLA_BROADCAST] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1836 [IFLA_MAP] = { .len = sizeof(struct rtnl_link_ifmap) },
1837 [IFLA_MTU] = { .type = NLA_U32 },
1838 [IFLA_LINK] = { .type = NLA_U32 },
1839 [IFLA_MASTER] = { .type = NLA_U32 },
1840 [IFLA_CARRIER] = { .type = NLA_U8 },
1841 [IFLA_TXQLEN] = { .type = NLA_U32 },
1842 [IFLA_WEIGHT] = { .type = NLA_U32 },
1843 [IFLA_OPERSTATE] = { .type = NLA_U8 },
1844 [IFLA_LINKMODE] = { .type = NLA_U8 },
1845 [IFLA_LINKINFO] = { .type = NLA_NESTED },
1846 [IFLA_NET_NS_PID] = { .type = NLA_U32 },
1847 [IFLA_NET_NS_FD] = { .type = NLA_U32 },
1848 /* IFLA_IFALIAS is a string, but policy is set to NLA_BINARY to
1849 * allow 0-length string (needed to remove an alias).
1851 [IFLA_IFALIAS] = { .type = NLA_BINARY, .len = IFALIASZ - 1 },
1852 [IFLA_VFINFO_LIST] = {. type = NLA_NESTED },
1853 [IFLA_VF_PORTS] = { .type = NLA_NESTED },
1854 [IFLA_PORT_SELF] = { .type = NLA_NESTED },
1855 [IFLA_AF_SPEC] = { .type = NLA_NESTED },
1856 [IFLA_EXT_MASK] = { .type = NLA_U32 },
1857 [IFLA_PROMISCUITY] = { .type = NLA_U32 },
1858 [IFLA_NUM_TX_QUEUES] = { .type = NLA_U32 },
1859 [IFLA_NUM_RX_QUEUES] = { .type = NLA_U32 },
1860 [IFLA_GSO_MAX_SEGS] = { .type = NLA_U32 },
1861 [IFLA_GSO_MAX_SIZE] = { .type = NLA_U32 },
1862 [IFLA_PHYS_PORT_ID] = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1863 [IFLA_CARRIER_CHANGES] = { .type = NLA_U32 }, /* ignored */
1864 [IFLA_PHYS_SWITCH_ID] = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1865 [IFLA_LINK_NETNSID] = { .type = NLA_S32 },
1866 [IFLA_PROTO_DOWN] = { .type = NLA_U8 },
1867 [IFLA_XDP] = { .type = NLA_NESTED },
1868 [IFLA_EVENT] = { .type = NLA_U32 },
1869 [IFLA_GROUP] = { .type = NLA_U32 },
1870 [IFLA_TARGET_NETNSID] = { .type = NLA_S32 },
1871 [IFLA_CARRIER_UP_COUNT] = { .type = NLA_U32 },
1872 [IFLA_CARRIER_DOWN_COUNT] = { .type = NLA_U32 },
1873 [IFLA_MIN_MTU] = { .type = NLA_U32 },
1874 [IFLA_MAX_MTU] = { .type = NLA_U32 },
1875 [IFLA_PROP_LIST] = { .type = NLA_NESTED },
1876 [IFLA_ALT_IFNAME] = { .type = NLA_STRING,
1877 .len = ALTIFNAMSIZ - 1 },
1878 [IFLA_PERM_ADDRESS] = { .type = NLA_REJECT },
1879 [IFLA_PROTO_DOWN_REASON] = { .type = NLA_NESTED },
1882 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1883 [IFLA_INFO_KIND] = { .type = NLA_STRING },
1884 [IFLA_INFO_DATA] = { .type = NLA_NESTED },
1885 [IFLA_INFO_SLAVE_KIND] = { .type = NLA_STRING },
1886 [IFLA_INFO_SLAVE_DATA] = { .type = NLA_NESTED },
1889 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1890 [IFLA_VF_MAC] = { .len = sizeof(struct ifla_vf_mac) },
1891 [IFLA_VF_BROADCAST] = { .type = NLA_REJECT },
1892 [IFLA_VF_VLAN] = { .len = sizeof(struct ifla_vf_vlan) },
1893 [IFLA_VF_VLAN_LIST] = { .type = NLA_NESTED },
1894 [IFLA_VF_TX_RATE] = { .len = sizeof(struct ifla_vf_tx_rate) },
1895 [IFLA_VF_SPOOFCHK] = { .len = sizeof(struct ifla_vf_spoofchk) },
1896 [IFLA_VF_RATE] = { .len = sizeof(struct ifla_vf_rate) },
1897 [IFLA_VF_LINK_STATE] = { .len = sizeof(struct ifla_vf_link_state) },
1898 [IFLA_VF_RSS_QUERY_EN] = { .len = sizeof(struct ifla_vf_rss_query_en) },
1899 [IFLA_VF_STATS] = { .type = NLA_NESTED },
1900 [IFLA_VF_TRUST] = { .len = sizeof(struct ifla_vf_trust) },
1901 [IFLA_VF_IB_NODE_GUID] = { .len = sizeof(struct ifla_vf_guid) },
1902 [IFLA_VF_IB_PORT_GUID] = { .len = sizeof(struct ifla_vf_guid) },
1905 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1906 [IFLA_PORT_VF] = { .type = NLA_U32 },
1907 [IFLA_PORT_PROFILE] = { .type = NLA_STRING,
1908 .len = PORT_PROFILE_MAX },
1909 [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1910 .len = PORT_UUID_MAX },
1911 [IFLA_PORT_HOST_UUID] = { .type = NLA_STRING,
1912 .len = PORT_UUID_MAX },
1913 [IFLA_PORT_REQUEST] = { .type = NLA_U8, },
1914 [IFLA_PORT_RESPONSE] = { .type = NLA_U16, },
1916 /* Unused, but we need to keep it here since user space could
1917 * fill it. It's also broken with regard to NLA_BINARY use in
1918 * combination with structs.
1920 [IFLA_PORT_VSI_TYPE] = { .type = NLA_BINARY,
1921 .len = sizeof(struct ifla_port_vsi) },
1924 static const struct nla_policy ifla_xdp_policy[IFLA_XDP_MAX + 1] = {
1925 [IFLA_XDP_UNSPEC] = { .strict_start_type = IFLA_XDP_EXPECTED_FD },
1926 [IFLA_XDP_FD] = { .type = NLA_S32 },
1927 [IFLA_XDP_EXPECTED_FD] = { .type = NLA_S32 },
1928 [IFLA_XDP_ATTACHED] = { .type = NLA_U8 },
1929 [IFLA_XDP_FLAGS] = { .type = NLA_U32 },
1930 [IFLA_XDP_PROG_ID] = { .type = NLA_U32 },
1933 static const struct rtnl_link_ops *linkinfo_to_kind_ops(const struct nlattr *nla)
1935 const struct rtnl_link_ops *ops = NULL;
1936 struct nlattr *linfo[IFLA_INFO_MAX + 1];
1938 if (nla_parse_nested_deprecated(linfo, IFLA_INFO_MAX, nla, ifla_info_policy, NULL) < 0)
1939 return NULL;
1941 if (linfo[IFLA_INFO_KIND]) {
1942 char kind[MODULE_NAME_LEN];
1944 nla_strscpy(kind, linfo[IFLA_INFO_KIND], sizeof(kind));
1945 ops = rtnl_link_ops_get(kind);
1948 return ops;
1951 static bool link_master_filtered(struct net_device *dev, int master_idx)
1953 struct net_device *master;
1955 if (!master_idx)
1956 return false;
1958 master = netdev_master_upper_dev_get(dev);
1959 if (!master || master->ifindex != master_idx)
1960 return true;
1962 return false;
1965 static bool link_kind_filtered(const struct net_device *dev,
1966 const struct rtnl_link_ops *kind_ops)
1968 if (kind_ops && dev->rtnl_link_ops != kind_ops)
1969 return true;
1971 return false;
1974 static bool link_dump_filtered(struct net_device *dev,
1975 int master_idx,
1976 const struct rtnl_link_ops *kind_ops)
1978 if (link_master_filtered(dev, master_idx) ||
1979 link_kind_filtered(dev, kind_ops))
1980 return true;
1982 return false;
1986 * rtnl_get_net_ns_capable - Get netns if sufficiently privileged.
1987 * @sk: netlink socket
1988 * @netnsid: network namespace identifier
1990 * Returns the network namespace identified by netnsid on success or an error
1991 * pointer on failure.
1993 struct net *rtnl_get_net_ns_capable(struct sock *sk, int netnsid)
1995 struct net *net;
1997 net = get_net_ns_by_id(sock_net(sk), netnsid);
1998 if (!net)
1999 return ERR_PTR(-EINVAL);
2001 /* For now, the caller is required to have CAP_NET_ADMIN in
2002 * the user namespace owning the target net ns.
2004 if (!sk_ns_capable(sk, net->user_ns, CAP_NET_ADMIN)) {
2005 put_net(net);
2006 return ERR_PTR(-EACCES);
2008 return net;
2010 EXPORT_SYMBOL_GPL(rtnl_get_net_ns_capable);
2012 static int rtnl_valid_dump_ifinfo_req(const struct nlmsghdr *nlh,
2013 bool strict_check, struct nlattr **tb,
2014 struct netlink_ext_ack *extack)
2016 int hdrlen;
2018 if (strict_check) {
2019 struct ifinfomsg *ifm;
2021 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
2022 NL_SET_ERR_MSG(extack, "Invalid header for link dump");
2023 return -EINVAL;
2026 ifm = nlmsg_data(nlh);
2027 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
2028 ifm->ifi_change) {
2029 NL_SET_ERR_MSG(extack, "Invalid values in header for link dump request");
2030 return -EINVAL;
2032 if (ifm->ifi_index) {
2033 NL_SET_ERR_MSG(extack, "Filter by device index not supported for link dumps");
2034 return -EINVAL;
2037 return nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb,
2038 IFLA_MAX, ifla_policy,
2039 extack);
2042 /* A hack to preserve kernel<->userspace interface.
2043 * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
2044 * However, before Linux v3.9 the code here assumed rtgenmsg and that's
2045 * what iproute2 < v3.9.0 used.
2046 * We can detect the old iproute2. Even including the IFLA_EXT_MASK
2047 * attribute, its netlink message is shorter than struct ifinfomsg.
2049 hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
2050 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
2052 return nlmsg_parse_deprecated(nlh, hdrlen, tb, IFLA_MAX, ifla_policy,
2053 extack);
2056 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
2058 struct netlink_ext_ack *extack = cb->extack;
2059 const struct nlmsghdr *nlh = cb->nlh;
2060 struct net *net = sock_net(skb->sk);
2061 struct net *tgt_net = net;
2062 int h, s_h;
2063 int idx = 0, s_idx;
2064 struct net_device *dev;
2065 struct hlist_head *head;
2066 struct nlattr *tb[IFLA_MAX+1];
2067 u32 ext_filter_mask = 0;
2068 const struct rtnl_link_ops *kind_ops = NULL;
2069 unsigned int flags = NLM_F_MULTI;
2070 int master_idx = 0;
2071 int netnsid = -1;
2072 int err, i;
2074 s_h = cb->args[0];
2075 s_idx = cb->args[1];
2077 err = rtnl_valid_dump_ifinfo_req(nlh, cb->strict_check, tb, extack);
2078 if (err < 0) {
2079 if (cb->strict_check)
2080 return err;
2082 goto walk_entries;
2085 for (i = 0; i <= IFLA_MAX; ++i) {
2086 if (!tb[i])
2087 continue;
2089 /* new attributes should only be added with strict checking */
2090 switch (i) {
2091 case IFLA_TARGET_NETNSID:
2092 netnsid = nla_get_s32(tb[i]);
2093 tgt_net = rtnl_get_net_ns_capable(skb->sk, netnsid);
2094 if (IS_ERR(tgt_net)) {
2095 NL_SET_ERR_MSG(extack, "Invalid target network namespace id");
2096 return PTR_ERR(tgt_net);
2098 break;
2099 case IFLA_EXT_MASK:
2100 ext_filter_mask = nla_get_u32(tb[i]);
2101 break;
2102 case IFLA_MASTER:
2103 master_idx = nla_get_u32(tb[i]);
2104 break;
2105 case IFLA_LINKINFO:
2106 kind_ops = linkinfo_to_kind_ops(tb[i]);
2107 break;
2108 default:
2109 if (cb->strict_check) {
2110 NL_SET_ERR_MSG(extack, "Unsupported attribute in link dump request");
2111 return -EINVAL;
2116 if (master_idx || kind_ops)
2117 flags |= NLM_F_DUMP_FILTERED;
2119 walk_entries:
2120 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
2121 idx = 0;
2122 head = &tgt_net->dev_index_head[h];
2123 hlist_for_each_entry(dev, head, index_hlist) {
2124 if (link_dump_filtered(dev, master_idx, kind_ops))
2125 goto cont;
2126 if (idx < s_idx)
2127 goto cont;
2128 err = rtnl_fill_ifinfo(skb, dev, net,
2129 RTM_NEWLINK,
2130 NETLINK_CB(cb->skb).portid,
2131 nlh->nlmsg_seq, 0, flags,
2132 ext_filter_mask, 0, NULL, 0,
2133 netnsid, GFP_KERNEL);
2135 if (err < 0) {
2136 if (likely(skb->len))
2137 goto out;
2139 goto out_err;
2141 cont:
2142 idx++;
2145 out:
2146 err = skb->len;
2147 out_err:
2148 cb->args[1] = idx;
2149 cb->args[0] = h;
2150 cb->seq = net->dev_base_seq;
2151 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
2152 if (netnsid >= 0)
2153 put_net(tgt_net);
2155 return err;
2158 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len,
2159 struct netlink_ext_ack *exterr)
2161 return nla_parse_deprecated(tb, IFLA_MAX, head, len, ifla_policy,
2162 exterr);
2164 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
2166 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
2168 struct net *net;
2169 /* Examine the link attributes and figure out which
2170 * network namespace we are talking about.
2172 if (tb[IFLA_NET_NS_PID])
2173 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
2174 else if (tb[IFLA_NET_NS_FD])
2175 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
2176 else
2177 net = get_net(src_net);
2178 return net;
2180 EXPORT_SYMBOL(rtnl_link_get_net);
2182 /* Figure out which network namespace we are talking about by
2183 * examining the link attributes in the following order:
2185 * 1. IFLA_NET_NS_PID
2186 * 2. IFLA_NET_NS_FD
2187 * 3. IFLA_TARGET_NETNSID
2189 static struct net *rtnl_link_get_net_by_nlattr(struct net *src_net,
2190 struct nlattr *tb[])
2192 struct net *net;
2194 if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD])
2195 return rtnl_link_get_net(src_net, tb);
2197 if (!tb[IFLA_TARGET_NETNSID])
2198 return get_net(src_net);
2200 net = get_net_ns_by_id(src_net, nla_get_u32(tb[IFLA_TARGET_NETNSID]));
2201 if (!net)
2202 return ERR_PTR(-EINVAL);
2204 return net;
2207 static struct net *rtnl_link_get_net_capable(const struct sk_buff *skb,
2208 struct net *src_net,
2209 struct nlattr *tb[], int cap)
2211 struct net *net;
2213 net = rtnl_link_get_net_by_nlattr(src_net, tb);
2214 if (IS_ERR(net))
2215 return net;
2217 if (!netlink_ns_capable(skb, net->user_ns, cap)) {
2218 put_net(net);
2219 return ERR_PTR(-EPERM);
2222 return net;
2225 /* Verify that rtnetlink requests do not pass additional properties
2226 * potentially referring to different network namespaces.
2228 static int rtnl_ensure_unique_netns(struct nlattr *tb[],
2229 struct netlink_ext_ack *extack,
2230 bool netns_id_only)
2233 if (netns_id_only) {
2234 if (!tb[IFLA_NET_NS_PID] && !tb[IFLA_NET_NS_FD])
2235 return 0;
2237 NL_SET_ERR_MSG(extack, "specified netns attribute not supported");
2238 return -EOPNOTSUPP;
2241 if (tb[IFLA_TARGET_NETNSID] && (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]))
2242 goto invalid_attr;
2244 if (tb[IFLA_NET_NS_PID] && (tb[IFLA_TARGET_NETNSID] || tb[IFLA_NET_NS_FD]))
2245 goto invalid_attr;
2247 if (tb[IFLA_NET_NS_FD] && (tb[IFLA_TARGET_NETNSID] || tb[IFLA_NET_NS_PID]))
2248 goto invalid_attr;
2250 return 0;
2252 invalid_attr:
2253 NL_SET_ERR_MSG(extack, "multiple netns identifying attributes specified");
2254 return -EINVAL;
2257 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
2259 if (dev) {
2260 if (tb[IFLA_ADDRESS] &&
2261 nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
2262 return -EINVAL;
2264 if (tb[IFLA_BROADCAST] &&
2265 nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
2266 return -EINVAL;
2269 if (tb[IFLA_AF_SPEC]) {
2270 struct nlattr *af;
2271 int rem, err;
2273 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2274 const struct rtnl_af_ops *af_ops;
2276 rcu_read_lock();
2277 af_ops = rtnl_af_lookup(nla_type(af));
2278 if (!af_ops) {
2279 rcu_read_unlock();
2280 return -EAFNOSUPPORT;
2283 if (!af_ops->set_link_af) {
2284 rcu_read_unlock();
2285 return -EOPNOTSUPP;
2288 if (af_ops->validate_link_af) {
2289 err = af_ops->validate_link_af(dev, af);
2290 if (err < 0) {
2291 rcu_read_unlock();
2292 return err;
2296 rcu_read_unlock();
2300 return 0;
2303 static int handle_infiniband_guid(struct net_device *dev, struct ifla_vf_guid *ivt,
2304 int guid_type)
2306 const struct net_device_ops *ops = dev->netdev_ops;
2308 return ops->ndo_set_vf_guid(dev, ivt->vf, ivt->guid, guid_type);
2311 static int handle_vf_guid(struct net_device *dev, struct ifla_vf_guid *ivt, int guid_type)
2313 if (dev->type != ARPHRD_INFINIBAND)
2314 return -EOPNOTSUPP;
2316 return handle_infiniband_guid(dev, ivt, guid_type);
2319 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb)
2321 const struct net_device_ops *ops = dev->netdev_ops;
2322 int err = -EINVAL;
2324 if (tb[IFLA_VF_MAC]) {
2325 struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]);
2327 if (ivm->vf >= INT_MAX)
2328 return -EINVAL;
2329 err = -EOPNOTSUPP;
2330 if (ops->ndo_set_vf_mac)
2331 err = ops->ndo_set_vf_mac(dev, ivm->vf,
2332 ivm->mac);
2333 if (err < 0)
2334 return err;
2337 if (tb[IFLA_VF_VLAN]) {
2338 struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]);
2340 if (ivv->vf >= INT_MAX)
2341 return -EINVAL;
2342 err = -EOPNOTSUPP;
2343 if (ops->ndo_set_vf_vlan)
2344 err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan,
2345 ivv->qos,
2346 htons(ETH_P_8021Q));
2347 if (err < 0)
2348 return err;
2351 if (tb[IFLA_VF_VLAN_LIST]) {
2352 struct ifla_vf_vlan_info *ivvl[MAX_VLAN_LIST_LEN];
2353 struct nlattr *attr;
2354 int rem, len = 0;
2356 err = -EOPNOTSUPP;
2357 if (!ops->ndo_set_vf_vlan)
2358 return err;
2360 nla_for_each_nested(attr, tb[IFLA_VF_VLAN_LIST], rem) {
2361 if (nla_type(attr) != IFLA_VF_VLAN_INFO ||
2362 nla_len(attr) < NLA_HDRLEN) {
2363 return -EINVAL;
2365 if (len >= MAX_VLAN_LIST_LEN)
2366 return -EOPNOTSUPP;
2367 ivvl[len] = nla_data(attr);
2369 len++;
2371 if (len == 0)
2372 return -EINVAL;
2374 if (ivvl[0]->vf >= INT_MAX)
2375 return -EINVAL;
2376 err = ops->ndo_set_vf_vlan(dev, ivvl[0]->vf, ivvl[0]->vlan,
2377 ivvl[0]->qos, ivvl[0]->vlan_proto);
2378 if (err < 0)
2379 return err;
2382 if (tb[IFLA_VF_TX_RATE]) {
2383 struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]);
2384 struct ifla_vf_info ivf;
2386 if (ivt->vf >= INT_MAX)
2387 return -EINVAL;
2388 err = -EOPNOTSUPP;
2389 if (ops->ndo_get_vf_config)
2390 err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf);
2391 if (err < 0)
2392 return err;
2394 err = -EOPNOTSUPP;
2395 if (ops->ndo_set_vf_rate)
2396 err = ops->ndo_set_vf_rate(dev, ivt->vf,
2397 ivf.min_tx_rate,
2398 ivt->rate);
2399 if (err < 0)
2400 return err;
2403 if (tb[IFLA_VF_RATE]) {
2404 struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]);
2406 if (ivt->vf >= INT_MAX)
2407 return -EINVAL;
2408 err = -EOPNOTSUPP;
2409 if (ops->ndo_set_vf_rate)
2410 err = ops->ndo_set_vf_rate(dev, ivt->vf,
2411 ivt->min_tx_rate,
2412 ivt->max_tx_rate);
2413 if (err < 0)
2414 return err;
2417 if (tb[IFLA_VF_SPOOFCHK]) {
2418 struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]);
2420 if (ivs->vf >= INT_MAX)
2421 return -EINVAL;
2422 err = -EOPNOTSUPP;
2423 if (ops->ndo_set_vf_spoofchk)
2424 err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
2425 ivs->setting);
2426 if (err < 0)
2427 return err;
2430 if (tb[IFLA_VF_LINK_STATE]) {
2431 struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]);
2433 if (ivl->vf >= INT_MAX)
2434 return -EINVAL;
2435 err = -EOPNOTSUPP;
2436 if (ops->ndo_set_vf_link_state)
2437 err = ops->ndo_set_vf_link_state(dev, ivl->vf,
2438 ivl->link_state);
2439 if (err < 0)
2440 return err;
2443 if (tb[IFLA_VF_RSS_QUERY_EN]) {
2444 struct ifla_vf_rss_query_en *ivrssq_en;
2446 err = -EOPNOTSUPP;
2447 ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]);
2448 if (ivrssq_en->vf >= INT_MAX)
2449 return -EINVAL;
2450 if (ops->ndo_set_vf_rss_query_en)
2451 err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf,
2452 ivrssq_en->setting);
2453 if (err < 0)
2454 return err;
2457 if (tb[IFLA_VF_TRUST]) {
2458 struct ifla_vf_trust *ivt = nla_data(tb[IFLA_VF_TRUST]);
2460 if (ivt->vf >= INT_MAX)
2461 return -EINVAL;
2462 err = -EOPNOTSUPP;
2463 if (ops->ndo_set_vf_trust)
2464 err = ops->ndo_set_vf_trust(dev, ivt->vf, ivt->setting);
2465 if (err < 0)
2466 return err;
2469 if (tb[IFLA_VF_IB_NODE_GUID]) {
2470 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_NODE_GUID]);
2472 if (ivt->vf >= INT_MAX)
2473 return -EINVAL;
2474 if (!ops->ndo_set_vf_guid)
2475 return -EOPNOTSUPP;
2476 return handle_vf_guid(dev, ivt, IFLA_VF_IB_NODE_GUID);
2479 if (tb[IFLA_VF_IB_PORT_GUID]) {
2480 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_PORT_GUID]);
2482 if (ivt->vf >= INT_MAX)
2483 return -EINVAL;
2484 if (!ops->ndo_set_vf_guid)
2485 return -EOPNOTSUPP;
2487 return handle_vf_guid(dev, ivt, IFLA_VF_IB_PORT_GUID);
2490 return err;
2493 static int do_set_master(struct net_device *dev, int ifindex,
2494 struct netlink_ext_ack *extack)
2496 struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
2497 const struct net_device_ops *ops;
2498 int err;
2500 if (upper_dev) {
2501 if (upper_dev->ifindex == ifindex)
2502 return 0;
2503 ops = upper_dev->netdev_ops;
2504 if (ops->ndo_del_slave) {
2505 err = ops->ndo_del_slave(upper_dev, dev);
2506 if (err)
2507 return err;
2508 } else {
2509 return -EOPNOTSUPP;
2513 if (ifindex) {
2514 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
2515 if (!upper_dev)
2516 return -EINVAL;
2517 ops = upper_dev->netdev_ops;
2518 if (ops->ndo_add_slave) {
2519 err = ops->ndo_add_slave(upper_dev, dev, extack);
2520 if (err)
2521 return err;
2522 } else {
2523 return -EOPNOTSUPP;
2526 return 0;
2529 static const struct nla_policy ifla_proto_down_reason_policy[IFLA_PROTO_DOWN_REASON_VALUE + 1] = {
2530 [IFLA_PROTO_DOWN_REASON_MASK] = { .type = NLA_U32 },
2531 [IFLA_PROTO_DOWN_REASON_VALUE] = { .type = NLA_U32 },
2534 static int do_set_proto_down(struct net_device *dev,
2535 struct nlattr *nl_proto_down,
2536 struct nlattr *nl_proto_down_reason,
2537 struct netlink_ext_ack *extack)
2539 struct nlattr *pdreason[IFLA_PROTO_DOWN_REASON_MAX + 1];
2540 const struct net_device_ops *ops = dev->netdev_ops;
2541 unsigned long mask = 0;
2542 u32 value;
2543 bool proto_down;
2544 int err;
2546 if (!ops->ndo_change_proto_down) {
2547 NL_SET_ERR_MSG(extack, "Protodown not supported by device");
2548 return -EOPNOTSUPP;
2551 if (nl_proto_down_reason) {
2552 err = nla_parse_nested_deprecated(pdreason,
2553 IFLA_PROTO_DOWN_REASON_MAX,
2554 nl_proto_down_reason,
2555 ifla_proto_down_reason_policy,
2556 NULL);
2557 if (err < 0)
2558 return err;
2560 if (!pdreason[IFLA_PROTO_DOWN_REASON_VALUE]) {
2561 NL_SET_ERR_MSG(extack, "Invalid protodown reason value");
2562 return -EINVAL;
2565 value = nla_get_u32(pdreason[IFLA_PROTO_DOWN_REASON_VALUE]);
2567 if (pdreason[IFLA_PROTO_DOWN_REASON_MASK])
2568 mask = nla_get_u32(pdreason[IFLA_PROTO_DOWN_REASON_MASK]);
2570 dev_change_proto_down_reason(dev, mask, value);
2573 if (nl_proto_down) {
2574 proto_down = nla_get_u8(nl_proto_down);
2576 /* Dont turn off protodown if there are active reasons */
2577 if (!proto_down && dev->proto_down_reason) {
2578 NL_SET_ERR_MSG(extack, "Cannot clear protodown, active reasons");
2579 return -EBUSY;
2581 err = dev_change_proto_down(dev,
2582 proto_down);
2583 if (err)
2584 return err;
2587 return 0;
2590 #define DO_SETLINK_MODIFIED 0x01
2591 /* notify flag means notify + modified. */
2592 #define DO_SETLINK_NOTIFY 0x03
2593 static int do_setlink(const struct sk_buff *skb,
2594 struct net_device *dev, struct ifinfomsg *ifm,
2595 struct netlink_ext_ack *extack,
2596 struct nlattr **tb, char *ifname, int status)
2598 const struct net_device_ops *ops = dev->netdev_ops;
2599 int err;
2601 err = validate_linkmsg(dev, tb);
2602 if (err < 0)
2603 return err;
2605 if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD] || tb[IFLA_TARGET_NETNSID]) {
2606 struct net *net = rtnl_link_get_net_capable(skb, dev_net(dev),
2607 tb, CAP_NET_ADMIN);
2608 if (IS_ERR(net)) {
2609 err = PTR_ERR(net);
2610 goto errout;
2613 err = dev_change_net_namespace(dev, net, ifname);
2614 put_net(net);
2615 if (err)
2616 goto errout;
2617 status |= DO_SETLINK_MODIFIED;
2620 if (tb[IFLA_MAP]) {
2621 struct rtnl_link_ifmap *u_map;
2622 struct ifmap k_map;
2624 if (!ops->ndo_set_config) {
2625 err = -EOPNOTSUPP;
2626 goto errout;
2629 if (!netif_device_present(dev)) {
2630 err = -ENODEV;
2631 goto errout;
2634 u_map = nla_data(tb[IFLA_MAP]);
2635 k_map.mem_start = (unsigned long) u_map->mem_start;
2636 k_map.mem_end = (unsigned long) u_map->mem_end;
2637 k_map.base_addr = (unsigned short) u_map->base_addr;
2638 k_map.irq = (unsigned char) u_map->irq;
2639 k_map.dma = (unsigned char) u_map->dma;
2640 k_map.port = (unsigned char) u_map->port;
2642 err = ops->ndo_set_config(dev, &k_map);
2643 if (err < 0)
2644 goto errout;
2646 status |= DO_SETLINK_NOTIFY;
2649 if (tb[IFLA_ADDRESS]) {
2650 struct sockaddr *sa;
2651 int len;
2653 len = sizeof(sa_family_t) + max_t(size_t, dev->addr_len,
2654 sizeof(*sa));
2655 sa = kmalloc(len, GFP_KERNEL);
2656 if (!sa) {
2657 err = -ENOMEM;
2658 goto errout;
2660 sa->sa_family = dev->type;
2661 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
2662 dev->addr_len);
2663 err = dev_set_mac_address(dev, sa, extack);
2664 kfree(sa);
2665 if (err)
2666 goto errout;
2667 status |= DO_SETLINK_MODIFIED;
2670 if (tb[IFLA_MTU]) {
2671 err = dev_set_mtu_ext(dev, nla_get_u32(tb[IFLA_MTU]), extack);
2672 if (err < 0)
2673 goto errout;
2674 status |= DO_SETLINK_MODIFIED;
2677 if (tb[IFLA_GROUP]) {
2678 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2679 status |= DO_SETLINK_NOTIFY;
2683 * Interface selected by interface index but interface
2684 * name provided implies that a name change has been
2685 * requested.
2687 if (ifm->ifi_index > 0 && ifname[0]) {
2688 err = dev_change_name(dev, ifname);
2689 if (err < 0)
2690 goto errout;
2691 status |= DO_SETLINK_MODIFIED;
2694 if (tb[IFLA_IFALIAS]) {
2695 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
2696 nla_len(tb[IFLA_IFALIAS]));
2697 if (err < 0)
2698 goto errout;
2699 status |= DO_SETLINK_NOTIFY;
2702 if (tb[IFLA_BROADCAST]) {
2703 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
2704 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
2707 if (ifm->ifi_flags || ifm->ifi_change) {
2708 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm),
2709 extack);
2710 if (err < 0)
2711 goto errout;
2714 if (tb[IFLA_MASTER]) {
2715 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]), extack);
2716 if (err)
2717 goto errout;
2718 status |= DO_SETLINK_MODIFIED;
2721 if (tb[IFLA_CARRIER]) {
2722 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
2723 if (err)
2724 goto errout;
2725 status |= DO_SETLINK_MODIFIED;
2728 if (tb[IFLA_TXQLEN]) {
2729 unsigned int value = nla_get_u32(tb[IFLA_TXQLEN]);
2731 err = dev_change_tx_queue_len(dev, value);
2732 if (err)
2733 goto errout;
2734 status |= DO_SETLINK_MODIFIED;
2737 if (tb[IFLA_GSO_MAX_SIZE]) {
2738 u32 max_size = nla_get_u32(tb[IFLA_GSO_MAX_SIZE]);
2740 if (max_size > GSO_MAX_SIZE) {
2741 err = -EINVAL;
2742 goto errout;
2745 if (dev->gso_max_size ^ max_size) {
2746 netif_set_gso_max_size(dev, max_size);
2747 status |= DO_SETLINK_MODIFIED;
2751 if (tb[IFLA_GSO_MAX_SEGS]) {
2752 u32 max_segs = nla_get_u32(tb[IFLA_GSO_MAX_SEGS]);
2754 if (max_segs > GSO_MAX_SEGS) {
2755 err = -EINVAL;
2756 goto errout;
2759 if (dev->gso_max_segs ^ max_segs) {
2760 dev->gso_max_segs = max_segs;
2761 status |= DO_SETLINK_MODIFIED;
2765 if (tb[IFLA_OPERSTATE])
2766 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2768 if (tb[IFLA_LINKMODE]) {
2769 unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
2771 write_lock_bh(&dev_base_lock);
2772 if (dev->link_mode ^ value)
2773 status |= DO_SETLINK_NOTIFY;
2774 dev->link_mode = value;
2775 write_unlock_bh(&dev_base_lock);
2778 if (tb[IFLA_VFINFO_LIST]) {
2779 struct nlattr *vfinfo[IFLA_VF_MAX + 1];
2780 struct nlattr *attr;
2781 int rem;
2783 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
2784 if (nla_type(attr) != IFLA_VF_INFO ||
2785 nla_len(attr) < NLA_HDRLEN) {
2786 err = -EINVAL;
2787 goto errout;
2789 err = nla_parse_nested_deprecated(vfinfo, IFLA_VF_MAX,
2790 attr,
2791 ifla_vf_policy,
2792 NULL);
2793 if (err < 0)
2794 goto errout;
2795 err = do_setvfinfo(dev, vfinfo);
2796 if (err < 0)
2797 goto errout;
2798 status |= DO_SETLINK_NOTIFY;
2801 err = 0;
2803 if (tb[IFLA_VF_PORTS]) {
2804 struct nlattr *port[IFLA_PORT_MAX+1];
2805 struct nlattr *attr;
2806 int vf;
2807 int rem;
2809 err = -EOPNOTSUPP;
2810 if (!ops->ndo_set_vf_port)
2811 goto errout;
2813 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
2814 if (nla_type(attr) != IFLA_VF_PORT ||
2815 nla_len(attr) < NLA_HDRLEN) {
2816 err = -EINVAL;
2817 goto errout;
2819 err = nla_parse_nested_deprecated(port, IFLA_PORT_MAX,
2820 attr,
2821 ifla_port_policy,
2822 NULL);
2823 if (err < 0)
2824 goto errout;
2825 if (!port[IFLA_PORT_VF]) {
2826 err = -EOPNOTSUPP;
2827 goto errout;
2829 vf = nla_get_u32(port[IFLA_PORT_VF]);
2830 err = ops->ndo_set_vf_port(dev, vf, port);
2831 if (err < 0)
2832 goto errout;
2833 status |= DO_SETLINK_NOTIFY;
2836 err = 0;
2838 if (tb[IFLA_PORT_SELF]) {
2839 struct nlattr *port[IFLA_PORT_MAX+1];
2841 err = nla_parse_nested_deprecated(port, IFLA_PORT_MAX,
2842 tb[IFLA_PORT_SELF],
2843 ifla_port_policy, NULL);
2844 if (err < 0)
2845 goto errout;
2847 err = -EOPNOTSUPP;
2848 if (ops->ndo_set_vf_port)
2849 err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
2850 if (err < 0)
2851 goto errout;
2852 status |= DO_SETLINK_NOTIFY;
2855 if (tb[IFLA_AF_SPEC]) {
2856 struct nlattr *af;
2857 int rem;
2859 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2860 const struct rtnl_af_ops *af_ops;
2862 rcu_read_lock();
2864 BUG_ON(!(af_ops = rtnl_af_lookup(nla_type(af))));
2866 err = af_ops->set_link_af(dev, af);
2867 if (err < 0) {
2868 rcu_read_unlock();
2869 goto errout;
2872 rcu_read_unlock();
2873 status |= DO_SETLINK_NOTIFY;
2876 err = 0;
2878 if (tb[IFLA_PROTO_DOWN] || tb[IFLA_PROTO_DOWN_REASON]) {
2879 err = do_set_proto_down(dev, tb[IFLA_PROTO_DOWN],
2880 tb[IFLA_PROTO_DOWN_REASON], extack);
2881 if (err)
2882 goto errout;
2883 status |= DO_SETLINK_NOTIFY;
2886 if (tb[IFLA_XDP]) {
2887 struct nlattr *xdp[IFLA_XDP_MAX + 1];
2888 u32 xdp_flags = 0;
2890 err = nla_parse_nested_deprecated(xdp, IFLA_XDP_MAX,
2891 tb[IFLA_XDP],
2892 ifla_xdp_policy, NULL);
2893 if (err < 0)
2894 goto errout;
2896 if (xdp[IFLA_XDP_ATTACHED] || xdp[IFLA_XDP_PROG_ID]) {
2897 err = -EINVAL;
2898 goto errout;
2901 if (xdp[IFLA_XDP_FLAGS]) {
2902 xdp_flags = nla_get_u32(xdp[IFLA_XDP_FLAGS]);
2903 if (xdp_flags & ~XDP_FLAGS_MASK) {
2904 err = -EINVAL;
2905 goto errout;
2907 if (hweight32(xdp_flags & XDP_FLAGS_MODES) > 1) {
2908 err = -EINVAL;
2909 goto errout;
2913 if (xdp[IFLA_XDP_FD]) {
2914 int expected_fd = -1;
2916 if (xdp_flags & XDP_FLAGS_REPLACE) {
2917 if (!xdp[IFLA_XDP_EXPECTED_FD]) {
2918 err = -EINVAL;
2919 goto errout;
2921 expected_fd =
2922 nla_get_s32(xdp[IFLA_XDP_EXPECTED_FD]);
2925 err = dev_change_xdp_fd(dev, extack,
2926 nla_get_s32(xdp[IFLA_XDP_FD]),
2927 expected_fd,
2928 xdp_flags);
2929 if (err)
2930 goto errout;
2931 status |= DO_SETLINK_NOTIFY;
2935 errout:
2936 if (status & DO_SETLINK_MODIFIED) {
2937 if ((status & DO_SETLINK_NOTIFY) == DO_SETLINK_NOTIFY)
2938 netdev_state_change(dev);
2940 if (err < 0)
2941 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",
2942 dev->name);
2945 return err;
2948 static struct net_device *rtnl_dev_get(struct net *net,
2949 struct nlattr *ifname_attr,
2950 struct nlattr *altifname_attr,
2951 char *ifname)
2953 char buffer[ALTIFNAMSIZ];
2955 if (!ifname) {
2956 ifname = buffer;
2957 if (ifname_attr)
2958 nla_strscpy(ifname, ifname_attr, IFNAMSIZ);
2959 else if (altifname_attr)
2960 nla_strscpy(ifname, altifname_attr, ALTIFNAMSIZ);
2961 else
2962 return NULL;
2965 return __dev_get_by_name(net, ifname);
2968 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
2969 struct netlink_ext_ack *extack)
2971 struct net *net = sock_net(skb->sk);
2972 struct ifinfomsg *ifm;
2973 struct net_device *dev;
2974 int err;
2975 struct nlattr *tb[IFLA_MAX+1];
2976 char ifname[IFNAMSIZ];
2978 err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
2979 ifla_policy, extack);
2980 if (err < 0)
2981 goto errout;
2983 err = rtnl_ensure_unique_netns(tb, extack, false);
2984 if (err < 0)
2985 goto errout;
2987 if (tb[IFLA_IFNAME])
2988 nla_strscpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2989 else
2990 ifname[0] = '\0';
2992 err = -EINVAL;
2993 ifm = nlmsg_data(nlh);
2994 if (ifm->ifi_index > 0)
2995 dev = __dev_get_by_index(net, ifm->ifi_index);
2996 else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
2997 dev = rtnl_dev_get(net, NULL, tb[IFLA_ALT_IFNAME], ifname);
2998 else
2999 goto errout;
3001 if (dev == NULL) {
3002 err = -ENODEV;
3003 goto errout;
3006 err = do_setlink(skb, dev, ifm, extack, tb, ifname, 0);
3007 errout:
3008 return err;
3011 static int rtnl_group_dellink(const struct net *net, int group)
3013 struct net_device *dev, *aux;
3014 LIST_HEAD(list_kill);
3015 bool found = false;
3017 if (!group)
3018 return -EPERM;
3020 for_each_netdev(net, dev) {
3021 if (dev->group == group) {
3022 const struct rtnl_link_ops *ops;
3024 found = true;
3025 ops = dev->rtnl_link_ops;
3026 if (!ops || !ops->dellink)
3027 return -EOPNOTSUPP;
3031 if (!found)
3032 return -ENODEV;
3034 for_each_netdev_safe(net, dev, aux) {
3035 if (dev->group == group) {
3036 const struct rtnl_link_ops *ops;
3038 ops = dev->rtnl_link_ops;
3039 ops->dellink(dev, &list_kill);
3042 unregister_netdevice_many(&list_kill);
3044 return 0;
3047 int rtnl_delete_link(struct net_device *dev)
3049 const struct rtnl_link_ops *ops;
3050 LIST_HEAD(list_kill);
3052 ops = dev->rtnl_link_ops;
3053 if (!ops || !ops->dellink)
3054 return -EOPNOTSUPP;
3056 ops->dellink(dev, &list_kill);
3057 unregister_netdevice_many(&list_kill);
3059 return 0;
3061 EXPORT_SYMBOL_GPL(rtnl_delete_link);
3063 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
3064 struct netlink_ext_ack *extack)
3066 struct net *net = sock_net(skb->sk);
3067 struct net *tgt_net = net;
3068 struct net_device *dev = NULL;
3069 struct ifinfomsg *ifm;
3070 struct nlattr *tb[IFLA_MAX+1];
3071 int err;
3072 int netnsid = -1;
3074 err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3075 ifla_policy, extack);
3076 if (err < 0)
3077 return err;
3079 err = rtnl_ensure_unique_netns(tb, extack, true);
3080 if (err < 0)
3081 return err;
3083 if (tb[IFLA_TARGET_NETNSID]) {
3084 netnsid = nla_get_s32(tb[IFLA_TARGET_NETNSID]);
3085 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(skb).sk, netnsid);
3086 if (IS_ERR(tgt_net))
3087 return PTR_ERR(tgt_net);
3090 err = -EINVAL;
3091 ifm = nlmsg_data(nlh);
3092 if (ifm->ifi_index > 0)
3093 dev = __dev_get_by_index(tgt_net, ifm->ifi_index);
3094 else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3095 dev = rtnl_dev_get(net, tb[IFLA_IFNAME],
3096 tb[IFLA_ALT_IFNAME], NULL);
3097 else if (tb[IFLA_GROUP])
3098 err = rtnl_group_dellink(tgt_net, nla_get_u32(tb[IFLA_GROUP]));
3099 else
3100 goto out;
3102 if (!dev) {
3103 if (tb[IFLA_IFNAME] || ifm->ifi_index > 0)
3104 err = -ENODEV;
3106 goto out;
3109 err = rtnl_delete_link(dev);
3111 out:
3112 if (netnsid >= 0)
3113 put_net(tgt_net);
3115 return err;
3118 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
3120 unsigned int old_flags;
3121 int err;
3123 old_flags = dev->flags;
3124 if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
3125 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm),
3126 NULL);
3127 if (err < 0)
3128 return err;
3131 if (dev->rtnl_link_state == RTNL_LINK_INITIALIZED) {
3132 __dev_notify_flags(dev, old_flags, (old_flags ^ dev->flags));
3133 } else {
3134 dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
3135 __dev_notify_flags(dev, old_flags, ~0U);
3137 return 0;
3139 EXPORT_SYMBOL(rtnl_configure_link);
3141 struct net_device *rtnl_create_link(struct net *net, const char *ifname,
3142 unsigned char name_assign_type,
3143 const struct rtnl_link_ops *ops,
3144 struct nlattr *tb[],
3145 struct netlink_ext_ack *extack)
3147 struct net_device *dev;
3148 unsigned int num_tx_queues = 1;
3149 unsigned int num_rx_queues = 1;
3151 if (tb[IFLA_NUM_TX_QUEUES])
3152 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
3153 else if (ops->get_num_tx_queues)
3154 num_tx_queues = ops->get_num_tx_queues();
3156 if (tb[IFLA_NUM_RX_QUEUES])
3157 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
3158 else if (ops->get_num_rx_queues)
3159 num_rx_queues = ops->get_num_rx_queues();
3161 if (num_tx_queues < 1 || num_tx_queues > 4096) {
3162 NL_SET_ERR_MSG(extack, "Invalid number of transmit queues");
3163 return ERR_PTR(-EINVAL);
3166 if (num_rx_queues < 1 || num_rx_queues > 4096) {
3167 NL_SET_ERR_MSG(extack, "Invalid number of receive queues");
3168 return ERR_PTR(-EINVAL);
3171 dev = alloc_netdev_mqs(ops->priv_size, ifname, name_assign_type,
3172 ops->setup, num_tx_queues, num_rx_queues);
3173 if (!dev)
3174 return ERR_PTR(-ENOMEM);
3176 dev_net_set(dev, net);
3177 dev->rtnl_link_ops = ops;
3178 dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
3180 if (tb[IFLA_MTU]) {
3181 u32 mtu = nla_get_u32(tb[IFLA_MTU]);
3182 int err;
3184 err = dev_validate_mtu(dev, mtu, extack);
3185 if (err) {
3186 free_netdev(dev);
3187 return ERR_PTR(err);
3189 dev->mtu = mtu;
3191 if (tb[IFLA_ADDRESS]) {
3192 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
3193 nla_len(tb[IFLA_ADDRESS]));
3194 dev->addr_assign_type = NET_ADDR_SET;
3196 if (tb[IFLA_BROADCAST])
3197 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
3198 nla_len(tb[IFLA_BROADCAST]));
3199 if (tb[IFLA_TXQLEN])
3200 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
3201 if (tb[IFLA_OPERSTATE])
3202 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
3203 if (tb[IFLA_LINKMODE])
3204 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
3205 if (tb[IFLA_GROUP])
3206 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
3207 if (tb[IFLA_GSO_MAX_SIZE])
3208 netif_set_gso_max_size(dev, nla_get_u32(tb[IFLA_GSO_MAX_SIZE]));
3209 if (tb[IFLA_GSO_MAX_SEGS])
3210 dev->gso_max_segs = nla_get_u32(tb[IFLA_GSO_MAX_SEGS]);
3212 return dev;
3214 EXPORT_SYMBOL(rtnl_create_link);
3216 static int rtnl_group_changelink(const struct sk_buff *skb,
3217 struct net *net, int group,
3218 struct ifinfomsg *ifm,
3219 struct netlink_ext_ack *extack,
3220 struct nlattr **tb)
3222 struct net_device *dev, *aux;
3223 int err;
3225 for_each_netdev_safe(net, dev, aux) {
3226 if (dev->group == group) {
3227 err = do_setlink(skb, dev, ifm, extack, tb, NULL, 0);
3228 if (err < 0)
3229 return err;
3233 return 0;
3236 static int __rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3237 struct nlattr **attr, struct netlink_ext_ack *extack)
3239 struct nlattr *slave_attr[RTNL_SLAVE_MAX_TYPE + 1];
3240 unsigned char name_assign_type = NET_NAME_USER;
3241 struct nlattr *linkinfo[IFLA_INFO_MAX + 1];
3242 const struct rtnl_link_ops *m_ops = NULL;
3243 struct net_device *master_dev = NULL;
3244 struct net *net = sock_net(skb->sk);
3245 const struct rtnl_link_ops *ops;
3246 struct nlattr *tb[IFLA_MAX + 1];
3247 struct net *dest_net, *link_net;
3248 struct nlattr **slave_data;
3249 char kind[MODULE_NAME_LEN];
3250 struct net_device *dev;
3251 struct ifinfomsg *ifm;
3252 char ifname[IFNAMSIZ];
3253 struct nlattr **data;
3254 int err;
3256 #ifdef CONFIG_MODULES
3257 replay:
3258 #endif
3259 err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3260 ifla_policy, extack);
3261 if (err < 0)
3262 return err;
3264 err = rtnl_ensure_unique_netns(tb, extack, false);
3265 if (err < 0)
3266 return err;
3268 if (tb[IFLA_IFNAME])
3269 nla_strscpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
3270 else
3271 ifname[0] = '\0';
3273 ifm = nlmsg_data(nlh);
3274 if (ifm->ifi_index > 0)
3275 dev = __dev_get_by_index(net, ifm->ifi_index);
3276 else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3277 dev = rtnl_dev_get(net, NULL, tb[IFLA_ALT_IFNAME], ifname);
3278 else
3279 dev = NULL;
3281 if (dev) {
3282 master_dev = netdev_master_upper_dev_get(dev);
3283 if (master_dev)
3284 m_ops = master_dev->rtnl_link_ops;
3287 err = validate_linkmsg(dev, tb);
3288 if (err < 0)
3289 return err;
3291 if (tb[IFLA_LINKINFO]) {
3292 err = nla_parse_nested_deprecated(linkinfo, IFLA_INFO_MAX,
3293 tb[IFLA_LINKINFO],
3294 ifla_info_policy, NULL);
3295 if (err < 0)
3296 return err;
3297 } else
3298 memset(linkinfo, 0, sizeof(linkinfo));
3300 if (linkinfo[IFLA_INFO_KIND]) {
3301 nla_strscpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
3302 ops = rtnl_link_ops_get(kind);
3303 } else {
3304 kind[0] = '\0';
3305 ops = NULL;
3308 data = NULL;
3309 if (ops) {
3310 if (ops->maxtype > RTNL_MAX_TYPE)
3311 return -EINVAL;
3313 if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
3314 err = nla_parse_nested_deprecated(attr, ops->maxtype,
3315 linkinfo[IFLA_INFO_DATA],
3316 ops->policy, extack);
3317 if (err < 0)
3318 return err;
3319 data = attr;
3321 if (ops->validate) {
3322 err = ops->validate(tb, data, extack);
3323 if (err < 0)
3324 return err;
3328 slave_data = NULL;
3329 if (m_ops) {
3330 if (m_ops->slave_maxtype > RTNL_SLAVE_MAX_TYPE)
3331 return -EINVAL;
3333 if (m_ops->slave_maxtype &&
3334 linkinfo[IFLA_INFO_SLAVE_DATA]) {
3335 err = nla_parse_nested_deprecated(slave_attr,
3336 m_ops->slave_maxtype,
3337 linkinfo[IFLA_INFO_SLAVE_DATA],
3338 m_ops->slave_policy,
3339 extack);
3340 if (err < 0)
3341 return err;
3342 slave_data = slave_attr;
3346 if (dev) {
3347 int status = 0;
3349 if (nlh->nlmsg_flags & NLM_F_EXCL)
3350 return -EEXIST;
3351 if (nlh->nlmsg_flags & NLM_F_REPLACE)
3352 return -EOPNOTSUPP;
3354 if (linkinfo[IFLA_INFO_DATA]) {
3355 if (!ops || ops != dev->rtnl_link_ops ||
3356 !ops->changelink)
3357 return -EOPNOTSUPP;
3359 err = ops->changelink(dev, tb, data, extack);
3360 if (err < 0)
3361 return err;
3362 status |= DO_SETLINK_NOTIFY;
3365 if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
3366 if (!m_ops || !m_ops->slave_changelink)
3367 return -EOPNOTSUPP;
3369 err = m_ops->slave_changelink(master_dev, dev, tb,
3370 slave_data, extack);
3371 if (err < 0)
3372 return err;
3373 status |= DO_SETLINK_NOTIFY;
3376 return do_setlink(skb, dev, ifm, extack, tb, ifname, status);
3379 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
3380 if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
3381 return rtnl_group_changelink(skb, net,
3382 nla_get_u32(tb[IFLA_GROUP]),
3383 ifm, extack, tb);
3384 return -ENODEV;
3387 if (tb[IFLA_MAP] || tb[IFLA_PROTINFO])
3388 return -EOPNOTSUPP;
3390 if (!ops) {
3391 #ifdef CONFIG_MODULES
3392 if (kind[0]) {
3393 __rtnl_unlock();
3394 request_module("rtnl-link-%s", kind);
3395 rtnl_lock();
3396 ops = rtnl_link_ops_get(kind);
3397 if (ops)
3398 goto replay;
3400 #endif
3401 NL_SET_ERR_MSG(extack, "Unknown device type");
3402 return -EOPNOTSUPP;
3405 if (!ops->setup)
3406 return -EOPNOTSUPP;
3408 if (!ifname[0]) {
3409 snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
3410 name_assign_type = NET_NAME_ENUM;
3413 dest_net = rtnl_link_get_net_capable(skb, net, tb, CAP_NET_ADMIN);
3414 if (IS_ERR(dest_net))
3415 return PTR_ERR(dest_net);
3417 if (tb[IFLA_LINK_NETNSID]) {
3418 int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
3420 link_net = get_net_ns_by_id(dest_net, id);
3421 if (!link_net) {
3422 NL_SET_ERR_MSG(extack, "Unknown network namespace id");
3423 err = -EINVAL;
3424 goto out;
3426 err = -EPERM;
3427 if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
3428 goto out;
3429 } else {
3430 link_net = NULL;
3433 dev = rtnl_create_link(link_net ? : dest_net, ifname,
3434 name_assign_type, ops, tb, extack);
3435 if (IS_ERR(dev)) {
3436 err = PTR_ERR(dev);
3437 goto out;
3440 dev->ifindex = ifm->ifi_index;
3442 if (ops->newlink) {
3443 err = ops->newlink(link_net ? : net, dev, tb, data, extack);
3444 /* Drivers should call free_netdev() in ->destructor
3445 * and unregister it on failure after registration
3446 * so that device could be finally freed in rtnl_unlock.
3448 if (err < 0) {
3449 /* If device is not registered at all, free it now */
3450 if (dev->reg_state == NETREG_UNINITIALIZED ||
3451 dev->reg_state == NETREG_UNREGISTERED)
3452 free_netdev(dev);
3453 goto out;
3455 } else {
3456 err = register_netdevice(dev);
3457 if (err < 0) {
3458 free_netdev(dev);
3459 goto out;
3462 err = rtnl_configure_link(dev, ifm);
3463 if (err < 0)
3464 goto out_unregister;
3465 if (link_net) {
3466 err = dev_change_net_namespace(dev, dest_net, ifname);
3467 if (err < 0)
3468 goto out_unregister;
3470 if (tb[IFLA_MASTER]) {
3471 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]), extack);
3472 if (err)
3473 goto out_unregister;
3475 out:
3476 if (link_net)
3477 put_net(link_net);
3478 put_net(dest_net);
3479 return err;
3480 out_unregister:
3481 if (ops->newlink) {
3482 LIST_HEAD(list_kill);
3484 ops->dellink(dev, &list_kill);
3485 unregister_netdevice_many(&list_kill);
3486 } else {
3487 unregister_netdevice(dev);
3489 goto out;
3492 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3493 struct netlink_ext_ack *extack)
3495 struct nlattr **attr;
3496 int ret;
3498 attr = kmalloc_array(RTNL_MAX_TYPE + 1, sizeof(*attr), GFP_KERNEL);
3499 if (!attr)
3500 return -ENOMEM;
3502 ret = __rtnl_newlink(skb, nlh, attr, extack);
3503 kfree(attr);
3504 return ret;
3507 static int rtnl_valid_getlink_req(struct sk_buff *skb,
3508 const struct nlmsghdr *nlh,
3509 struct nlattr **tb,
3510 struct netlink_ext_ack *extack)
3512 struct ifinfomsg *ifm;
3513 int i, err;
3515 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
3516 NL_SET_ERR_MSG(extack, "Invalid header for get link");
3517 return -EINVAL;
3520 if (!netlink_strict_get_check(skb))
3521 return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3522 ifla_policy, extack);
3524 ifm = nlmsg_data(nlh);
3525 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
3526 ifm->ifi_change) {
3527 NL_SET_ERR_MSG(extack, "Invalid values in header for get link request");
3528 return -EINVAL;
3531 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFLA_MAX,
3532 ifla_policy, extack);
3533 if (err)
3534 return err;
3536 for (i = 0; i <= IFLA_MAX; i++) {
3537 if (!tb[i])
3538 continue;
3540 switch (i) {
3541 case IFLA_IFNAME:
3542 case IFLA_ALT_IFNAME:
3543 case IFLA_EXT_MASK:
3544 case IFLA_TARGET_NETNSID:
3545 break;
3546 default:
3547 NL_SET_ERR_MSG(extack, "Unsupported attribute in get link request");
3548 return -EINVAL;
3552 return 0;
3555 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3556 struct netlink_ext_ack *extack)
3558 struct net *net = sock_net(skb->sk);
3559 struct net *tgt_net = net;
3560 struct ifinfomsg *ifm;
3561 struct nlattr *tb[IFLA_MAX+1];
3562 struct net_device *dev = NULL;
3563 struct sk_buff *nskb;
3564 int netnsid = -1;
3565 int err;
3566 u32 ext_filter_mask = 0;
3568 err = rtnl_valid_getlink_req(skb, nlh, tb, extack);
3569 if (err < 0)
3570 return err;
3572 err = rtnl_ensure_unique_netns(tb, extack, true);
3573 if (err < 0)
3574 return err;
3576 if (tb[IFLA_TARGET_NETNSID]) {
3577 netnsid = nla_get_s32(tb[IFLA_TARGET_NETNSID]);
3578 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(skb).sk, netnsid);
3579 if (IS_ERR(tgt_net))
3580 return PTR_ERR(tgt_net);
3583 if (tb[IFLA_EXT_MASK])
3584 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
3586 err = -EINVAL;
3587 ifm = nlmsg_data(nlh);
3588 if (ifm->ifi_index > 0)
3589 dev = __dev_get_by_index(tgt_net, ifm->ifi_index);
3590 else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3591 dev = rtnl_dev_get(tgt_net, tb[IFLA_IFNAME],
3592 tb[IFLA_ALT_IFNAME], NULL);
3593 else
3594 goto out;
3596 err = -ENODEV;
3597 if (dev == NULL)
3598 goto out;
3600 err = -ENOBUFS;
3601 nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
3602 if (nskb == NULL)
3603 goto out;
3605 err = rtnl_fill_ifinfo(nskb, dev, net,
3606 RTM_NEWLINK, NETLINK_CB(skb).portid,
3607 nlh->nlmsg_seq, 0, 0, ext_filter_mask,
3608 0, NULL, 0, netnsid, GFP_KERNEL);
3609 if (err < 0) {
3610 /* -EMSGSIZE implies BUG in if_nlmsg_size */
3611 WARN_ON(err == -EMSGSIZE);
3612 kfree_skb(nskb);
3613 } else
3614 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
3615 out:
3616 if (netnsid >= 0)
3617 put_net(tgt_net);
3619 return err;
3622 static int rtnl_alt_ifname(int cmd, struct net_device *dev, struct nlattr *attr,
3623 bool *changed, struct netlink_ext_ack *extack)
3625 char *alt_ifname;
3626 int err;
3628 err = nla_validate(attr, attr->nla_len, IFLA_MAX, ifla_policy, extack);
3629 if (err)
3630 return err;
3632 alt_ifname = nla_strdup(attr, GFP_KERNEL);
3633 if (!alt_ifname)
3634 return -ENOMEM;
3636 if (cmd == RTM_NEWLINKPROP) {
3637 err = netdev_name_node_alt_create(dev, alt_ifname);
3638 if (!err)
3639 alt_ifname = NULL;
3640 } else if (cmd == RTM_DELLINKPROP) {
3641 err = netdev_name_node_alt_destroy(dev, alt_ifname);
3642 } else {
3643 WARN_ON_ONCE(1);
3644 err = -EINVAL;
3647 kfree(alt_ifname);
3648 if (!err)
3649 *changed = true;
3650 return err;
3653 static int rtnl_linkprop(int cmd, struct sk_buff *skb, struct nlmsghdr *nlh,
3654 struct netlink_ext_ack *extack)
3656 struct net *net = sock_net(skb->sk);
3657 struct nlattr *tb[IFLA_MAX + 1];
3658 struct net_device *dev;
3659 struct ifinfomsg *ifm;
3660 bool changed = false;
3661 struct nlattr *attr;
3662 int err, rem;
3664 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
3665 if (err)
3666 return err;
3668 err = rtnl_ensure_unique_netns(tb, extack, true);
3669 if (err)
3670 return err;
3672 ifm = nlmsg_data(nlh);
3673 if (ifm->ifi_index > 0)
3674 dev = __dev_get_by_index(net, ifm->ifi_index);
3675 else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3676 dev = rtnl_dev_get(net, tb[IFLA_IFNAME],
3677 tb[IFLA_ALT_IFNAME], NULL);
3678 else
3679 return -EINVAL;
3681 if (!dev)
3682 return -ENODEV;
3684 if (!tb[IFLA_PROP_LIST])
3685 return 0;
3687 nla_for_each_nested(attr, tb[IFLA_PROP_LIST], rem) {
3688 switch (nla_type(attr)) {
3689 case IFLA_ALT_IFNAME:
3690 err = rtnl_alt_ifname(cmd, dev, attr, &changed, extack);
3691 if (err)
3692 return err;
3693 break;
3697 if (changed)
3698 netdev_state_change(dev);
3699 return 0;
3702 static int rtnl_newlinkprop(struct sk_buff *skb, struct nlmsghdr *nlh,
3703 struct netlink_ext_ack *extack)
3705 return rtnl_linkprop(RTM_NEWLINKPROP, skb, nlh, extack);
3708 static int rtnl_dellinkprop(struct sk_buff *skb, struct nlmsghdr *nlh,
3709 struct netlink_ext_ack *extack)
3711 return rtnl_linkprop(RTM_DELLINKPROP, skb, nlh, extack);
3714 static u32 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
3716 struct net *net = sock_net(skb->sk);
3717 size_t min_ifinfo_dump_size = 0;
3718 struct nlattr *tb[IFLA_MAX+1];
3719 u32 ext_filter_mask = 0;
3720 struct net_device *dev;
3721 int hdrlen;
3723 /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
3724 hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
3725 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
3727 if (nlmsg_parse_deprecated(nlh, hdrlen, tb, IFLA_MAX, ifla_policy, NULL) >= 0) {
3728 if (tb[IFLA_EXT_MASK])
3729 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
3732 if (!ext_filter_mask)
3733 return NLMSG_GOODSIZE;
3735 * traverse the list of net devices and compute the minimum
3736 * buffer size based upon the filter mask.
3738 rcu_read_lock();
3739 for_each_netdev_rcu(net, dev) {
3740 min_ifinfo_dump_size = max(min_ifinfo_dump_size,
3741 if_nlmsg_size(dev, ext_filter_mask));
3743 rcu_read_unlock();
3745 return nlmsg_total_size(min_ifinfo_dump_size);
3748 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
3750 int idx;
3751 int s_idx = cb->family;
3752 int type = cb->nlh->nlmsg_type - RTM_BASE;
3753 int ret = 0;
3755 if (s_idx == 0)
3756 s_idx = 1;
3758 for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
3759 struct rtnl_link __rcu **tab;
3760 struct rtnl_link *link;
3761 rtnl_dumpit_func dumpit;
3763 if (idx < s_idx || idx == PF_PACKET)
3764 continue;
3766 if (type < 0 || type >= RTM_NR_MSGTYPES)
3767 continue;
3769 tab = rcu_dereference_rtnl(rtnl_msg_handlers[idx]);
3770 if (!tab)
3771 continue;
3773 link = rcu_dereference_rtnl(tab[type]);
3774 if (!link)
3775 continue;
3777 dumpit = link->dumpit;
3778 if (!dumpit)
3779 continue;
3781 if (idx > s_idx) {
3782 memset(&cb->args[0], 0, sizeof(cb->args));
3783 cb->prev_seq = 0;
3784 cb->seq = 0;
3786 ret = dumpit(skb, cb);
3787 if (ret)
3788 break;
3790 cb->family = idx;
3792 return skb->len ? : ret;
3795 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
3796 unsigned int change,
3797 u32 event, gfp_t flags, int *new_nsid,
3798 int new_ifindex)
3800 struct net *net = dev_net(dev);
3801 struct sk_buff *skb;
3802 int err = -ENOBUFS;
3803 size_t if_info_size;
3805 skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
3806 if (skb == NULL)
3807 goto errout;
3809 err = rtnl_fill_ifinfo(skb, dev, dev_net(dev),
3810 type, 0, 0, change, 0, 0, event,
3811 new_nsid, new_ifindex, -1, flags);
3812 if (err < 0) {
3813 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
3814 WARN_ON(err == -EMSGSIZE);
3815 kfree_skb(skb);
3816 goto errout;
3818 return skb;
3819 errout:
3820 if (err < 0)
3821 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
3822 return NULL;
3825 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags)
3827 struct net *net = dev_net(dev);
3829 rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
3832 static void rtmsg_ifinfo_event(int type, struct net_device *dev,
3833 unsigned int change, u32 event,
3834 gfp_t flags, int *new_nsid, int new_ifindex)
3836 struct sk_buff *skb;
3838 if (dev->reg_state != NETREG_REGISTERED)
3839 return;
3841 skb = rtmsg_ifinfo_build_skb(type, dev, change, event, flags, new_nsid,
3842 new_ifindex);
3843 if (skb)
3844 rtmsg_ifinfo_send(skb, dev, flags);
3847 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
3848 gfp_t flags)
3850 rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
3851 NULL, 0);
3854 void rtmsg_ifinfo_newnet(int type, struct net_device *dev, unsigned int change,
3855 gfp_t flags, int *new_nsid, int new_ifindex)
3857 rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
3858 new_nsid, new_ifindex);
3861 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
3862 struct net_device *dev,
3863 u8 *addr, u16 vid, u32 pid, u32 seq,
3864 int type, unsigned int flags,
3865 int nlflags, u16 ndm_state)
3867 struct nlmsghdr *nlh;
3868 struct ndmsg *ndm;
3870 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
3871 if (!nlh)
3872 return -EMSGSIZE;
3874 ndm = nlmsg_data(nlh);
3875 ndm->ndm_family = AF_BRIDGE;
3876 ndm->ndm_pad1 = 0;
3877 ndm->ndm_pad2 = 0;
3878 ndm->ndm_flags = flags;
3879 ndm->ndm_type = 0;
3880 ndm->ndm_ifindex = dev->ifindex;
3881 ndm->ndm_state = ndm_state;
3883 if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
3884 goto nla_put_failure;
3885 if (vid)
3886 if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
3887 goto nla_put_failure;
3889 nlmsg_end(skb, nlh);
3890 return 0;
3892 nla_put_failure:
3893 nlmsg_cancel(skb, nlh);
3894 return -EMSGSIZE;
3897 static inline size_t rtnl_fdb_nlmsg_size(void)
3899 return NLMSG_ALIGN(sizeof(struct ndmsg)) +
3900 nla_total_size(ETH_ALEN) + /* NDA_LLADDR */
3901 nla_total_size(sizeof(u16)) + /* NDA_VLAN */
3905 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type,
3906 u16 ndm_state)
3908 struct net *net = dev_net(dev);
3909 struct sk_buff *skb;
3910 int err = -ENOBUFS;
3912 skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
3913 if (!skb)
3914 goto errout;
3916 err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
3917 0, 0, type, NTF_SELF, 0, ndm_state);
3918 if (err < 0) {
3919 kfree_skb(skb);
3920 goto errout;
3923 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
3924 return;
3925 errout:
3926 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
3930 * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
3932 int ndo_dflt_fdb_add(struct ndmsg *ndm,
3933 struct nlattr *tb[],
3934 struct net_device *dev,
3935 const unsigned char *addr, u16 vid,
3936 u16 flags)
3938 int err = -EINVAL;
3940 /* If aging addresses are supported device will need to
3941 * implement its own handler for this.
3943 if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
3944 pr_info("%s: FDB only supports static addresses\n", dev->name);
3945 return err;
3948 if (vid) {
3949 pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
3950 return err;
3953 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
3954 err = dev_uc_add_excl(dev, addr);
3955 else if (is_multicast_ether_addr(addr))
3956 err = dev_mc_add_excl(dev, addr);
3958 /* Only return duplicate errors if NLM_F_EXCL is set */
3959 if (err == -EEXIST && !(flags & NLM_F_EXCL))
3960 err = 0;
3962 return err;
3964 EXPORT_SYMBOL(ndo_dflt_fdb_add);
3966 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid,
3967 struct netlink_ext_ack *extack)
3969 u16 vid = 0;
3971 if (vlan_attr) {
3972 if (nla_len(vlan_attr) != sizeof(u16)) {
3973 NL_SET_ERR_MSG(extack, "invalid vlan attribute size");
3974 return -EINVAL;
3977 vid = nla_get_u16(vlan_attr);
3979 if (!vid || vid >= VLAN_VID_MASK) {
3980 NL_SET_ERR_MSG(extack, "invalid vlan id");
3981 return -EINVAL;
3984 *p_vid = vid;
3985 return 0;
3988 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh,
3989 struct netlink_ext_ack *extack)
3991 struct net *net = sock_net(skb->sk);
3992 struct ndmsg *ndm;
3993 struct nlattr *tb[NDA_MAX+1];
3994 struct net_device *dev;
3995 u8 *addr;
3996 u16 vid;
3997 int err;
3999 err = nlmsg_parse_deprecated(nlh, sizeof(*ndm), tb, NDA_MAX, NULL,
4000 extack);
4001 if (err < 0)
4002 return err;
4004 ndm = nlmsg_data(nlh);
4005 if (ndm->ndm_ifindex == 0) {
4006 NL_SET_ERR_MSG(extack, "invalid ifindex");
4007 return -EINVAL;
4010 dev = __dev_get_by_index(net, ndm->ndm_ifindex);
4011 if (dev == NULL) {
4012 NL_SET_ERR_MSG(extack, "unknown ifindex");
4013 return -ENODEV;
4016 if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
4017 NL_SET_ERR_MSG(extack, "invalid address");
4018 return -EINVAL;
4021 if (dev->type != ARPHRD_ETHER) {
4022 NL_SET_ERR_MSG(extack, "FDB add only supported for Ethernet devices");
4023 return -EINVAL;
4026 addr = nla_data(tb[NDA_LLADDR]);
4028 err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
4029 if (err)
4030 return err;
4032 err = -EOPNOTSUPP;
4034 /* Support fdb on master device the net/bridge default case */
4035 if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
4036 netif_is_bridge_port(dev)) {
4037 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4038 const struct net_device_ops *ops = br_dev->netdev_ops;
4040 err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
4041 nlh->nlmsg_flags, extack);
4042 if (err)
4043 goto out;
4044 else
4045 ndm->ndm_flags &= ~NTF_MASTER;
4048 /* Embedded bridge, macvlan, and any other device support */
4049 if ((ndm->ndm_flags & NTF_SELF)) {
4050 if (dev->netdev_ops->ndo_fdb_add)
4051 err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
4052 vid,
4053 nlh->nlmsg_flags,
4054 extack);
4055 else
4056 err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
4057 nlh->nlmsg_flags);
4059 if (!err) {
4060 rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH,
4061 ndm->ndm_state);
4062 ndm->ndm_flags &= ~NTF_SELF;
4065 out:
4066 return err;
4070 * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
4072 int ndo_dflt_fdb_del(struct ndmsg *ndm,
4073 struct nlattr *tb[],
4074 struct net_device *dev,
4075 const unsigned char *addr, u16 vid)
4077 int err = -EINVAL;
4079 /* If aging addresses are supported device will need to
4080 * implement its own handler for this.
4082 if (!(ndm->ndm_state & NUD_PERMANENT)) {
4083 pr_info("%s: FDB only supports static addresses\n", dev->name);
4084 return err;
4087 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
4088 err = dev_uc_del(dev, addr);
4089 else if (is_multicast_ether_addr(addr))
4090 err = dev_mc_del(dev, addr);
4092 return err;
4094 EXPORT_SYMBOL(ndo_dflt_fdb_del);
4096 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh,
4097 struct netlink_ext_ack *extack)
4099 struct net *net = sock_net(skb->sk);
4100 struct ndmsg *ndm;
4101 struct nlattr *tb[NDA_MAX+1];
4102 struct net_device *dev;
4103 __u8 *addr;
4104 int err;
4105 u16 vid;
4107 if (!netlink_capable(skb, CAP_NET_ADMIN))
4108 return -EPERM;
4110 err = nlmsg_parse_deprecated(nlh, sizeof(*ndm), tb, NDA_MAX, NULL,
4111 extack);
4112 if (err < 0)
4113 return err;
4115 ndm = nlmsg_data(nlh);
4116 if (ndm->ndm_ifindex == 0) {
4117 NL_SET_ERR_MSG(extack, "invalid ifindex");
4118 return -EINVAL;
4121 dev = __dev_get_by_index(net, ndm->ndm_ifindex);
4122 if (dev == NULL) {
4123 NL_SET_ERR_MSG(extack, "unknown ifindex");
4124 return -ENODEV;
4127 if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
4128 NL_SET_ERR_MSG(extack, "invalid address");
4129 return -EINVAL;
4132 if (dev->type != ARPHRD_ETHER) {
4133 NL_SET_ERR_MSG(extack, "FDB delete only supported for Ethernet devices");
4134 return -EINVAL;
4137 addr = nla_data(tb[NDA_LLADDR]);
4139 err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
4140 if (err)
4141 return err;
4143 err = -EOPNOTSUPP;
4145 /* Support fdb on master device the net/bridge default case */
4146 if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
4147 netif_is_bridge_port(dev)) {
4148 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4149 const struct net_device_ops *ops = br_dev->netdev_ops;
4151 if (ops->ndo_fdb_del)
4152 err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid);
4154 if (err)
4155 goto out;
4156 else
4157 ndm->ndm_flags &= ~NTF_MASTER;
4160 /* Embedded bridge, macvlan, and any other device support */
4161 if (ndm->ndm_flags & NTF_SELF) {
4162 if (dev->netdev_ops->ndo_fdb_del)
4163 err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr,
4164 vid);
4165 else
4166 err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
4168 if (!err) {
4169 rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH,
4170 ndm->ndm_state);
4171 ndm->ndm_flags &= ~NTF_SELF;
4174 out:
4175 return err;
4178 static int nlmsg_populate_fdb(struct sk_buff *skb,
4179 struct netlink_callback *cb,
4180 struct net_device *dev,
4181 int *idx,
4182 struct netdev_hw_addr_list *list)
4184 struct netdev_hw_addr *ha;
4185 int err;
4186 u32 portid, seq;
4188 portid = NETLINK_CB(cb->skb).portid;
4189 seq = cb->nlh->nlmsg_seq;
4191 list_for_each_entry(ha, &list->list, list) {
4192 if (*idx < cb->args[2])
4193 goto skip;
4195 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
4196 portid, seq,
4197 RTM_NEWNEIGH, NTF_SELF,
4198 NLM_F_MULTI, NUD_PERMANENT);
4199 if (err < 0)
4200 return err;
4201 skip:
4202 *idx += 1;
4204 return 0;
4208 * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
4209 * @skb: socket buffer to store message in
4210 * @cb: netlink callback
4211 * @dev: netdevice
4212 * @filter_dev: ignored
4213 * @idx: the number of FDB table entries dumped is added to *@idx
4215 * Default netdevice operation to dump the existing unicast address list.
4216 * Returns number of addresses from list put in skb.
4218 int ndo_dflt_fdb_dump(struct sk_buff *skb,
4219 struct netlink_callback *cb,
4220 struct net_device *dev,
4221 struct net_device *filter_dev,
4222 int *idx)
4224 int err;
4226 if (dev->type != ARPHRD_ETHER)
4227 return -EINVAL;
4229 netif_addr_lock_bh(dev);
4230 err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->uc);
4231 if (err)
4232 goto out;
4233 err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->mc);
4234 out:
4235 netif_addr_unlock_bh(dev);
4236 return err;
4238 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
4240 static int valid_fdb_dump_strict(const struct nlmsghdr *nlh,
4241 int *br_idx, int *brport_idx,
4242 struct netlink_ext_ack *extack)
4244 struct nlattr *tb[NDA_MAX + 1];
4245 struct ndmsg *ndm;
4246 int err, i;
4248 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ndm))) {
4249 NL_SET_ERR_MSG(extack, "Invalid header for fdb dump request");
4250 return -EINVAL;
4253 ndm = nlmsg_data(nlh);
4254 if (ndm->ndm_pad1 || ndm->ndm_pad2 || ndm->ndm_state ||
4255 ndm->ndm_flags || ndm->ndm_type) {
4256 NL_SET_ERR_MSG(extack, "Invalid values in header for fdb dump request");
4257 return -EINVAL;
4260 err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct ndmsg), tb,
4261 NDA_MAX, NULL, extack);
4262 if (err < 0)
4263 return err;
4265 *brport_idx = ndm->ndm_ifindex;
4266 for (i = 0; i <= NDA_MAX; ++i) {
4267 if (!tb[i])
4268 continue;
4270 switch (i) {
4271 case NDA_IFINDEX:
4272 if (nla_len(tb[i]) != sizeof(u32)) {
4273 NL_SET_ERR_MSG(extack, "Invalid IFINDEX attribute in fdb dump request");
4274 return -EINVAL;
4276 *brport_idx = nla_get_u32(tb[NDA_IFINDEX]);
4277 break;
4278 case NDA_MASTER:
4279 if (nla_len(tb[i]) != sizeof(u32)) {
4280 NL_SET_ERR_MSG(extack, "Invalid MASTER attribute in fdb dump request");
4281 return -EINVAL;
4283 *br_idx = nla_get_u32(tb[NDA_MASTER]);
4284 break;
4285 default:
4286 NL_SET_ERR_MSG(extack, "Unsupported attribute in fdb dump request");
4287 return -EINVAL;
4291 return 0;
4294 static int valid_fdb_dump_legacy(const struct nlmsghdr *nlh,
4295 int *br_idx, int *brport_idx,
4296 struct netlink_ext_ack *extack)
4298 struct nlattr *tb[IFLA_MAX+1];
4299 int err;
4301 /* A hack to preserve kernel<->userspace interface.
4302 * Before Linux v4.12 this code accepted ndmsg since iproute2 v3.3.0.
4303 * However, ndmsg is shorter than ifinfomsg thus nlmsg_parse() bails.
4304 * So, check for ndmsg with an optional u32 attribute (not used here).
4305 * Fortunately these sizes don't conflict with the size of ifinfomsg
4306 * with an optional attribute.
4308 if (nlmsg_len(nlh) != sizeof(struct ndmsg) &&
4309 (nlmsg_len(nlh) != sizeof(struct ndmsg) +
4310 nla_attr_size(sizeof(u32)))) {
4311 struct ifinfomsg *ifm;
4313 err = nlmsg_parse_deprecated(nlh, sizeof(struct ifinfomsg),
4314 tb, IFLA_MAX, ifla_policy,
4315 extack);
4316 if (err < 0) {
4317 return -EINVAL;
4318 } else if (err == 0) {
4319 if (tb[IFLA_MASTER])
4320 *br_idx = nla_get_u32(tb[IFLA_MASTER]);
4323 ifm = nlmsg_data(nlh);
4324 *brport_idx = ifm->ifi_index;
4326 return 0;
4329 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
4331 struct net_device *dev;
4332 struct net_device *br_dev = NULL;
4333 const struct net_device_ops *ops = NULL;
4334 const struct net_device_ops *cops = NULL;
4335 struct net *net = sock_net(skb->sk);
4336 struct hlist_head *head;
4337 int brport_idx = 0;
4338 int br_idx = 0;
4339 int h, s_h;
4340 int idx = 0, s_idx;
4341 int err = 0;
4342 int fidx = 0;
4344 if (cb->strict_check)
4345 err = valid_fdb_dump_strict(cb->nlh, &br_idx, &brport_idx,
4346 cb->extack);
4347 else
4348 err = valid_fdb_dump_legacy(cb->nlh, &br_idx, &brport_idx,
4349 cb->extack);
4350 if (err < 0)
4351 return err;
4353 if (br_idx) {
4354 br_dev = __dev_get_by_index(net, br_idx);
4355 if (!br_dev)
4356 return -ENODEV;
4358 ops = br_dev->netdev_ops;
4361 s_h = cb->args[0];
4362 s_idx = cb->args[1];
4364 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4365 idx = 0;
4366 head = &net->dev_index_head[h];
4367 hlist_for_each_entry(dev, head, index_hlist) {
4369 if (brport_idx && (dev->ifindex != brport_idx))
4370 continue;
4372 if (!br_idx) { /* user did not specify a specific bridge */
4373 if (netif_is_bridge_port(dev)) {
4374 br_dev = netdev_master_upper_dev_get(dev);
4375 cops = br_dev->netdev_ops;
4377 } else {
4378 if (dev != br_dev &&
4379 !netif_is_bridge_port(dev))
4380 continue;
4382 if (br_dev != netdev_master_upper_dev_get(dev) &&
4383 !(dev->priv_flags & IFF_EBRIDGE))
4384 continue;
4385 cops = ops;
4388 if (idx < s_idx)
4389 goto cont;
4391 if (netif_is_bridge_port(dev)) {
4392 if (cops && cops->ndo_fdb_dump) {
4393 err = cops->ndo_fdb_dump(skb, cb,
4394 br_dev, dev,
4395 &fidx);
4396 if (err == -EMSGSIZE)
4397 goto out;
4401 if (dev->netdev_ops->ndo_fdb_dump)
4402 err = dev->netdev_ops->ndo_fdb_dump(skb, cb,
4403 dev, NULL,
4404 &fidx);
4405 else
4406 err = ndo_dflt_fdb_dump(skb, cb, dev, NULL,
4407 &fidx);
4408 if (err == -EMSGSIZE)
4409 goto out;
4411 cops = NULL;
4413 /* reset fdb offset to 0 for rest of the interfaces */
4414 cb->args[2] = 0;
4415 fidx = 0;
4416 cont:
4417 idx++;
4421 out:
4422 cb->args[0] = h;
4423 cb->args[1] = idx;
4424 cb->args[2] = fidx;
4426 return skb->len;
4429 static int valid_fdb_get_strict(const struct nlmsghdr *nlh,
4430 struct nlattr **tb, u8 *ndm_flags,
4431 int *br_idx, int *brport_idx, u8 **addr,
4432 u16 *vid, struct netlink_ext_ack *extack)
4434 struct ndmsg *ndm;
4435 int err, i;
4437 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ndm))) {
4438 NL_SET_ERR_MSG(extack, "Invalid header for fdb get request");
4439 return -EINVAL;
4442 ndm = nlmsg_data(nlh);
4443 if (ndm->ndm_pad1 || ndm->ndm_pad2 || ndm->ndm_state ||
4444 ndm->ndm_type) {
4445 NL_SET_ERR_MSG(extack, "Invalid values in header for fdb get request");
4446 return -EINVAL;
4449 if (ndm->ndm_flags & ~(NTF_MASTER | NTF_SELF)) {
4450 NL_SET_ERR_MSG(extack, "Invalid flags in header for fdb get request");
4451 return -EINVAL;
4454 err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct ndmsg), tb,
4455 NDA_MAX, nda_policy, extack);
4456 if (err < 0)
4457 return err;
4459 *ndm_flags = ndm->ndm_flags;
4460 *brport_idx = ndm->ndm_ifindex;
4461 for (i = 0; i <= NDA_MAX; ++i) {
4462 if (!tb[i])
4463 continue;
4465 switch (i) {
4466 case NDA_MASTER:
4467 *br_idx = nla_get_u32(tb[i]);
4468 break;
4469 case NDA_LLADDR:
4470 if (nla_len(tb[i]) != ETH_ALEN) {
4471 NL_SET_ERR_MSG(extack, "Invalid address in fdb get request");
4472 return -EINVAL;
4474 *addr = nla_data(tb[i]);
4475 break;
4476 case NDA_VLAN:
4477 err = fdb_vid_parse(tb[i], vid, extack);
4478 if (err)
4479 return err;
4480 break;
4481 case NDA_VNI:
4482 break;
4483 default:
4484 NL_SET_ERR_MSG(extack, "Unsupported attribute in fdb get request");
4485 return -EINVAL;
4489 return 0;
4492 static int rtnl_fdb_get(struct sk_buff *in_skb, struct nlmsghdr *nlh,
4493 struct netlink_ext_ack *extack)
4495 struct net_device *dev = NULL, *br_dev = NULL;
4496 const struct net_device_ops *ops = NULL;
4497 struct net *net = sock_net(in_skb->sk);
4498 struct nlattr *tb[NDA_MAX + 1];
4499 struct sk_buff *skb;
4500 int brport_idx = 0;
4501 u8 ndm_flags = 0;
4502 int br_idx = 0;
4503 u8 *addr = NULL;
4504 u16 vid = 0;
4505 int err;
4507 err = valid_fdb_get_strict(nlh, tb, &ndm_flags, &br_idx,
4508 &brport_idx, &addr, &vid, extack);
4509 if (err < 0)
4510 return err;
4512 if (!addr) {
4513 NL_SET_ERR_MSG(extack, "Missing lookup address for fdb get request");
4514 return -EINVAL;
4517 if (brport_idx) {
4518 dev = __dev_get_by_index(net, brport_idx);
4519 if (!dev) {
4520 NL_SET_ERR_MSG(extack, "Unknown device ifindex");
4521 return -ENODEV;
4525 if (br_idx) {
4526 if (dev) {
4527 NL_SET_ERR_MSG(extack, "Master and device are mutually exclusive");
4528 return -EINVAL;
4531 br_dev = __dev_get_by_index(net, br_idx);
4532 if (!br_dev) {
4533 NL_SET_ERR_MSG(extack, "Invalid master ifindex");
4534 return -EINVAL;
4536 ops = br_dev->netdev_ops;
4539 if (dev) {
4540 if (!ndm_flags || (ndm_flags & NTF_MASTER)) {
4541 if (!netif_is_bridge_port(dev)) {
4542 NL_SET_ERR_MSG(extack, "Device is not a bridge port");
4543 return -EINVAL;
4545 br_dev = netdev_master_upper_dev_get(dev);
4546 if (!br_dev) {
4547 NL_SET_ERR_MSG(extack, "Master of device not found");
4548 return -EINVAL;
4550 ops = br_dev->netdev_ops;
4551 } else {
4552 if (!(ndm_flags & NTF_SELF)) {
4553 NL_SET_ERR_MSG(extack, "Missing NTF_SELF");
4554 return -EINVAL;
4556 ops = dev->netdev_ops;
4560 if (!br_dev && !dev) {
4561 NL_SET_ERR_MSG(extack, "No device specified");
4562 return -ENODEV;
4565 if (!ops || !ops->ndo_fdb_get) {
4566 NL_SET_ERR_MSG(extack, "Fdb get operation not supported by device");
4567 return -EOPNOTSUPP;
4570 skb = nlmsg_new(NLMSG_GOODSIZE, GFP_KERNEL);
4571 if (!skb)
4572 return -ENOBUFS;
4574 if (br_dev)
4575 dev = br_dev;
4576 err = ops->ndo_fdb_get(skb, tb, dev, addr, vid,
4577 NETLINK_CB(in_skb).portid,
4578 nlh->nlmsg_seq, extack);
4579 if (err)
4580 goto out;
4582 return rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4583 out:
4584 kfree_skb(skb);
4585 return err;
4588 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
4589 unsigned int attrnum, unsigned int flag)
4591 if (mask & flag)
4592 return nla_put_u8(skb, attrnum, !!(flags & flag));
4593 return 0;
4596 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
4597 struct net_device *dev, u16 mode,
4598 u32 flags, u32 mask, int nlflags,
4599 u32 filter_mask,
4600 int (*vlan_fill)(struct sk_buff *skb,
4601 struct net_device *dev,
4602 u32 filter_mask))
4604 struct nlmsghdr *nlh;
4605 struct ifinfomsg *ifm;
4606 struct nlattr *br_afspec;
4607 struct nlattr *protinfo;
4608 u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
4609 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4610 int err = 0;
4612 nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
4613 if (nlh == NULL)
4614 return -EMSGSIZE;
4616 ifm = nlmsg_data(nlh);
4617 ifm->ifi_family = AF_BRIDGE;
4618 ifm->__ifi_pad = 0;
4619 ifm->ifi_type = dev->type;
4620 ifm->ifi_index = dev->ifindex;
4621 ifm->ifi_flags = dev_get_flags(dev);
4622 ifm->ifi_change = 0;
4625 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4626 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4627 nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
4628 (br_dev &&
4629 nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
4630 (dev->addr_len &&
4631 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4632 (dev->ifindex != dev_get_iflink(dev) &&
4633 nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
4634 goto nla_put_failure;
4636 br_afspec = nla_nest_start_noflag(skb, IFLA_AF_SPEC);
4637 if (!br_afspec)
4638 goto nla_put_failure;
4640 if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
4641 nla_nest_cancel(skb, br_afspec);
4642 goto nla_put_failure;
4645 if (mode != BRIDGE_MODE_UNDEF) {
4646 if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
4647 nla_nest_cancel(skb, br_afspec);
4648 goto nla_put_failure;
4651 if (vlan_fill) {
4652 err = vlan_fill(skb, dev, filter_mask);
4653 if (err) {
4654 nla_nest_cancel(skb, br_afspec);
4655 goto nla_put_failure;
4658 nla_nest_end(skb, br_afspec);
4660 protinfo = nla_nest_start(skb, IFLA_PROTINFO);
4661 if (!protinfo)
4662 goto nla_put_failure;
4664 if (brport_nla_put_flag(skb, flags, mask,
4665 IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
4666 brport_nla_put_flag(skb, flags, mask,
4667 IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
4668 brport_nla_put_flag(skb, flags, mask,
4669 IFLA_BRPORT_FAST_LEAVE,
4670 BR_MULTICAST_FAST_LEAVE) ||
4671 brport_nla_put_flag(skb, flags, mask,
4672 IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
4673 brport_nla_put_flag(skb, flags, mask,
4674 IFLA_BRPORT_LEARNING, BR_LEARNING) ||
4675 brport_nla_put_flag(skb, flags, mask,
4676 IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
4677 brport_nla_put_flag(skb, flags, mask,
4678 IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
4679 brport_nla_put_flag(skb, flags, mask,
4680 IFLA_BRPORT_PROXYARP, BR_PROXYARP) ||
4681 brport_nla_put_flag(skb, flags, mask,
4682 IFLA_BRPORT_MCAST_FLOOD, BR_MCAST_FLOOD) ||
4683 brport_nla_put_flag(skb, flags, mask,
4684 IFLA_BRPORT_BCAST_FLOOD, BR_BCAST_FLOOD)) {
4685 nla_nest_cancel(skb, protinfo);
4686 goto nla_put_failure;
4689 nla_nest_end(skb, protinfo);
4691 nlmsg_end(skb, nlh);
4692 return 0;
4693 nla_put_failure:
4694 nlmsg_cancel(skb, nlh);
4695 return err ? err : -EMSGSIZE;
4697 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink);
4699 static int valid_bridge_getlink_req(const struct nlmsghdr *nlh,
4700 bool strict_check, u32 *filter_mask,
4701 struct netlink_ext_ack *extack)
4703 struct nlattr *tb[IFLA_MAX+1];
4704 int err, i;
4706 if (strict_check) {
4707 struct ifinfomsg *ifm;
4709 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
4710 NL_SET_ERR_MSG(extack, "Invalid header for bridge link dump");
4711 return -EINVAL;
4714 ifm = nlmsg_data(nlh);
4715 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
4716 ifm->ifi_change || ifm->ifi_index) {
4717 NL_SET_ERR_MSG(extack, "Invalid values in header for bridge link dump request");
4718 return -EINVAL;
4721 err = nlmsg_parse_deprecated_strict(nlh,
4722 sizeof(struct ifinfomsg),
4723 tb, IFLA_MAX, ifla_policy,
4724 extack);
4725 } else {
4726 err = nlmsg_parse_deprecated(nlh, sizeof(struct ifinfomsg),
4727 tb, IFLA_MAX, ifla_policy,
4728 extack);
4730 if (err < 0)
4731 return err;
4733 /* new attributes should only be added with strict checking */
4734 for (i = 0; i <= IFLA_MAX; ++i) {
4735 if (!tb[i])
4736 continue;
4738 switch (i) {
4739 case IFLA_EXT_MASK:
4740 *filter_mask = nla_get_u32(tb[i]);
4741 break;
4742 default:
4743 if (strict_check) {
4744 NL_SET_ERR_MSG(extack, "Unsupported attribute in bridge link dump request");
4745 return -EINVAL;
4750 return 0;
4753 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
4755 const struct nlmsghdr *nlh = cb->nlh;
4756 struct net *net = sock_net(skb->sk);
4757 struct net_device *dev;
4758 int idx = 0;
4759 u32 portid = NETLINK_CB(cb->skb).portid;
4760 u32 seq = nlh->nlmsg_seq;
4761 u32 filter_mask = 0;
4762 int err;
4764 err = valid_bridge_getlink_req(nlh, cb->strict_check, &filter_mask,
4765 cb->extack);
4766 if (err < 0 && cb->strict_check)
4767 return err;
4769 rcu_read_lock();
4770 for_each_netdev_rcu(net, dev) {
4771 const struct net_device_ops *ops = dev->netdev_ops;
4772 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4774 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
4775 if (idx >= cb->args[0]) {
4776 err = br_dev->netdev_ops->ndo_bridge_getlink(
4777 skb, portid, seq, dev,
4778 filter_mask, NLM_F_MULTI);
4779 if (err < 0 && err != -EOPNOTSUPP) {
4780 if (likely(skb->len))
4781 break;
4783 goto out_err;
4786 idx++;
4789 if (ops->ndo_bridge_getlink) {
4790 if (idx >= cb->args[0]) {
4791 err = ops->ndo_bridge_getlink(skb, portid,
4792 seq, dev,
4793 filter_mask,
4794 NLM_F_MULTI);
4795 if (err < 0 && err != -EOPNOTSUPP) {
4796 if (likely(skb->len))
4797 break;
4799 goto out_err;
4802 idx++;
4805 err = skb->len;
4806 out_err:
4807 rcu_read_unlock();
4808 cb->args[0] = idx;
4810 return err;
4813 static inline size_t bridge_nlmsg_size(void)
4815 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4816 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4817 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4818 + nla_total_size(sizeof(u32)) /* IFLA_MASTER */
4819 + nla_total_size(sizeof(u32)) /* IFLA_MTU */
4820 + nla_total_size(sizeof(u32)) /* IFLA_LINK */
4821 + nla_total_size(sizeof(u32)) /* IFLA_OPERSTATE */
4822 + nla_total_size(sizeof(u8)) /* IFLA_PROTINFO */
4823 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
4824 + nla_total_size(sizeof(u16)) /* IFLA_BRIDGE_FLAGS */
4825 + nla_total_size(sizeof(u16)); /* IFLA_BRIDGE_MODE */
4828 static int rtnl_bridge_notify(struct net_device *dev)
4830 struct net *net = dev_net(dev);
4831 struct sk_buff *skb;
4832 int err = -EOPNOTSUPP;
4834 if (!dev->netdev_ops->ndo_bridge_getlink)
4835 return 0;
4837 skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
4838 if (!skb) {
4839 err = -ENOMEM;
4840 goto errout;
4843 err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
4844 if (err < 0)
4845 goto errout;
4847 if (!skb->len)
4848 goto errout;
4850 rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
4851 return 0;
4852 errout:
4853 WARN_ON(err == -EMSGSIZE);
4854 kfree_skb(skb);
4855 if (err)
4856 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
4857 return err;
4860 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
4861 struct netlink_ext_ack *extack)
4863 struct net *net = sock_net(skb->sk);
4864 struct ifinfomsg *ifm;
4865 struct net_device *dev;
4866 struct nlattr *br_spec, *attr = NULL;
4867 int rem, err = -EOPNOTSUPP;
4868 u16 flags = 0;
4869 bool have_flags = false;
4871 if (nlmsg_len(nlh) < sizeof(*ifm))
4872 return -EINVAL;
4874 ifm = nlmsg_data(nlh);
4875 if (ifm->ifi_family != AF_BRIDGE)
4876 return -EPFNOSUPPORT;
4878 dev = __dev_get_by_index(net, ifm->ifi_index);
4879 if (!dev) {
4880 NL_SET_ERR_MSG(extack, "unknown ifindex");
4881 return -ENODEV;
4884 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
4885 if (br_spec) {
4886 nla_for_each_nested(attr, br_spec, rem) {
4887 if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
4888 if (nla_len(attr) < sizeof(flags))
4889 return -EINVAL;
4891 have_flags = true;
4892 flags = nla_get_u16(attr);
4893 break;
4898 if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
4899 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4901 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
4902 err = -EOPNOTSUPP;
4903 goto out;
4906 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags,
4907 extack);
4908 if (err)
4909 goto out;
4911 flags &= ~BRIDGE_FLAGS_MASTER;
4914 if ((flags & BRIDGE_FLAGS_SELF)) {
4915 if (!dev->netdev_ops->ndo_bridge_setlink)
4916 err = -EOPNOTSUPP;
4917 else
4918 err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
4919 flags,
4920 extack);
4921 if (!err) {
4922 flags &= ~BRIDGE_FLAGS_SELF;
4924 /* Generate event to notify upper layer of bridge
4925 * change
4927 err = rtnl_bridge_notify(dev);
4931 if (have_flags)
4932 memcpy(nla_data(attr), &flags, sizeof(flags));
4933 out:
4934 return err;
4937 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
4938 struct netlink_ext_ack *extack)
4940 struct net *net = sock_net(skb->sk);
4941 struct ifinfomsg *ifm;
4942 struct net_device *dev;
4943 struct nlattr *br_spec, *attr = NULL;
4944 int rem, err = -EOPNOTSUPP;
4945 u16 flags = 0;
4946 bool have_flags = false;
4948 if (nlmsg_len(nlh) < sizeof(*ifm))
4949 return -EINVAL;
4951 ifm = nlmsg_data(nlh);
4952 if (ifm->ifi_family != AF_BRIDGE)
4953 return -EPFNOSUPPORT;
4955 dev = __dev_get_by_index(net, ifm->ifi_index);
4956 if (!dev) {
4957 NL_SET_ERR_MSG(extack, "unknown ifindex");
4958 return -ENODEV;
4961 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
4962 if (br_spec) {
4963 nla_for_each_nested(attr, br_spec, rem) {
4964 if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
4965 if (nla_len(attr) < sizeof(flags))
4966 return -EINVAL;
4968 have_flags = true;
4969 flags = nla_get_u16(attr);
4970 break;
4975 if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
4976 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4978 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
4979 err = -EOPNOTSUPP;
4980 goto out;
4983 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
4984 if (err)
4985 goto out;
4987 flags &= ~BRIDGE_FLAGS_MASTER;
4990 if ((flags & BRIDGE_FLAGS_SELF)) {
4991 if (!dev->netdev_ops->ndo_bridge_dellink)
4992 err = -EOPNOTSUPP;
4993 else
4994 err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
4995 flags);
4997 if (!err) {
4998 flags &= ~BRIDGE_FLAGS_SELF;
5000 /* Generate event to notify upper layer of bridge
5001 * change
5003 err = rtnl_bridge_notify(dev);
5007 if (have_flags)
5008 memcpy(nla_data(attr), &flags, sizeof(flags));
5009 out:
5010 return err;
5013 static bool stats_attr_valid(unsigned int mask, int attrid, int idxattr)
5015 return (mask & IFLA_STATS_FILTER_BIT(attrid)) &&
5016 (!idxattr || idxattr == attrid);
5019 #define IFLA_OFFLOAD_XSTATS_FIRST (IFLA_OFFLOAD_XSTATS_UNSPEC + 1)
5020 static int rtnl_get_offload_stats_attr_size(int attr_id)
5022 switch (attr_id) {
5023 case IFLA_OFFLOAD_XSTATS_CPU_HIT:
5024 return sizeof(struct rtnl_link_stats64);
5027 return 0;
5030 static int rtnl_get_offload_stats(struct sk_buff *skb, struct net_device *dev,
5031 int *prividx)
5033 struct nlattr *attr = NULL;
5034 int attr_id, size;
5035 void *attr_data;
5036 int err;
5038 if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
5039 dev->netdev_ops->ndo_get_offload_stats))
5040 return -ENODATA;
5042 for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
5043 attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
5044 if (attr_id < *prividx)
5045 continue;
5047 size = rtnl_get_offload_stats_attr_size(attr_id);
5048 if (!size)
5049 continue;
5051 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
5052 continue;
5054 attr = nla_reserve_64bit(skb, attr_id, size,
5055 IFLA_OFFLOAD_XSTATS_UNSPEC);
5056 if (!attr)
5057 goto nla_put_failure;
5059 attr_data = nla_data(attr);
5060 memset(attr_data, 0, size);
5061 err = dev->netdev_ops->ndo_get_offload_stats(attr_id, dev,
5062 attr_data);
5063 if (err)
5064 goto get_offload_stats_failure;
5067 if (!attr)
5068 return -ENODATA;
5070 *prividx = 0;
5071 return 0;
5073 nla_put_failure:
5074 err = -EMSGSIZE;
5075 get_offload_stats_failure:
5076 *prividx = attr_id;
5077 return err;
5080 static int rtnl_get_offload_stats_size(const struct net_device *dev)
5082 int nla_size = 0;
5083 int attr_id;
5084 int size;
5086 if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
5087 dev->netdev_ops->ndo_get_offload_stats))
5088 return 0;
5090 for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
5091 attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
5092 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
5093 continue;
5094 size = rtnl_get_offload_stats_attr_size(attr_id);
5095 nla_size += nla_total_size_64bit(size);
5098 if (nla_size != 0)
5099 nla_size += nla_total_size(0);
5101 return nla_size;
5104 static int rtnl_fill_statsinfo(struct sk_buff *skb, struct net_device *dev,
5105 int type, u32 pid, u32 seq, u32 change,
5106 unsigned int flags, unsigned int filter_mask,
5107 int *idxattr, int *prividx)
5109 struct if_stats_msg *ifsm;
5110 struct nlmsghdr *nlh;
5111 struct nlattr *attr;
5112 int s_prividx = *prividx;
5113 int err;
5115 ASSERT_RTNL();
5117 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifsm), flags);
5118 if (!nlh)
5119 return -EMSGSIZE;
5121 ifsm = nlmsg_data(nlh);
5122 ifsm->family = PF_UNSPEC;
5123 ifsm->pad1 = 0;
5124 ifsm->pad2 = 0;
5125 ifsm->ifindex = dev->ifindex;
5126 ifsm->filter_mask = filter_mask;
5128 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, *idxattr)) {
5129 struct rtnl_link_stats64 *sp;
5131 attr = nla_reserve_64bit(skb, IFLA_STATS_LINK_64,
5132 sizeof(struct rtnl_link_stats64),
5133 IFLA_STATS_UNSPEC);
5134 if (!attr)
5135 goto nla_put_failure;
5137 sp = nla_data(attr);
5138 dev_get_stats(dev, sp);
5141 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, *idxattr)) {
5142 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
5144 if (ops && ops->fill_linkxstats) {
5145 *idxattr = IFLA_STATS_LINK_XSTATS;
5146 attr = nla_nest_start_noflag(skb,
5147 IFLA_STATS_LINK_XSTATS);
5148 if (!attr)
5149 goto nla_put_failure;
5151 err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
5152 nla_nest_end(skb, attr);
5153 if (err)
5154 goto nla_put_failure;
5155 *idxattr = 0;
5159 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE,
5160 *idxattr)) {
5161 const struct rtnl_link_ops *ops = NULL;
5162 const struct net_device *master;
5164 master = netdev_master_upper_dev_get(dev);
5165 if (master)
5166 ops = master->rtnl_link_ops;
5167 if (ops && ops->fill_linkxstats) {
5168 *idxattr = IFLA_STATS_LINK_XSTATS_SLAVE;
5169 attr = nla_nest_start_noflag(skb,
5170 IFLA_STATS_LINK_XSTATS_SLAVE);
5171 if (!attr)
5172 goto nla_put_failure;
5174 err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
5175 nla_nest_end(skb, attr);
5176 if (err)
5177 goto nla_put_failure;
5178 *idxattr = 0;
5182 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS,
5183 *idxattr)) {
5184 *idxattr = IFLA_STATS_LINK_OFFLOAD_XSTATS;
5185 attr = nla_nest_start_noflag(skb,
5186 IFLA_STATS_LINK_OFFLOAD_XSTATS);
5187 if (!attr)
5188 goto nla_put_failure;
5190 err = rtnl_get_offload_stats(skb, dev, prividx);
5191 if (err == -ENODATA)
5192 nla_nest_cancel(skb, attr);
5193 else
5194 nla_nest_end(skb, attr);
5196 if (err && err != -ENODATA)
5197 goto nla_put_failure;
5198 *idxattr = 0;
5201 if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, *idxattr)) {
5202 struct rtnl_af_ops *af_ops;
5204 *idxattr = IFLA_STATS_AF_SPEC;
5205 attr = nla_nest_start_noflag(skb, IFLA_STATS_AF_SPEC);
5206 if (!attr)
5207 goto nla_put_failure;
5209 rcu_read_lock();
5210 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
5211 if (af_ops->fill_stats_af) {
5212 struct nlattr *af;
5213 int err;
5215 af = nla_nest_start_noflag(skb,
5216 af_ops->family);
5217 if (!af) {
5218 rcu_read_unlock();
5219 goto nla_put_failure;
5221 err = af_ops->fill_stats_af(skb, dev);
5223 if (err == -ENODATA) {
5224 nla_nest_cancel(skb, af);
5225 } else if (err < 0) {
5226 rcu_read_unlock();
5227 goto nla_put_failure;
5230 nla_nest_end(skb, af);
5233 rcu_read_unlock();
5235 nla_nest_end(skb, attr);
5237 *idxattr = 0;
5240 nlmsg_end(skb, nlh);
5242 return 0;
5244 nla_put_failure:
5245 /* not a multi message or no progress mean a real error */
5246 if (!(flags & NLM_F_MULTI) || s_prividx == *prividx)
5247 nlmsg_cancel(skb, nlh);
5248 else
5249 nlmsg_end(skb, nlh);
5251 return -EMSGSIZE;
5254 static size_t if_nlmsg_stats_size(const struct net_device *dev,
5255 u32 filter_mask)
5257 size_t size = 0;
5259 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, 0))
5260 size += nla_total_size_64bit(sizeof(struct rtnl_link_stats64));
5262 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, 0)) {
5263 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
5264 int attr = IFLA_STATS_LINK_XSTATS;
5266 if (ops && ops->get_linkxstats_size) {
5267 size += nla_total_size(ops->get_linkxstats_size(dev,
5268 attr));
5269 /* for IFLA_STATS_LINK_XSTATS */
5270 size += nla_total_size(0);
5274 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE, 0)) {
5275 struct net_device *_dev = (struct net_device *)dev;
5276 const struct rtnl_link_ops *ops = NULL;
5277 const struct net_device *master;
5279 /* netdev_master_upper_dev_get can't take const */
5280 master = netdev_master_upper_dev_get(_dev);
5281 if (master)
5282 ops = master->rtnl_link_ops;
5283 if (ops && ops->get_linkxstats_size) {
5284 int attr = IFLA_STATS_LINK_XSTATS_SLAVE;
5286 size += nla_total_size(ops->get_linkxstats_size(dev,
5287 attr));
5288 /* for IFLA_STATS_LINK_XSTATS_SLAVE */
5289 size += nla_total_size(0);
5293 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS, 0))
5294 size += rtnl_get_offload_stats_size(dev);
5296 if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, 0)) {
5297 struct rtnl_af_ops *af_ops;
5299 /* for IFLA_STATS_AF_SPEC */
5300 size += nla_total_size(0);
5302 rcu_read_lock();
5303 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
5304 if (af_ops->get_stats_af_size) {
5305 size += nla_total_size(
5306 af_ops->get_stats_af_size(dev));
5308 /* for AF_* */
5309 size += nla_total_size(0);
5312 rcu_read_unlock();
5315 return size;
5318 static int rtnl_valid_stats_req(const struct nlmsghdr *nlh, bool strict_check,
5319 bool is_dump, struct netlink_ext_ack *extack)
5321 struct if_stats_msg *ifsm;
5323 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifsm))) {
5324 NL_SET_ERR_MSG(extack, "Invalid header for stats dump");
5325 return -EINVAL;
5328 if (!strict_check)
5329 return 0;
5331 ifsm = nlmsg_data(nlh);
5333 /* only requests using strict checks can pass data to influence
5334 * the dump. The legacy exception is filter_mask.
5336 if (ifsm->pad1 || ifsm->pad2 || (is_dump && ifsm->ifindex)) {
5337 NL_SET_ERR_MSG(extack, "Invalid values in header for stats dump request");
5338 return -EINVAL;
5340 if (nlmsg_attrlen(nlh, sizeof(*ifsm))) {
5341 NL_SET_ERR_MSG(extack, "Invalid attributes after stats header");
5342 return -EINVAL;
5344 if (ifsm->filter_mask >= IFLA_STATS_FILTER_BIT(IFLA_STATS_MAX + 1)) {
5345 NL_SET_ERR_MSG(extack, "Invalid stats requested through filter mask");
5346 return -EINVAL;
5349 return 0;
5352 static int rtnl_stats_get(struct sk_buff *skb, struct nlmsghdr *nlh,
5353 struct netlink_ext_ack *extack)
5355 struct net *net = sock_net(skb->sk);
5356 struct net_device *dev = NULL;
5357 int idxattr = 0, prividx = 0;
5358 struct if_stats_msg *ifsm;
5359 struct sk_buff *nskb;
5360 u32 filter_mask;
5361 int err;
5363 err = rtnl_valid_stats_req(nlh, netlink_strict_get_check(skb),
5364 false, extack);
5365 if (err)
5366 return err;
5368 ifsm = nlmsg_data(nlh);
5369 if (ifsm->ifindex > 0)
5370 dev = __dev_get_by_index(net, ifsm->ifindex);
5371 else
5372 return -EINVAL;
5374 if (!dev)
5375 return -ENODEV;
5377 filter_mask = ifsm->filter_mask;
5378 if (!filter_mask)
5379 return -EINVAL;
5381 nskb = nlmsg_new(if_nlmsg_stats_size(dev, filter_mask), GFP_KERNEL);
5382 if (!nskb)
5383 return -ENOBUFS;
5385 err = rtnl_fill_statsinfo(nskb, dev, RTM_NEWSTATS,
5386 NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
5387 0, filter_mask, &idxattr, &prividx);
5388 if (err < 0) {
5389 /* -EMSGSIZE implies BUG in if_nlmsg_stats_size */
5390 WARN_ON(err == -EMSGSIZE);
5391 kfree_skb(nskb);
5392 } else {
5393 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
5396 return err;
5399 static int rtnl_stats_dump(struct sk_buff *skb, struct netlink_callback *cb)
5401 struct netlink_ext_ack *extack = cb->extack;
5402 int h, s_h, err, s_idx, s_idxattr, s_prividx;
5403 struct net *net = sock_net(skb->sk);
5404 unsigned int flags = NLM_F_MULTI;
5405 struct if_stats_msg *ifsm;
5406 struct hlist_head *head;
5407 struct net_device *dev;
5408 u32 filter_mask = 0;
5409 int idx = 0;
5411 s_h = cb->args[0];
5412 s_idx = cb->args[1];
5413 s_idxattr = cb->args[2];
5414 s_prividx = cb->args[3];
5416 cb->seq = net->dev_base_seq;
5418 err = rtnl_valid_stats_req(cb->nlh, cb->strict_check, true, extack);
5419 if (err)
5420 return err;
5422 ifsm = nlmsg_data(cb->nlh);
5423 filter_mask = ifsm->filter_mask;
5424 if (!filter_mask) {
5425 NL_SET_ERR_MSG(extack, "Filter mask must be set for stats dump");
5426 return -EINVAL;
5429 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5430 idx = 0;
5431 head = &net->dev_index_head[h];
5432 hlist_for_each_entry(dev, head, index_hlist) {
5433 if (idx < s_idx)
5434 goto cont;
5435 err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS,
5436 NETLINK_CB(cb->skb).portid,
5437 cb->nlh->nlmsg_seq, 0,
5438 flags, filter_mask,
5439 &s_idxattr, &s_prividx);
5440 /* If we ran out of room on the first message,
5441 * we're in trouble
5443 WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
5445 if (err < 0)
5446 goto out;
5447 s_prividx = 0;
5448 s_idxattr = 0;
5449 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
5450 cont:
5451 idx++;
5454 out:
5455 cb->args[3] = s_prividx;
5456 cb->args[2] = s_idxattr;
5457 cb->args[1] = idx;
5458 cb->args[0] = h;
5460 return skb->len;
5463 /* Process one rtnetlink message. */
5465 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh,
5466 struct netlink_ext_ack *extack)
5468 struct net *net = sock_net(skb->sk);
5469 struct rtnl_link *link;
5470 struct module *owner;
5471 int err = -EOPNOTSUPP;
5472 rtnl_doit_func doit;
5473 unsigned int flags;
5474 int kind;
5475 int family;
5476 int type;
5478 type = nlh->nlmsg_type;
5479 if (type > RTM_MAX)
5480 return -EOPNOTSUPP;
5482 type -= RTM_BASE;
5484 /* All the messages must have at least 1 byte length */
5485 if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
5486 return 0;
5488 family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
5489 kind = type&3;
5491 if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
5492 return -EPERM;
5494 rcu_read_lock();
5495 if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
5496 struct sock *rtnl;
5497 rtnl_dumpit_func dumpit;
5498 u32 min_dump_alloc = 0;
5500 link = rtnl_get_link(family, type);
5501 if (!link || !link->dumpit) {
5502 family = PF_UNSPEC;
5503 link = rtnl_get_link(family, type);
5504 if (!link || !link->dumpit)
5505 goto err_unlock;
5507 owner = link->owner;
5508 dumpit = link->dumpit;
5510 if (type == RTM_GETLINK - RTM_BASE)
5511 min_dump_alloc = rtnl_calcit(skb, nlh);
5513 err = 0;
5514 /* need to do this before rcu_read_unlock() */
5515 if (!try_module_get(owner))
5516 err = -EPROTONOSUPPORT;
5518 rcu_read_unlock();
5520 rtnl = net->rtnl;
5521 if (err == 0) {
5522 struct netlink_dump_control c = {
5523 .dump = dumpit,
5524 .min_dump_alloc = min_dump_alloc,
5525 .module = owner,
5527 err = netlink_dump_start(rtnl, skb, nlh, &c);
5528 /* netlink_dump_start() will keep a reference on
5529 * module if dump is still in progress.
5531 module_put(owner);
5533 return err;
5536 link = rtnl_get_link(family, type);
5537 if (!link || !link->doit) {
5538 family = PF_UNSPEC;
5539 link = rtnl_get_link(PF_UNSPEC, type);
5540 if (!link || !link->doit)
5541 goto out_unlock;
5544 owner = link->owner;
5545 if (!try_module_get(owner)) {
5546 err = -EPROTONOSUPPORT;
5547 goto out_unlock;
5550 flags = link->flags;
5551 if (flags & RTNL_FLAG_DOIT_UNLOCKED) {
5552 doit = link->doit;
5553 rcu_read_unlock();
5554 if (doit)
5555 err = doit(skb, nlh, extack);
5556 module_put(owner);
5557 return err;
5559 rcu_read_unlock();
5561 rtnl_lock();
5562 link = rtnl_get_link(family, type);
5563 if (link && link->doit)
5564 err = link->doit(skb, nlh, extack);
5565 rtnl_unlock();
5567 module_put(owner);
5569 return err;
5571 out_unlock:
5572 rcu_read_unlock();
5573 return err;
5575 err_unlock:
5576 rcu_read_unlock();
5577 return -EOPNOTSUPP;
5580 static void rtnetlink_rcv(struct sk_buff *skb)
5582 netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
5585 static int rtnetlink_bind(struct net *net, int group)
5587 switch (group) {
5588 case RTNLGRP_IPV4_MROUTE_R:
5589 case RTNLGRP_IPV6_MROUTE_R:
5590 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
5591 return -EPERM;
5592 break;
5594 return 0;
5597 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
5599 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
5601 switch (event) {
5602 case NETDEV_REBOOT:
5603 case NETDEV_CHANGEMTU:
5604 case NETDEV_CHANGEADDR:
5605 case NETDEV_CHANGENAME:
5606 case NETDEV_FEAT_CHANGE:
5607 case NETDEV_BONDING_FAILOVER:
5608 case NETDEV_POST_TYPE_CHANGE:
5609 case NETDEV_NOTIFY_PEERS:
5610 case NETDEV_CHANGEUPPER:
5611 case NETDEV_RESEND_IGMP:
5612 case NETDEV_CHANGEINFODATA:
5613 case NETDEV_CHANGELOWERSTATE:
5614 case NETDEV_CHANGE_TX_QUEUE_LEN:
5615 rtmsg_ifinfo_event(RTM_NEWLINK, dev, 0, rtnl_get_event(event),
5616 GFP_KERNEL, NULL, 0);
5617 break;
5618 default:
5619 break;
5621 return NOTIFY_DONE;
5624 static struct notifier_block rtnetlink_dev_notifier = {
5625 .notifier_call = rtnetlink_event,
5629 static int __net_init rtnetlink_net_init(struct net *net)
5631 struct sock *sk;
5632 struct netlink_kernel_cfg cfg = {
5633 .groups = RTNLGRP_MAX,
5634 .input = rtnetlink_rcv,
5635 .cb_mutex = &rtnl_mutex,
5636 .flags = NL_CFG_F_NONROOT_RECV,
5637 .bind = rtnetlink_bind,
5640 sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
5641 if (!sk)
5642 return -ENOMEM;
5643 net->rtnl = sk;
5644 return 0;
5647 static void __net_exit rtnetlink_net_exit(struct net *net)
5649 netlink_kernel_release(net->rtnl);
5650 net->rtnl = NULL;
5653 static struct pernet_operations rtnetlink_net_ops = {
5654 .init = rtnetlink_net_init,
5655 .exit = rtnetlink_net_exit,
5658 void __init rtnetlink_init(void)
5660 if (register_pernet_subsys(&rtnetlink_net_ops))
5661 panic("rtnetlink_init: cannot initialize rtnetlink\n");
5663 register_netdevice_notifier(&rtnetlink_dev_notifier);
5665 rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
5666 rtnl_dump_ifinfo, 0);
5667 rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, 0);
5668 rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, 0);
5669 rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, 0);
5671 rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, 0);
5672 rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, 0);
5673 rtnl_register(PF_UNSPEC, RTM_GETNETCONF, NULL, rtnl_dump_all, 0);
5675 rtnl_register(PF_UNSPEC, RTM_NEWLINKPROP, rtnl_newlinkprop, NULL, 0);
5676 rtnl_register(PF_UNSPEC, RTM_DELLINKPROP, rtnl_dellinkprop, NULL, 0);
5678 rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, 0);
5679 rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, 0);
5680 rtnl_register(PF_BRIDGE, RTM_GETNEIGH, rtnl_fdb_get, rtnl_fdb_dump, 0);
5682 rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, 0);
5683 rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, 0);
5684 rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, 0);
5686 rtnl_register(PF_UNSPEC, RTM_GETSTATS, rtnl_stats_get, rtnl_stats_dump,