1 // SPDX-License-Identifier: GPL-2.0-or-later
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>
12 * Vitaly E. Lavrov RTA_OK arithmetic 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>
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>
46 #include <net/protocol.h>
48 #include <net/route.h>
52 #include <net/pkt_sched.h>
53 #include <net/fib_rules.h>
54 #include <net/rtnetlink.h>
55 #include <net/net_namespace.h>
56 #include <net/devlink.h>
60 #define RTNL_MAX_TYPE 50
61 #define RTNL_SLAVE_MAX_TYPE 40
65 rtnl_dumpit_func dumpit;
71 static DEFINE_MUTEX(rtnl_mutex);
75 mutex_lock(&rtnl_mutex);
77 EXPORT_SYMBOL(rtnl_lock);
79 int rtnl_lock_killable(void)
81 return mutex_lock_killable(&rtnl_mutex);
83 EXPORT_SYMBOL(rtnl_lock_killable);
85 static struct sk_buff *defer_kfree_skb_list;
86 void rtnl_kfree_skbs(struct sk_buff *head, struct sk_buff *tail)
89 tail->next = defer_kfree_skb_list;
90 defer_kfree_skb_list = head;
93 EXPORT_SYMBOL(rtnl_kfree_skbs);
95 void __rtnl_unlock(void)
97 struct sk_buff *head = defer_kfree_skb_list;
99 defer_kfree_skb_list = NULL;
101 /* Ensure that we didn't actually add any TODO item when __rtnl_unlock()
102 * is used. In some places, e.g. in cfg80211, we have code that will do
109 * and because netdev_run_todo() acquires the RTNL for items on the list
110 * we could cause a situation such as this:
113 * unregister_netdevice()
121 * // list not empty now
122 * // because of thread 2
124 * while (!list_empty(...))
129 * However, usage of __rtnl_unlock() is rare, and so we can ensure that
130 * it's not used in cases where something is added to do the list.
132 WARN_ON(!list_empty(&net_todo_list));
134 mutex_unlock(&rtnl_mutex);
137 struct sk_buff *next = head->next;
145 void rtnl_unlock(void)
147 /* This fellow will unlock it for us. */
150 EXPORT_SYMBOL(rtnl_unlock);
152 int rtnl_trylock(void)
154 return mutex_trylock(&rtnl_mutex);
156 EXPORT_SYMBOL(rtnl_trylock);
158 int rtnl_is_locked(void)
160 return mutex_is_locked(&rtnl_mutex);
162 EXPORT_SYMBOL(rtnl_is_locked);
164 bool refcount_dec_and_rtnl_lock(refcount_t *r)
166 return refcount_dec_and_mutex_lock(r, &rtnl_mutex);
168 EXPORT_SYMBOL(refcount_dec_and_rtnl_lock);
170 #ifdef CONFIG_PROVE_LOCKING
171 bool lockdep_rtnl_is_held(void)
173 return lockdep_is_held(&rtnl_mutex);
175 EXPORT_SYMBOL(lockdep_rtnl_is_held);
176 #endif /* #ifdef CONFIG_PROVE_LOCKING */
178 static struct rtnl_link __rcu *__rcu *rtnl_msg_handlers[RTNL_FAMILY_MAX + 1];
180 static inline int rtm_msgindex(int msgtype)
182 int msgindex = msgtype - RTM_BASE;
185 * msgindex < 0 implies someone tried to register a netlink
186 * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
187 * the message type has not been added to linux/rtnetlink.h
189 BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
194 static struct rtnl_link *rtnl_get_link(int protocol, int msgtype)
196 struct rtnl_link __rcu **tab;
198 if (protocol >= ARRAY_SIZE(rtnl_msg_handlers))
199 protocol = PF_UNSPEC;
201 tab = rcu_dereference_rtnl(rtnl_msg_handlers[protocol]);
203 tab = rcu_dereference_rtnl(rtnl_msg_handlers[PF_UNSPEC]);
205 return rcu_dereference_rtnl(tab[msgtype]);
208 static int rtnl_register_internal(struct module *owner,
209 int protocol, int msgtype,
210 rtnl_doit_func doit, rtnl_dumpit_func dumpit,
213 struct rtnl_link *link, *old;
214 struct rtnl_link __rcu **tab;
218 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
219 msgindex = rtm_msgindex(msgtype);
222 tab = rtnl_dereference(rtnl_msg_handlers[protocol]);
224 tab = kcalloc(RTM_NR_MSGTYPES, sizeof(void *), GFP_KERNEL);
228 /* ensures we see the 0 stores */
229 rcu_assign_pointer(rtnl_msg_handlers[protocol], tab);
232 old = rtnl_dereference(tab[msgindex]);
234 link = kmemdup(old, sizeof(*old), GFP_KERNEL);
238 link = kzalloc(sizeof(*link), GFP_KERNEL);
243 WARN_ON(link->owner && link->owner != owner);
246 WARN_ON(doit && link->doit && link->doit != doit);
249 WARN_ON(dumpit && link->dumpit && link->dumpit != dumpit);
251 link->dumpit = dumpit;
253 WARN_ON(rtnl_msgtype_kind(msgtype) != RTNL_KIND_DEL &&
254 (flags & RTNL_FLAG_BULK_DEL_SUPPORTED));
255 link->flags |= flags;
257 /* publish protocol:msgtype */
258 rcu_assign_pointer(tab[msgindex], link);
268 * rtnl_register_module - Register a rtnetlink message type
270 * @owner: module registering the hook (THIS_MODULE)
271 * @protocol: Protocol family or PF_UNSPEC
272 * @msgtype: rtnetlink message type
273 * @doit: Function pointer called for each request message
274 * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
275 * @flags: rtnl_link_flags to modify behaviour of doit/dumpit functions
277 * Like rtnl_register, but for use by removable modules.
279 int rtnl_register_module(struct module *owner,
280 int protocol, int msgtype,
281 rtnl_doit_func doit, rtnl_dumpit_func dumpit,
284 return rtnl_register_internal(owner, protocol, msgtype,
285 doit, dumpit, flags);
287 EXPORT_SYMBOL_GPL(rtnl_register_module);
290 * rtnl_register - Register a rtnetlink message type
291 * @protocol: Protocol family or PF_UNSPEC
292 * @msgtype: rtnetlink message type
293 * @doit: Function pointer called for each request message
294 * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
295 * @flags: rtnl_link_flags to modify behaviour of doit/dumpit functions
297 * Registers the specified function pointers (at least one of them has
298 * to be non-NULL) to be called whenever a request message for the
299 * specified protocol family and message type is received.
301 * The special protocol family PF_UNSPEC may be used to define fallback
302 * function pointers for the case when no entry for the specific protocol
305 void rtnl_register(int protocol, int msgtype,
306 rtnl_doit_func doit, rtnl_dumpit_func dumpit,
311 err = rtnl_register_internal(NULL, protocol, msgtype, doit, dumpit,
314 pr_err("Unable to register rtnetlink message handler, "
315 "protocol = %d, message type = %d\n", protocol, msgtype);
319 * rtnl_unregister - Unregister a rtnetlink message type
320 * @protocol: Protocol family or PF_UNSPEC
321 * @msgtype: rtnetlink message type
323 * Returns 0 on success or a negative error code.
325 int rtnl_unregister(int protocol, int msgtype)
327 struct rtnl_link __rcu **tab;
328 struct rtnl_link *link;
331 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
332 msgindex = rtm_msgindex(msgtype);
335 tab = rtnl_dereference(rtnl_msg_handlers[protocol]);
341 link = rtnl_dereference(tab[msgindex]);
342 RCU_INIT_POINTER(tab[msgindex], NULL);
345 kfree_rcu(link, rcu);
349 EXPORT_SYMBOL_GPL(rtnl_unregister);
352 * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
353 * @protocol : Protocol family or PF_UNSPEC
355 * Identical to calling rtnl_unregster() for all registered message types
356 * of a certain protocol family.
358 void rtnl_unregister_all(int protocol)
360 struct rtnl_link __rcu **tab;
361 struct rtnl_link *link;
364 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
367 tab = rtnl_dereference(rtnl_msg_handlers[protocol]);
372 RCU_INIT_POINTER(rtnl_msg_handlers[protocol], NULL);
373 for (msgindex = 0; msgindex < RTM_NR_MSGTYPES; msgindex++) {
374 link = rtnl_dereference(tab[msgindex]);
378 RCU_INIT_POINTER(tab[msgindex], NULL);
379 kfree_rcu(link, rcu);
387 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
389 static LIST_HEAD(link_ops);
391 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
393 const struct rtnl_link_ops *ops;
395 list_for_each_entry(ops, &link_ops, list) {
396 if (!strcmp(ops->kind, kind))
403 * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
404 * @ops: struct rtnl_link_ops * to register
406 * The caller must hold the rtnl_mutex. This function should be used
407 * by drivers that create devices during module initialization. It
408 * must be called before registering the devices.
410 * Returns 0 on success or a negative error code.
412 int __rtnl_link_register(struct rtnl_link_ops *ops)
414 if (rtnl_link_ops_get(ops->kind))
417 /* The check for alloc/setup is here because if ops
418 * does not have that filled up, it is not possible
419 * to use the ops for creating device. So do not
420 * fill up dellink as well. That disables rtnl_dellink.
422 if ((ops->alloc || ops->setup) && !ops->dellink)
423 ops->dellink = unregister_netdevice_queue;
425 list_add_tail(&ops->list, &link_ops);
428 EXPORT_SYMBOL_GPL(__rtnl_link_register);
431 * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
432 * @ops: struct rtnl_link_ops * to register
434 * Returns 0 on success or a negative error code.
436 int rtnl_link_register(struct rtnl_link_ops *ops)
440 /* Sanity-check max sizes to avoid stack buffer overflow. */
441 if (WARN_ON(ops->maxtype > RTNL_MAX_TYPE ||
442 ops->slave_maxtype > RTNL_SLAVE_MAX_TYPE))
446 err = __rtnl_link_register(ops);
450 EXPORT_SYMBOL_GPL(rtnl_link_register);
452 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
454 struct net_device *dev;
455 LIST_HEAD(list_kill);
457 for_each_netdev(net, dev) {
458 if (dev->rtnl_link_ops == ops)
459 ops->dellink(dev, &list_kill);
461 unregister_netdevice_many(&list_kill);
465 * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
466 * @ops: struct rtnl_link_ops * to unregister
468 * The caller must hold the rtnl_mutex and guarantee net_namespace_list
469 * integrity (hold pernet_ops_rwsem for writing to close the race
470 * with setup_net() and cleanup_net()).
472 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
477 __rtnl_kill_links(net, ops);
479 list_del(&ops->list);
481 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
483 /* Return with the rtnl_lock held when there are no network
484 * devices unregistering in any network namespace.
486 static void rtnl_lock_unregistering_all(void)
490 DEFINE_WAIT_FUNC(wait, woken_wake_function);
492 add_wait_queue(&netdev_unregistering_wq, &wait);
494 unregistering = false;
496 /* We held write locked pernet_ops_rwsem, and parallel
497 * setup_net() and cleanup_net() are not possible.
500 if (atomic_read(&net->dev_unreg_count) > 0) {
501 unregistering = true;
509 wait_woken(&wait, TASK_UNINTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
511 remove_wait_queue(&netdev_unregistering_wq, &wait);
515 * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
516 * @ops: struct rtnl_link_ops * to unregister
518 void rtnl_link_unregister(struct rtnl_link_ops *ops)
520 /* Close the race with setup_net() and cleanup_net() */
521 down_write(&pernet_ops_rwsem);
522 rtnl_lock_unregistering_all();
523 __rtnl_link_unregister(ops);
525 up_write(&pernet_ops_rwsem);
527 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
529 static size_t rtnl_link_get_slave_info_data_size(const struct net_device *dev)
531 struct net_device *master_dev;
532 const struct rtnl_link_ops *ops;
537 master_dev = netdev_master_upper_dev_get_rcu((struct net_device *)dev);
541 ops = master_dev->rtnl_link_ops;
542 if (!ops || !ops->get_slave_size)
544 /* IFLA_INFO_SLAVE_DATA + nested data */
545 size = nla_total_size(sizeof(struct nlattr)) +
546 ops->get_slave_size(master_dev, dev);
553 static size_t rtnl_link_get_size(const struct net_device *dev)
555 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
561 size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
562 nla_total_size(strlen(ops->kind) + 1); /* IFLA_INFO_KIND */
565 /* IFLA_INFO_DATA + nested data */
566 size += nla_total_size(sizeof(struct nlattr)) +
569 if (ops->get_xstats_size)
570 /* IFLA_INFO_XSTATS */
571 size += nla_total_size(ops->get_xstats_size(dev));
573 size += rtnl_link_get_slave_info_data_size(dev);
578 static LIST_HEAD(rtnl_af_ops);
580 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
582 const struct rtnl_af_ops *ops;
586 list_for_each_entry(ops, &rtnl_af_ops, list) {
587 if (ops->family == family)
595 * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
596 * @ops: struct rtnl_af_ops * to register
598 * Returns 0 on success or a negative error code.
600 void rtnl_af_register(struct rtnl_af_ops *ops)
603 list_add_tail_rcu(&ops->list, &rtnl_af_ops);
606 EXPORT_SYMBOL_GPL(rtnl_af_register);
609 * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
610 * @ops: struct rtnl_af_ops * to unregister
612 void rtnl_af_unregister(struct rtnl_af_ops *ops)
615 list_del_rcu(&ops->list);
620 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
622 static size_t rtnl_link_get_af_size(const struct net_device *dev,
625 struct rtnl_af_ops *af_ops;
629 size = nla_total_size(sizeof(struct nlattr));
632 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
633 if (af_ops->get_link_af_size) {
634 /* AF_* + nested data */
635 size += nla_total_size(sizeof(struct nlattr)) +
636 af_ops->get_link_af_size(dev, ext_filter_mask);
644 static bool rtnl_have_link_slave_info(const struct net_device *dev)
646 struct net_device *master_dev;
651 master_dev = netdev_master_upper_dev_get_rcu((struct net_device *)dev);
652 if (master_dev && master_dev->rtnl_link_ops)
658 static int rtnl_link_slave_info_fill(struct sk_buff *skb,
659 const struct net_device *dev)
661 struct net_device *master_dev;
662 const struct rtnl_link_ops *ops;
663 struct nlattr *slave_data;
666 master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
669 ops = master_dev->rtnl_link_ops;
672 if (nla_put_string(skb, IFLA_INFO_SLAVE_KIND, ops->kind) < 0)
674 if (ops->fill_slave_info) {
675 slave_data = nla_nest_start_noflag(skb, IFLA_INFO_SLAVE_DATA);
678 err = ops->fill_slave_info(skb, master_dev, dev);
680 goto err_cancel_slave_data;
681 nla_nest_end(skb, slave_data);
685 err_cancel_slave_data:
686 nla_nest_cancel(skb, slave_data);
690 static int rtnl_link_info_fill(struct sk_buff *skb,
691 const struct net_device *dev)
693 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
699 if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
701 if (ops->fill_xstats) {
702 err = ops->fill_xstats(skb, dev);
706 if (ops->fill_info) {
707 data = nla_nest_start_noflag(skb, IFLA_INFO_DATA);
710 err = ops->fill_info(skb, dev);
712 goto err_cancel_data;
713 nla_nest_end(skb, data);
718 nla_nest_cancel(skb, data);
722 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
724 struct nlattr *linkinfo;
727 linkinfo = nla_nest_start_noflag(skb, IFLA_LINKINFO);
728 if (linkinfo == NULL)
731 err = rtnl_link_info_fill(skb, dev);
733 goto err_cancel_link;
735 err = rtnl_link_slave_info_fill(skb, dev);
737 goto err_cancel_link;
739 nla_nest_end(skb, linkinfo);
743 nla_nest_cancel(skb, linkinfo);
748 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
750 struct sock *rtnl = net->rtnl;
752 return nlmsg_notify(rtnl, skb, pid, group, echo, GFP_KERNEL);
755 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
757 struct sock *rtnl = net->rtnl;
759 return nlmsg_unicast(rtnl, skb, pid);
761 EXPORT_SYMBOL(rtnl_unicast);
763 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
764 const struct nlmsghdr *nlh, gfp_t flags)
766 struct sock *rtnl = net->rtnl;
768 nlmsg_notify(rtnl, skb, pid, group, nlmsg_report(nlh), flags);
770 EXPORT_SYMBOL(rtnl_notify);
772 void rtnl_set_sk_err(struct net *net, u32 group, int error)
774 struct sock *rtnl = net->rtnl;
776 netlink_set_err(rtnl, 0, group, error);
778 EXPORT_SYMBOL(rtnl_set_sk_err);
780 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
785 /* nothing is dumped for dst_default_metrics, so just skip the loop */
786 if (metrics == dst_default_metrics.metrics)
789 mx = nla_nest_start_noflag(skb, RTA_METRICS);
793 for (i = 0; i < RTAX_MAX; i++) {
795 if (i == RTAX_CC_ALGO - 1) {
796 char tmp[TCP_CA_NAME_MAX], *name;
798 name = tcp_ca_get_name_by_key(metrics[i], tmp);
801 if (nla_put_string(skb, i + 1, name))
802 goto nla_put_failure;
803 } else if (i == RTAX_FEATURES - 1) {
804 u32 user_features = metrics[i] & RTAX_FEATURE_MASK;
808 BUILD_BUG_ON(RTAX_FEATURE_MASK & DST_FEATURE_MASK);
809 if (nla_put_u32(skb, i + 1, user_features))
810 goto nla_put_failure;
812 if (nla_put_u32(skb, i + 1, metrics[i]))
813 goto nla_put_failure;
820 nla_nest_cancel(skb, mx);
824 return nla_nest_end(skb, mx);
827 nla_nest_cancel(skb, mx);
830 EXPORT_SYMBOL(rtnetlink_put_metrics);
832 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
833 long expires, u32 error)
835 struct rta_cacheinfo ci = {
841 ci.rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse);
842 ci.rta_used = dst->__use;
843 ci.rta_clntref = atomic_read(&dst->__refcnt);
848 clock = jiffies_to_clock_t(abs(expires));
849 clock = min_t(unsigned long, clock, INT_MAX);
850 ci.rta_expires = (expires > 0) ? clock : -clock;
852 return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
854 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
856 static void set_operstate(struct net_device *dev, unsigned char transition)
858 unsigned char operstate = dev->operstate;
860 switch (transition) {
862 if ((operstate == IF_OPER_DORMANT ||
863 operstate == IF_OPER_TESTING ||
864 operstate == IF_OPER_UNKNOWN) &&
865 !netif_dormant(dev) && !netif_testing(dev))
866 operstate = IF_OPER_UP;
869 case IF_OPER_TESTING:
870 if (netif_oper_up(dev))
871 operstate = IF_OPER_TESTING;
874 case IF_OPER_DORMANT:
875 if (netif_oper_up(dev))
876 operstate = IF_OPER_DORMANT;
880 if (dev->operstate != operstate) {
881 write_lock(&dev_base_lock);
882 dev->operstate = operstate;
883 write_unlock(&dev_base_lock);
884 netdev_state_change(dev);
888 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
890 return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
891 (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
894 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
895 const struct ifinfomsg *ifm)
897 unsigned int flags = ifm->ifi_flags;
899 /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
901 flags = (flags & ifm->ifi_change) |
902 (rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
907 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
908 const struct rtnl_link_stats64 *b)
910 a->rx_packets = b->rx_packets;
911 a->tx_packets = b->tx_packets;
912 a->rx_bytes = b->rx_bytes;
913 a->tx_bytes = b->tx_bytes;
914 a->rx_errors = b->rx_errors;
915 a->tx_errors = b->tx_errors;
916 a->rx_dropped = b->rx_dropped;
917 a->tx_dropped = b->tx_dropped;
919 a->multicast = b->multicast;
920 a->collisions = b->collisions;
922 a->rx_length_errors = b->rx_length_errors;
923 a->rx_over_errors = b->rx_over_errors;
924 a->rx_crc_errors = b->rx_crc_errors;
925 a->rx_frame_errors = b->rx_frame_errors;
926 a->rx_fifo_errors = b->rx_fifo_errors;
927 a->rx_missed_errors = b->rx_missed_errors;
929 a->tx_aborted_errors = b->tx_aborted_errors;
930 a->tx_carrier_errors = b->tx_carrier_errors;
931 a->tx_fifo_errors = b->tx_fifo_errors;
932 a->tx_heartbeat_errors = b->tx_heartbeat_errors;
933 a->tx_window_errors = b->tx_window_errors;
935 a->rx_compressed = b->rx_compressed;
936 a->tx_compressed = b->tx_compressed;
938 a->rx_nohandler = b->rx_nohandler;
942 static inline int rtnl_vfinfo_size(const struct net_device *dev,
945 if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF)) {
946 int num_vfs = dev_num_vf(dev->dev.parent);
947 size_t size = nla_total_size(0);
950 nla_total_size(sizeof(struct ifla_vf_mac)) +
951 nla_total_size(sizeof(struct ifla_vf_broadcast)) +
952 nla_total_size(sizeof(struct ifla_vf_vlan)) +
953 nla_total_size(0) + /* nest IFLA_VF_VLAN_LIST */
954 nla_total_size(MAX_VLAN_LIST_LEN *
955 sizeof(struct ifla_vf_vlan_info)) +
956 nla_total_size(sizeof(struct ifla_vf_spoofchk)) +
957 nla_total_size(sizeof(struct ifla_vf_tx_rate)) +
958 nla_total_size(sizeof(struct ifla_vf_rate)) +
959 nla_total_size(sizeof(struct ifla_vf_link_state)) +
960 nla_total_size(sizeof(struct ifla_vf_rss_query_en)) +
961 nla_total_size(0) + /* nest IFLA_VF_STATS */
962 /* IFLA_VF_STATS_RX_PACKETS */
963 nla_total_size_64bit(sizeof(__u64)) +
964 /* IFLA_VF_STATS_TX_PACKETS */
965 nla_total_size_64bit(sizeof(__u64)) +
966 /* IFLA_VF_STATS_RX_BYTES */
967 nla_total_size_64bit(sizeof(__u64)) +
968 /* IFLA_VF_STATS_TX_BYTES */
969 nla_total_size_64bit(sizeof(__u64)) +
970 /* IFLA_VF_STATS_BROADCAST */
971 nla_total_size_64bit(sizeof(__u64)) +
972 /* IFLA_VF_STATS_MULTICAST */
973 nla_total_size_64bit(sizeof(__u64)) +
974 /* IFLA_VF_STATS_RX_DROPPED */
975 nla_total_size_64bit(sizeof(__u64)) +
976 /* IFLA_VF_STATS_TX_DROPPED */
977 nla_total_size_64bit(sizeof(__u64)) +
978 nla_total_size(sizeof(struct ifla_vf_trust)));
984 static size_t rtnl_port_size(const struct net_device *dev,
987 size_t port_size = nla_total_size(4) /* PORT_VF */
988 + nla_total_size(PORT_PROFILE_MAX) /* PORT_PROFILE */
989 + nla_total_size(PORT_UUID_MAX) /* PORT_INSTANCE_UUID */
990 + nla_total_size(PORT_UUID_MAX) /* PORT_HOST_UUID */
991 + nla_total_size(1) /* PROT_VDP_REQUEST */
992 + nla_total_size(2); /* PORT_VDP_RESPONSE */
993 size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
994 size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
996 size_t port_self_size = nla_total_size(sizeof(struct nlattr))
999 if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
1000 !(ext_filter_mask & RTEXT_FILTER_VF))
1002 if (dev_num_vf(dev->dev.parent))
1003 return port_self_size + vf_ports_size +
1004 vf_port_size * dev_num_vf(dev->dev.parent);
1006 return port_self_size;
1009 static size_t rtnl_xdp_size(void)
1011 size_t xdp_size = nla_total_size(0) + /* nest IFLA_XDP */
1012 nla_total_size(1) + /* XDP_ATTACHED */
1013 nla_total_size(4) + /* XDP_PROG_ID (or 1st mode) */
1014 nla_total_size(4); /* XDP_<mode>_PROG_ID */
1019 static size_t rtnl_prop_list_size(const struct net_device *dev)
1021 struct netdev_name_node *name_node;
1024 if (list_empty(&dev->name_node->list))
1026 size = nla_total_size(0);
1027 list_for_each_entry(name_node, &dev->name_node->list, list)
1028 size += nla_total_size(ALTIFNAMSIZ);
1032 static size_t rtnl_proto_down_size(const struct net_device *dev)
1034 size_t size = nla_total_size(1);
1036 if (dev->proto_down_reason)
1037 size += nla_total_size(0) + nla_total_size(4);
1042 static size_t rtnl_devlink_port_size(const struct net_device *dev)
1044 size_t size = nla_total_size(0); /* nest IFLA_DEVLINK_PORT */
1046 if (dev->devlink_port)
1047 size += devlink_nl_port_handle_size(dev->devlink_port);
1052 static noinline size_t if_nlmsg_size(const struct net_device *dev,
1053 u32 ext_filter_mask)
1055 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
1056 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
1057 + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
1058 + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
1059 + nla_total_size_64bit(sizeof(struct rtnl_link_ifmap))
1060 + nla_total_size(sizeof(struct rtnl_link_stats))
1061 + nla_total_size_64bit(sizeof(struct rtnl_link_stats64))
1062 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
1063 + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
1064 + nla_total_size(4) /* IFLA_TXQLEN */
1065 + nla_total_size(4) /* IFLA_WEIGHT */
1066 + nla_total_size(4) /* IFLA_MTU */
1067 + nla_total_size(4) /* IFLA_LINK */
1068 + nla_total_size(4) /* IFLA_MASTER */
1069 + nla_total_size(1) /* IFLA_CARRIER */
1070 + nla_total_size(4) /* IFLA_PROMISCUITY */
1071 + nla_total_size(4) /* IFLA_ALLMULTI */
1072 + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
1073 + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
1074 + nla_total_size(4) /* IFLA_GSO_MAX_SEGS */
1075 + nla_total_size(4) /* IFLA_GSO_MAX_SIZE */
1076 + nla_total_size(4) /* IFLA_GRO_MAX_SIZE */
1077 + nla_total_size(4) /* IFLA_TSO_MAX_SIZE */
1078 + nla_total_size(4) /* IFLA_TSO_MAX_SEGS */
1079 + nla_total_size(1) /* IFLA_OPERSTATE */
1080 + nla_total_size(1) /* IFLA_LINKMODE */
1081 + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
1082 + nla_total_size(4) /* IFLA_LINK_NETNSID */
1083 + nla_total_size(4) /* IFLA_GROUP */
1084 + nla_total_size(ext_filter_mask
1085 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
1086 + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
1087 + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
1088 + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
1089 + rtnl_link_get_af_size(dev, ext_filter_mask) /* IFLA_AF_SPEC */
1090 + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */
1091 + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_SWITCH_ID */
1092 + nla_total_size(IFNAMSIZ) /* IFLA_PHYS_PORT_NAME */
1093 + rtnl_xdp_size() /* IFLA_XDP */
1094 + nla_total_size(4) /* IFLA_EVENT */
1095 + nla_total_size(4) /* IFLA_NEW_NETNSID */
1096 + nla_total_size(4) /* IFLA_NEW_IFINDEX */
1097 + rtnl_proto_down_size(dev) /* proto down */
1098 + nla_total_size(4) /* IFLA_TARGET_NETNSID */
1099 + nla_total_size(4) /* IFLA_CARRIER_UP_COUNT */
1100 + nla_total_size(4) /* IFLA_CARRIER_DOWN_COUNT */
1101 + nla_total_size(4) /* IFLA_MIN_MTU */
1102 + nla_total_size(4) /* IFLA_MAX_MTU */
1103 + rtnl_prop_list_size(dev)
1104 + nla_total_size(MAX_ADDR_LEN) /* IFLA_PERM_ADDRESS */
1105 + rtnl_devlink_port_size(dev)
1109 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
1111 struct nlattr *vf_ports;
1112 struct nlattr *vf_port;
1116 vf_ports = nla_nest_start_noflag(skb, IFLA_VF_PORTS);
1120 for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
1121 vf_port = nla_nest_start_noflag(skb, IFLA_VF_PORT);
1123 goto nla_put_failure;
1124 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
1125 goto nla_put_failure;
1126 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
1127 if (err == -EMSGSIZE)
1128 goto nla_put_failure;
1130 nla_nest_cancel(skb, vf_port);
1133 nla_nest_end(skb, vf_port);
1136 nla_nest_end(skb, vf_ports);
1141 nla_nest_cancel(skb, vf_ports);
1145 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
1147 struct nlattr *port_self;
1150 port_self = nla_nest_start_noflag(skb, IFLA_PORT_SELF);
1154 err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
1156 nla_nest_cancel(skb, port_self);
1157 return (err == -EMSGSIZE) ? err : 0;
1160 nla_nest_end(skb, port_self);
1165 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
1166 u32 ext_filter_mask)
1170 if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
1171 !(ext_filter_mask & RTEXT_FILTER_VF))
1174 err = rtnl_port_self_fill(skb, dev);
1178 if (dev_num_vf(dev->dev.parent)) {
1179 err = rtnl_vf_ports_fill(skb, dev);
1187 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
1190 struct netdev_phys_item_id ppid;
1192 err = dev_get_phys_port_id(dev, &ppid);
1194 if (err == -EOPNOTSUPP)
1199 if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
1205 static int rtnl_phys_port_name_fill(struct sk_buff *skb, struct net_device *dev)
1207 char name[IFNAMSIZ];
1210 err = dev_get_phys_port_name(dev, name, sizeof(name));
1212 if (err == -EOPNOTSUPP)
1217 if (nla_put_string(skb, IFLA_PHYS_PORT_NAME, name))
1223 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev)
1225 struct netdev_phys_item_id ppid = { };
1228 err = dev_get_port_parent_id(dev, &ppid, false);
1230 if (err == -EOPNOTSUPP)
1235 if (nla_put(skb, IFLA_PHYS_SWITCH_ID, ppid.id_len, ppid.id))
1241 static noinline_for_stack int rtnl_fill_stats(struct sk_buff *skb,
1242 struct net_device *dev)
1244 struct rtnl_link_stats64 *sp;
1245 struct nlattr *attr;
1247 attr = nla_reserve_64bit(skb, IFLA_STATS64,
1248 sizeof(struct rtnl_link_stats64), IFLA_PAD);
1252 sp = nla_data(attr);
1253 dev_get_stats(dev, sp);
1255 attr = nla_reserve(skb, IFLA_STATS,
1256 sizeof(struct rtnl_link_stats));
1260 copy_rtnl_link_stats(nla_data(attr), sp);
1265 static noinline_for_stack int rtnl_fill_vfinfo(struct sk_buff *skb,
1266 struct net_device *dev,
1268 struct nlattr *vfinfo)
1270 struct ifla_vf_rss_query_en vf_rss_query_en;
1271 struct nlattr *vf, *vfstats, *vfvlanlist;
1272 struct ifla_vf_link_state vf_linkstate;
1273 struct ifla_vf_vlan_info vf_vlan_info;
1274 struct ifla_vf_spoofchk vf_spoofchk;
1275 struct ifla_vf_tx_rate vf_tx_rate;
1276 struct ifla_vf_stats vf_stats;
1277 struct ifla_vf_trust vf_trust;
1278 struct ifla_vf_vlan vf_vlan;
1279 struct ifla_vf_rate vf_rate;
1280 struct ifla_vf_mac vf_mac;
1281 struct ifla_vf_broadcast vf_broadcast;
1282 struct ifla_vf_info ivi;
1283 struct ifla_vf_guid node_guid;
1284 struct ifla_vf_guid port_guid;
1286 memset(&ivi, 0, sizeof(ivi));
1288 /* Not all SR-IOV capable drivers support the
1289 * spoofcheck and "RSS query enable" query. Preset to
1290 * -1 so the user space tool can detect that the driver
1291 * didn't report anything.
1294 ivi.rss_query_en = -1;
1296 /* The default value for VF link state is "auto"
1297 * IFLA_VF_LINK_STATE_AUTO which equals zero
1300 /* VLAN Protocol by default is 802.1Q */
1301 ivi.vlan_proto = htons(ETH_P_8021Q);
1302 if (dev->netdev_ops->ndo_get_vf_config(dev, vfs_num, &ivi))
1305 memset(&vf_vlan_info, 0, sizeof(vf_vlan_info));
1306 memset(&node_guid, 0, sizeof(node_guid));
1307 memset(&port_guid, 0, sizeof(port_guid));
1316 vf_rss_query_en.vf =
1319 port_guid.vf = ivi.vf;
1321 memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1322 memcpy(vf_broadcast.broadcast, dev->broadcast, dev->addr_len);
1323 vf_vlan.vlan = ivi.vlan;
1324 vf_vlan.qos = ivi.qos;
1325 vf_vlan_info.vlan = ivi.vlan;
1326 vf_vlan_info.qos = ivi.qos;
1327 vf_vlan_info.vlan_proto = ivi.vlan_proto;
1328 vf_tx_rate.rate = ivi.max_tx_rate;
1329 vf_rate.min_tx_rate = ivi.min_tx_rate;
1330 vf_rate.max_tx_rate = ivi.max_tx_rate;
1331 vf_spoofchk.setting = ivi.spoofchk;
1332 vf_linkstate.link_state = ivi.linkstate;
1333 vf_rss_query_en.setting = ivi.rss_query_en;
1334 vf_trust.setting = ivi.trusted;
1335 vf = nla_nest_start_noflag(skb, IFLA_VF_INFO);
1337 goto nla_put_vfinfo_failure;
1338 if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1339 nla_put(skb, IFLA_VF_BROADCAST, sizeof(vf_broadcast), &vf_broadcast) ||
1340 nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1341 nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1343 nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1345 nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1347 nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1349 nla_put(skb, IFLA_VF_RSS_QUERY_EN,
1350 sizeof(vf_rss_query_en),
1351 &vf_rss_query_en) ||
1352 nla_put(skb, IFLA_VF_TRUST,
1353 sizeof(vf_trust), &vf_trust))
1354 goto nla_put_vf_failure;
1356 if (dev->netdev_ops->ndo_get_vf_guid &&
1357 !dev->netdev_ops->ndo_get_vf_guid(dev, vfs_num, &node_guid,
1359 if (nla_put(skb, IFLA_VF_IB_NODE_GUID, sizeof(node_guid),
1361 nla_put(skb, IFLA_VF_IB_PORT_GUID, sizeof(port_guid),
1363 goto nla_put_vf_failure;
1365 vfvlanlist = nla_nest_start_noflag(skb, IFLA_VF_VLAN_LIST);
1367 goto nla_put_vf_failure;
1368 if (nla_put(skb, IFLA_VF_VLAN_INFO, sizeof(vf_vlan_info),
1370 nla_nest_cancel(skb, vfvlanlist);
1371 goto nla_put_vf_failure;
1373 nla_nest_end(skb, vfvlanlist);
1374 memset(&vf_stats, 0, sizeof(vf_stats));
1375 if (dev->netdev_ops->ndo_get_vf_stats)
1376 dev->netdev_ops->ndo_get_vf_stats(dev, vfs_num,
1378 vfstats = nla_nest_start_noflag(skb, IFLA_VF_STATS);
1380 goto nla_put_vf_failure;
1381 if (nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_PACKETS,
1382 vf_stats.rx_packets, IFLA_VF_STATS_PAD) ||
1383 nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_PACKETS,
1384 vf_stats.tx_packets, IFLA_VF_STATS_PAD) ||
1385 nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_BYTES,
1386 vf_stats.rx_bytes, IFLA_VF_STATS_PAD) ||
1387 nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_BYTES,
1388 vf_stats.tx_bytes, IFLA_VF_STATS_PAD) ||
1389 nla_put_u64_64bit(skb, IFLA_VF_STATS_BROADCAST,
1390 vf_stats.broadcast, IFLA_VF_STATS_PAD) ||
1391 nla_put_u64_64bit(skb, IFLA_VF_STATS_MULTICAST,
1392 vf_stats.multicast, IFLA_VF_STATS_PAD) ||
1393 nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_DROPPED,
1394 vf_stats.rx_dropped, IFLA_VF_STATS_PAD) ||
1395 nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_DROPPED,
1396 vf_stats.tx_dropped, IFLA_VF_STATS_PAD)) {
1397 nla_nest_cancel(skb, vfstats);
1398 goto nla_put_vf_failure;
1400 nla_nest_end(skb, vfstats);
1401 nla_nest_end(skb, vf);
1405 nla_nest_cancel(skb, vf);
1406 nla_put_vfinfo_failure:
1407 nla_nest_cancel(skb, vfinfo);
1411 static noinline_for_stack int rtnl_fill_vf(struct sk_buff *skb,
1412 struct net_device *dev,
1413 u32 ext_filter_mask)
1415 struct nlattr *vfinfo;
1418 if (!dev->dev.parent || ((ext_filter_mask & RTEXT_FILTER_VF) == 0))
1421 num_vfs = dev_num_vf(dev->dev.parent);
1422 if (nla_put_u32(skb, IFLA_NUM_VF, num_vfs))
1425 if (!dev->netdev_ops->ndo_get_vf_config)
1428 vfinfo = nla_nest_start_noflag(skb, IFLA_VFINFO_LIST);
1432 for (i = 0; i < num_vfs; i++) {
1433 if (rtnl_fill_vfinfo(skb, dev, i, vfinfo))
1437 nla_nest_end(skb, vfinfo);
1441 static int rtnl_fill_link_ifmap(struct sk_buff *skb, struct net_device *dev)
1443 struct rtnl_link_ifmap map;
1445 memset(&map, 0, sizeof(map));
1446 map.mem_start = dev->mem_start;
1447 map.mem_end = dev->mem_end;
1448 map.base_addr = dev->base_addr;
1451 map.port = dev->if_port;
1453 if (nla_put_64bit(skb, IFLA_MAP, sizeof(map), &map, IFLA_PAD))
1459 static u32 rtnl_xdp_prog_skb(struct net_device *dev)
1461 const struct bpf_prog *generic_xdp_prog;
1465 generic_xdp_prog = rtnl_dereference(dev->xdp_prog);
1466 if (!generic_xdp_prog)
1468 return generic_xdp_prog->aux->id;
1471 static u32 rtnl_xdp_prog_drv(struct net_device *dev)
1473 return dev_xdp_prog_id(dev, XDP_MODE_DRV);
1476 static u32 rtnl_xdp_prog_hw(struct net_device *dev)
1478 return dev_xdp_prog_id(dev, XDP_MODE_HW);
1481 static int rtnl_xdp_report_one(struct sk_buff *skb, struct net_device *dev,
1482 u32 *prog_id, u8 *mode, u8 tgt_mode, u32 attr,
1483 u32 (*get_prog_id)(struct net_device *dev))
1488 curr_id = get_prog_id(dev);
1493 err = nla_put_u32(skb, attr, curr_id);
1497 if (*mode != XDP_ATTACHED_NONE)
1498 *mode = XDP_ATTACHED_MULTI;
1505 static int rtnl_xdp_fill(struct sk_buff *skb, struct net_device *dev)
1512 xdp = nla_nest_start_noflag(skb, IFLA_XDP);
1517 mode = XDP_ATTACHED_NONE;
1518 err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_SKB,
1519 IFLA_XDP_SKB_PROG_ID, rtnl_xdp_prog_skb);
1522 err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_DRV,
1523 IFLA_XDP_DRV_PROG_ID, rtnl_xdp_prog_drv);
1526 err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_HW,
1527 IFLA_XDP_HW_PROG_ID, rtnl_xdp_prog_hw);
1531 err = nla_put_u8(skb, IFLA_XDP_ATTACHED, mode);
1535 if (prog_id && mode != XDP_ATTACHED_MULTI) {
1536 err = nla_put_u32(skb, IFLA_XDP_PROG_ID, prog_id);
1541 nla_nest_end(skb, xdp);
1545 nla_nest_cancel(skb, xdp);
1549 static u32 rtnl_get_event(unsigned long event)
1551 u32 rtnl_event_type = IFLA_EVENT_NONE;
1555 rtnl_event_type = IFLA_EVENT_REBOOT;
1557 case NETDEV_FEAT_CHANGE:
1558 rtnl_event_type = IFLA_EVENT_FEATURES;
1560 case NETDEV_BONDING_FAILOVER:
1561 rtnl_event_type = IFLA_EVENT_BONDING_FAILOVER;
1563 case NETDEV_NOTIFY_PEERS:
1564 rtnl_event_type = IFLA_EVENT_NOTIFY_PEERS;
1566 case NETDEV_RESEND_IGMP:
1567 rtnl_event_type = IFLA_EVENT_IGMP_RESEND;
1569 case NETDEV_CHANGEINFODATA:
1570 rtnl_event_type = IFLA_EVENT_BONDING_OPTIONS;
1576 return rtnl_event_type;
1579 static int put_master_ifindex(struct sk_buff *skb, struct net_device *dev)
1581 const struct net_device *upper_dev;
1586 upper_dev = netdev_master_upper_dev_get_rcu(dev);
1588 ret = nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex);
1594 static int nla_put_iflink(struct sk_buff *skb, const struct net_device *dev,
1597 int ifindex = dev_get_iflink(dev);
1599 if (force || dev->ifindex != ifindex)
1600 return nla_put_u32(skb, IFLA_LINK, ifindex);
1605 static noinline_for_stack int nla_put_ifalias(struct sk_buff *skb,
1606 struct net_device *dev)
1611 ret = dev_get_alias(dev, buf, sizeof(buf));
1612 return ret > 0 ? nla_put_string(skb, IFLA_IFALIAS, buf) : 0;
1615 static int rtnl_fill_link_netnsid(struct sk_buff *skb,
1616 const struct net_device *dev,
1617 struct net *src_net, gfp_t gfp)
1619 bool put_iflink = false;
1621 if (dev->rtnl_link_ops && dev->rtnl_link_ops->get_link_net) {
1622 struct net *link_net = dev->rtnl_link_ops->get_link_net(dev);
1624 if (!net_eq(dev_net(dev), link_net)) {
1625 int id = peernet2id_alloc(src_net, link_net, gfp);
1627 if (nla_put_s32(skb, IFLA_LINK_NETNSID, id))
1634 return nla_put_iflink(skb, dev, put_iflink);
1637 static int rtnl_fill_link_af(struct sk_buff *skb,
1638 const struct net_device *dev,
1639 u32 ext_filter_mask)
1641 const struct rtnl_af_ops *af_ops;
1642 struct nlattr *af_spec;
1644 af_spec = nla_nest_start_noflag(skb, IFLA_AF_SPEC);
1648 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
1652 if (!af_ops->fill_link_af)
1655 af = nla_nest_start_noflag(skb, af_ops->family);
1659 err = af_ops->fill_link_af(skb, dev, ext_filter_mask);
1661 * Caller may return ENODATA to indicate that there
1662 * was no data to be dumped. This is not an error, it
1663 * means we should trim the attribute header and
1666 if (err == -ENODATA)
1667 nla_nest_cancel(skb, af);
1671 nla_nest_end(skb, af);
1674 nla_nest_end(skb, af_spec);
1678 static int rtnl_fill_alt_ifnames(struct sk_buff *skb,
1679 const struct net_device *dev)
1681 struct netdev_name_node *name_node;
1684 list_for_each_entry(name_node, &dev->name_node->list, list) {
1685 if (nla_put_string(skb, IFLA_ALT_IFNAME, name_node->name))
1692 static int rtnl_fill_prop_list(struct sk_buff *skb,
1693 const struct net_device *dev)
1695 struct nlattr *prop_list;
1698 prop_list = nla_nest_start(skb, IFLA_PROP_LIST);
1702 ret = rtnl_fill_alt_ifnames(skb, dev);
1706 nla_nest_end(skb, prop_list);
1710 nla_nest_cancel(skb, prop_list);
1714 static int rtnl_fill_proto_down(struct sk_buff *skb,
1715 const struct net_device *dev)
1720 if (nla_put_u8(skb, IFLA_PROTO_DOWN, dev->proto_down))
1721 goto nla_put_failure;
1723 preason = dev->proto_down_reason;
1727 pr = nla_nest_start(skb, IFLA_PROTO_DOWN_REASON);
1731 if (nla_put_u32(skb, IFLA_PROTO_DOWN_REASON_VALUE, preason)) {
1732 nla_nest_cancel(skb, pr);
1733 goto nla_put_failure;
1736 nla_nest_end(skb, pr);
1743 static int rtnl_fill_devlink_port(struct sk_buff *skb,
1744 const struct net_device *dev)
1746 struct nlattr *devlink_port_nest;
1749 devlink_port_nest = nla_nest_start(skb, IFLA_DEVLINK_PORT);
1750 if (!devlink_port_nest)
1753 if (dev->devlink_port) {
1754 ret = devlink_nl_port_handle_fill(skb, dev->devlink_port);
1759 nla_nest_end(skb, devlink_port_nest);
1763 nla_nest_cancel(skb, devlink_port_nest);
1767 static int rtnl_fill_ifinfo(struct sk_buff *skb,
1768 struct net_device *dev, struct net *src_net,
1769 int type, u32 pid, u32 seq, u32 change,
1770 unsigned int flags, u32 ext_filter_mask,
1771 u32 event, int *new_nsid, int new_ifindex,
1772 int tgt_netnsid, gfp_t gfp)
1774 struct ifinfomsg *ifm;
1775 struct nlmsghdr *nlh;
1776 struct Qdisc *qdisc;
1779 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
1783 ifm = nlmsg_data(nlh);
1784 ifm->ifi_family = AF_UNSPEC;
1786 ifm->ifi_type = dev->type;
1787 ifm->ifi_index = dev->ifindex;
1788 ifm->ifi_flags = dev_get_flags(dev);
1789 ifm->ifi_change = change;
1791 if (tgt_netnsid >= 0 && nla_put_s32(skb, IFLA_TARGET_NETNSID, tgt_netnsid))
1792 goto nla_put_failure;
1794 qdisc = rtnl_dereference(dev->qdisc);
1795 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
1796 nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
1797 nla_put_u8(skb, IFLA_OPERSTATE,
1798 netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
1799 nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
1800 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
1801 nla_put_u32(skb, IFLA_MIN_MTU, dev->min_mtu) ||
1802 nla_put_u32(skb, IFLA_MAX_MTU, dev->max_mtu) ||
1803 nla_put_u32(skb, IFLA_GROUP, dev->group) ||
1804 nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
1805 nla_put_u32(skb, IFLA_ALLMULTI, dev->allmulti) ||
1806 nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
1807 nla_put_u32(skb, IFLA_GSO_MAX_SEGS, dev->gso_max_segs) ||
1808 nla_put_u32(skb, IFLA_GSO_MAX_SIZE, dev->gso_max_size) ||
1809 nla_put_u32(skb, IFLA_GRO_MAX_SIZE, dev->gro_max_size) ||
1810 nla_put_u32(skb, IFLA_TSO_MAX_SIZE, dev->tso_max_size) ||
1811 nla_put_u32(skb, IFLA_TSO_MAX_SEGS, dev->tso_max_segs) ||
1813 nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1815 put_master_ifindex(skb, dev) ||
1816 nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1818 nla_put_string(skb, IFLA_QDISC, qdisc->ops->id)) ||
1819 nla_put_ifalias(skb, dev) ||
1820 nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1821 atomic_read(&dev->carrier_up_count) +
1822 atomic_read(&dev->carrier_down_count)) ||
1823 nla_put_u32(skb, IFLA_CARRIER_UP_COUNT,
1824 atomic_read(&dev->carrier_up_count)) ||
1825 nla_put_u32(skb, IFLA_CARRIER_DOWN_COUNT,
1826 atomic_read(&dev->carrier_down_count)))
1827 goto nla_put_failure;
1829 if (rtnl_fill_proto_down(skb, dev))
1830 goto nla_put_failure;
1832 if (event != IFLA_EVENT_NONE) {
1833 if (nla_put_u32(skb, IFLA_EVENT, event))
1834 goto nla_put_failure;
1837 if (rtnl_fill_link_ifmap(skb, dev))
1838 goto nla_put_failure;
1840 if (dev->addr_len) {
1841 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1842 nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1843 goto nla_put_failure;
1846 if (rtnl_phys_port_id_fill(skb, dev))
1847 goto nla_put_failure;
1849 if (rtnl_phys_port_name_fill(skb, dev))
1850 goto nla_put_failure;
1852 if (rtnl_phys_switch_id_fill(skb, dev))
1853 goto nla_put_failure;
1855 if (rtnl_fill_stats(skb, dev))
1856 goto nla_put_failure;
1858 if (rtnl_fill_vf(skb, dev, ext_filter_mask))
1859 goto nla_put_failure;
1861 if (rtnl_port_fill(skb, dev, ext_filter_mask))
1862 goto nla_put_failure;
1864 if (rtnl_xdp_fill(skb, dev))
1865 goto nla_put_failure;
1867 if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1868 if (rtnl_link_fill(skb, dev) < 0)
1869 goto nla_put_failure;
1872 if (rtnl_fill_link_netnsid(skb, dev, src_net, gfp))
1873 goto nla_put_failure;
1876 nla_put_s32(skb, IFLA_NEW_NETNSID, *new_nsid) < 0)
1877 goto nla_put_failure;
1879 nla_put_s32(skb, IFLA_NEW_IFINDEX, new_ifindex) < 0)
1880 goto nla_put_failure;
1882 if (memchr_inv(dev->perm_addr, '\0', dev->addr_len) &&
1883 nla_put(skb, IFLA_PERM_ADDRESS, dev->addr_len, dev->perm_addr))
1884 goto nla_put_failure;
1887 if (rtnl_fill_link_af(skb, dev, ext_filter_mask))
1888 goto nla_put_failure_rcu;
1891 if (rtnl_fill_prop_list(skb, dev))
1892 goto nla_put_failure;
1894 if (dev->dev.parent &&
1895 nla_put_string(skb, IFLA_PARENT_DEV_NAME,
1896 dev_name(dev->dev.parent)))
1897 goto nla_put_failure;
1899 if (dev->dev.parent && dev->dev.parent->bus &&
1900 nla_put_string(skb, IFLA_PARENT_DEV_BUS_NAME,
1901 dev->dev.parent->bus->name))
1902 goto nla_put_failure;
1904 if (rtnl_fill_devlink_port(skb, dev))
1905 goto nla_put_failure;
1907 nlmsg_end(skb, nlh);
1910 nla_put_failure_rcu:
1913 nlmsg_cancel(skb, nlh);
1917 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1918 [IFLA_IFNAME] = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1919 [IFLA_ADDRESS] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1920 [IFLA_BROADCAST] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1921 [IFLA_MAP] = { .len = sizeof(struct rtnl_link_ifmap) },
1922 [IFLA_MTU] = { .type = NLA_U32 },
1923 [IFLA_LINK] = { .type = NLA_U32 },
1924 [IFLA_MASTER] = { .type = NLA_U32 },
1925 [IFLA_CARRIER] = { .type = NLA_U8 },
1926 [IFLA_TXQLEN] = { .type = NLA_U32 },
1927 [IFLA_WEIGHT] = { .type = NLA_U32 },
1928 [IFLA_OPERSTATE] = { .type = NLA_U8 },
1929 [IFLA_LINKMODE] = { .type = NLA_U8 },
1930 [IFLA_LINKINFO] = { .type = NLA_NESTED },
1931 [IFLA_NET_NS_PID] = { .type = NLA_U32 },
1932 [IFLA_NET_NS_FD] = { .type = NLA_U32 },
1933 /* IFLA_IFALIAS is a string, but policy is set to NLA_BINARY to
1934 * allow 0-length string (needed to remove an alias).
1936 [IFLA_IFALIAS] = { .type = NLA_BINARY, .len = IFALIASZ - 1 },
1937 [IFLA_VFINFO_LIST] = {. type = NLA_NESTED },
1938 [IFLA_VF_PORTS] = { .type = NLA_NESTED },
1939 [IFLA_PORT_SELF] = { .type = NLA_NESTED },
1940 [IFLA_AF_SPEC] = { .type = NLA_NESTED },
1941 [IFLA_EXT_MASK] = { .type = NLA_U32 },
1942 [IFLA_PROMISCUITY] = { .type = NLA_U32 },
1943 [IFLA_NUM_TX_QUEUES] = { .type = NLA_U32 },
1944 [IFLA_NUM_RX_QUEUES] = { .type = NLA_U32 },
1945 [IFLA_GSO_MAX_SEGS] = { .type = NLA_U32 },
1946 [IFLA_GSO_MAX_SIZE] = { .type = NLA_U32 },
1947 [IFLA_PHYS_PORT_ID] = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1948 [IFLA_CARRIER_CHANGES] = { .type = NLA_U32 }, /* ignored */
1949 [IFLA_PHYS_SWITCH_ID] = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1950 [IFLA_LINK_NETNSID] = { .type = NLA_S32 },
1951 [IFLA_PROTO_DOWN] = { .type = NLA_U8 },
1952 [IFLA_XDP] = { .type = NLA_NESTED },
1953 [IFLA_EVENT] = { .type = NLA_U32 },
1954 [IFLA_GROUP] = { .type = NLA_U32 },
1955 [IFLA_TARGET_NETNSID] = { .type = NLA_S32 },
1956 [IFLA_CARRIER_UP_COUNT] = { .type = NLA_U32 },
1957 [IFLA_CARRIER_DOWN_COUNT] = { .type = NLA_U32 },
1958 [IFLA_MIN_MTU] = { .type = NLA_U32 },
1959 [IFLA_MAX_MTU] = { .type = NLA_U32 },
1960 [IFLA_PROP_LIST] = { .type = NLA_NESTED },
1961 [IFLA_ALT_IFNAME] = { .type = NLA_STRING,
1962 .len = ALTIFNAMSIZ - 1 },
1963 [IFLA_PERM_ADDRESS] = { .type = NLA_REJECT },
1964 [IFLA_PROTO_DOWN_REASON] = { .type = NLA_NESTED },
1965 [IFLA_NEW_IFINDEX] = NLA_POLICY_MIN(NLA_S32, 1),
1966 [IFLA_PARENT_DEV_NAME] = { .type = NLA_NUL_STRING },
1967 [IFLA_GRO_MAX_SIZE] = { .type = NLA_U32 },
1968 [IFLA_TSO_MAX_SIZE] = { .type = NLA_REJECT },
1969 [IFLA_TSO_MAX_SEGS] = { .type = NLA_REJECT },
1970 [IFLA_ALLMULTI] = { .type = NLA_REJECT },
1973 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1974 [IFLA_INFO_KIND] = { .type = NLA_STRING },
1975 [IFLA_INFO_DATA] = { .type = NLA_NESTED },
1976 [IFLA_INFO_SLAVE_KIND] = { .type = NLA_STRING },
1977 [IFLA_INFO_SLAVE_DATA] = { .type = NLA_NESTED },
1980 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1981 [IFLA_VF_MAC] = { .len = sizeof(struct ifla_vf_mac) },
1982 [IFLA_VF_BROADCAST] = { .type = NLA_REJECT },
1983 [IFLA_VF_VLAN] = { .len = sizeof(struct ifla_vf_vlan) },
1984 [IFLA_VF_VLAN_LIST] = { .type = NLA_NESTED },
1985 [IFLA_VF_TX_RATE] = { .len = sizeof(struct ifla_vf_tx_rate) },
1986 [IFLA_VF_SPOOFCHK] = { .len = sizeof(struct ifla_vf_spoofchk) },
1987 [IFLA_VF_RATE] = { .len = sizeof(struct ifla_vf_rate) },
1988 [IFLA_VF_LINK_STATE] = { .len = sizeof(struct ifla_vf_link_state) },
1989 [IFLA_VF_RSS_QUERY_EN] = { .len = sizeof(struct ifla_vf_rss_query_en) },
1990 [IFLA_VF_STATS] = { .type = NLA_NESTED },
1991 [IFLA_VF_TRUST] = { .len = sizeof(struct ifla_vf_trust) },
1992 [IFLA_VF_IB_NODE_GUID] = { .len = sizeof(struct ifla_vf_guid) },
1993 [IFLA_VF_IB_PORT_GUID] = { .len = sizeof(struct ifla_vf_guid) },
1996 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1997 [IFLA_PORT_VF] = { .type = NLA_U32 },
1998 [IFLA_PORT_PROFILE] = { .type = NLA_STRING,
1999 .len = PORT_PROFILE_MAX },
2000 [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
2001 .len = PORT_UUID_MAX },
2002 [IFLA_PORT_HOST_UUID] = { .type = NLA_STRING,
2003 .len = PORT_UUID_MAX },
2004 [IFLA_PORT_REQUEST] = { .type = NLA_U8, },
2005 [IFLA_PORT_RESPONSE] = { .type = NLA_U16, },
2007 /* Unused, but we need to keep it here since user space could
2008 * fill it. It's also broken with regard to NLA_BINARY use in
2009 * combination with structs.
2011 [IFLA_PORT_VSI_TYPE] = { .type = NLA_BINARY,
2012 .len = sizeof(struct ifla_port_vsi) },
2015 static const struct nla_policy ifla_xdp_policy[IFLA_XDP_MAX + 1] = {
2016 [IFLA_XDP_UNSPEC] = { .strict_start_type = IFLA_XDP_EXPECTED_FD },
2017 [IFLA_XDP_FD] = { .type = NLA_S32 },
2018 [IFLA_XDP_EXPECTED_FD] = { .type = NLA_S32 },
2019 [IFLA_XDP_ATTACHED] = { .type = NLA_U8 },
2020 [IFLA_XDP_FLAGS] = { .type = NLA_U32 },
2021 [IFLA_XDP_PROG_ID] = { .type = NLA_U32 },
2024 static const struct rtnl_link_ops *linkinfo_to_kind_ops(const struct nlattr *nla)
2026 const struct rtnl_link_ops *ops = NULL;
2027 struct nlattr *linfo[IFLA_INFO_MAX + 1];
2029 if (nla_parse_nested_deprecated(linfo, IFLA_INFO_MAX, nla, ifla_info_policy, NULL) < 0)
2032 if (linfo[IFLA_INFO_KIND]) {
2033 char kind[MODULE_NAME_LEN];
2035 nla_strscpy(kind, linfo[IFLA_INFO_KIND], sizeof(kind));
2036 ops = rtnl_link_ops_get(kind);
2042 static bool link_master_filtered(struct net_device *dev, int master_idx)
2044 struct net_device *master;
2049 master = netdev_master_upper_dev_get(dev);
2051 /* 0 is already used to denote IFLA_MASTER wasn't passed, therefore need
2052 * another invalid value for ifindex to denote "no master".
2054 if (master_idx == -1)
2057 if (!master || master->ifindex != master_idx)
2063 static bool link_kind_filtered(const struct net_device *dev,
2064 const struct rtnl_link_ops *kind_ops)
2066 if (kind_ops && dev->rtnl_link_ops != kind_ops)
2072 static bool link_dump_filtered(struct net_device *dev,
2074 const struct rtnl_link_ops *kind_ops)
2076 if (link_master_filtered(dev, master_idx) ||
2077 link_kind_filtered(dev, kind_ops))
2084 * rtnl_get_net_ns_capable - Get netns if sufficiently privileged.
2085 * @sk: netlink socket
2086 * @netnsid: network namespace identifier
2088 * Returns the network namespace identified by netnsid on success or an error
2089 * pointer on failure.
2091 struct net *rtnl_get_net_ns_capable(struct sock *sk, int netnsid)
2095 net = get_net_ns_by_id(sock_net(sk), netnsid);
2097 return ERR_PTR(-EINVAL);
2099 /* For now, the caller is required to have CAP_NET_ADMIN in
2100 * the user namespace owning the target net ns.
2102 if (!sk_ns_capable(sk, net->user_ns, CAP_NET_ADMIN)) {
2104 return ERR_PTR(-EACCES);
2108 EXPORT_SYMBOL_GPL(rtnl_get_net_ns_capable);
2110 static int rtnl_valid_dump_ifinfo_req(const struct nlmsghdr *nlh,
2111 bool strict_check, struct nlattr **tb,
2112 struct netlink_ext_ack *extack)
2117 struct ifinfomsg *ifm;
2119 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
2120 NL_SET_ERR_MSG(extack, "Invalid header for link dump");
2124 ifm = nlmsg_data(nlh);
2125 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
2127 NL_SET_ERR_MSG(extack, "Invalid values in header for link dump request");
2130 if (ifm->ifi_index) {
2131 NL_SET_ERR_MSG(extack, "Filter by device index not supported for link dumps");
2135 return nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb,
2136 IFLA_MAX, ifla_policy,
2140 /* A hack to preserve kernel<->userspace interface.
2141 * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
2142 * However, before Linux v3.9 the code here assumed rtgenmsg and that's
2143 * what iproute2 < v3.9.0 used.
2144 * We can detect the old iproute2. Even including the IFLA_EXT_MASK
2145 * attribute, its netlink message is shorter than struct ifinfomsg.
2147 hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
2148 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
2150 return nlmsg_parse_deprecated(nlh, hdrlen, tb, IFLA_MAX, ifla_policy,
2154 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
2156 struct netlink_ext_ack *extack = cb->extack;
2157 const struct nlmsghdr *nlh = cb->nlh;
2158 struct net *net = sock_net(skb->sk);
2159 struct net *tgt_net = net;
2162 struct net_device *dev;
2163 struct hlist_head *head;
2164 struct nlattr *tb[IFLA_MAX+1];
2165 u32 ext_filter_mask = 0;
2166 const struct rtnl_link_ops *kind_ops = NULL;
2167 unsigned int flags = NLM_F_MULTI;
2173 s_idx = cb->args[1];
2175 err = rtnl_valid_dump_ifinfo_req(nlh, cb->strict_check, tb, extack);
2177 if (cb->strict_check)
2183 for (i = 0; i <= IFLA_MAX; ++i) {
2187 /* new attributes should only be added with strict checking */
2189 case IFLA_TARGET_NETNSID:
2190 netnsid = nla_get_s32(tb[i]);
2191 tgt_net = rtnl_get_net_ns_capable(skb->sk, netnsid);
2192 if (IS_ERR(tgt_net)) {
2193 NL_SET_ERR_MSG(extack, "Invalid target network namespace id");
2194 return PTR_ERR(tgt_net);
2198 ext_filter_mask = nla_get_u32(tb[i]);
2201 master_idx = nla_get_u32(tb[i]);
2204 kind_ops = linkinfo_to_kind_ops(tb[i]);
2207 if (cb->strict_check) {
2208 NL_SET_ERR_MSG(extack, "Unsupported attribute in link dump request");
2214 if (master_idx || kind_ops)
2215 flags |= NLM_F_DUMP_FILTERED;
2218 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
2220 head = &tgt_net->dev_index_head[h];
2221 hlist_for_each_entry(dev, head, index_hlist) {
2222 if (link_dump_filtered(dev, master_idx, kind_ops))
2226 err = rtnl_fill_ifinfo(skb, dev, net,
2228 NETLINK_CB(cb->skb).portid,
2229 nlh->nlmsg_seq, 0, flags,
2230 ext_filter_mask, 0, NULL, 0,
2231 netnsid, GFP_KERNEL);
2234 if (likely(skb->len))
2248 cb->seq = tgt_net->dev_base_seq;
2249 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
2256 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len,
2257 struct netlink_ext_ack *exterr)
2259 return nla_parse_deprecated(tb, IFLA_MAX, head, len, ifla_policy,
2262 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
2264 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
2267 /* Examine the link attributes and figure out which
2268 * network namespace we are talking about.
2270 if (tb[IFLA_NET_NS_PID])
2271 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
2272 else if (tb[IFLA_NET_NS_FD])
2273 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
2275 net = get_net(src_net);
2278 EXPORT_SYMBOL(rtnl_link_get_net);
2280 /* Figure out which network namespace we are talking about by
2281 * examining the link attributes in the following order:
2283 * 1. IFLA_NET_NS_PID
2285 * 3. IFLA_TARGET_NETNSID
2287 static struct net *rtnl_link_get_net_by_nlattr(struct net *src_net,
2288 struct nlattr *tb[])
2292 if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD])
2293 return rtnl_link_get_net(src_net, tb);
2295 if (!tb[IFLA_TARGET_NETNSID])
2296 return get_net(src_net);
2298 net = get_net_ns_by_id(src_net, nla_get_u32(tb[IFLA_TARGET_NETNSID]));
2300 return ERR_PTR(-EINVAL);
2305 static struct net *rtnl_link_get_net_capable(const struct sk_buff *skb,
2306 struct net *src_net,
2307 struct nlattr *tb[], int cap)
2311 net = rtnl_link_get_net_by_nlattr(src_net, tb);
2315 if (!netlink_ns_capable(skb, net->user_ns, cap)) {
2317 return ERR_PTR(-EPERM);
2323 /* Verify that rtnetlink requests do not pass additional properties
2324 * potentially referring to different network namespaces.
2326 static int rtnl_ensure_unique_netns(struct nlattr *tb[],
2327 struct netlink_ext_ack *extack,
2331 if (netns_id_only) {
2332 if (!tb[IFLA_NET_NS_PID] && !tb[IFLA_NET_NS_FD])
2335 NL_SET_ERR_MSG(extack, "specified netns attribute not supported");
2339 if (tb[IFLA_TARGET_NETNSID] && (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]))
2342 if (tb[IFLA_NET_NS_PID] && (tb[IFLA_TARGET_NETNSID] || tb[IFLA_NET_NS_FD]))
2345 if (tb[IFLA_NET_NS_FD] && (tb[IFLA_TARGET_NETNSID] || tb[IFLA_NET_NS_PID]))
2351 NL_SET_ERR_MSG(extack, "multiple netns identifying attributes specified");
2355 static int rtnl_set_vf_rate(struct net_device *dev, int vf, int min_tx_rate,
2358 const struct net_device_ops *ops = dev->netdev_ops;
2360 if (!ops->ndo_set_vf_rate)
2362 if (max_tx_rate && max_tx_rate < min_tx_rate)
2365 return ops->ndo_set_vf_rate(dev, vf, min_tx_rate, max_tx_rate);
2368 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[],
2369 struct netlink_ext_ack *extack)
2372 if (tb[IFLA_ADDRESS] &&
2373 nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
2376 if (tb[IFLA_BROADCAST] &&
2377 nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
2381 if (tb[IFLA_AF_SPEC]) {
2385 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2386 const struct rtnl_af_ops *af_ops;
2388 af_ops = rtnl_af_lookup(nla_type(af));
2390 return -EAFNOSUPPORT;
2392 if (!af_ops->set_link_af)
2395 if (af_ops->validate_link_af) {
2396 err = af_ops->validate_link_af(dev, af, extack);
2406 static int handle_infiniband_guid(struct net_device *dev, struct ifla_vf_guid *ivt,
2409 const struct net_device_ops *ops = dev->netdev_ops;
2411 return ops->ndo_set_vf_guid(dev, ivt->vf, ivt->guid, guid_type);
2414 static int handle_vf_guid(struct net_device *dev, struct ifla_vf_guid *ivt, int guid_type)
2416 if (dev->type != ARPHRD_INFINIBAND)
2419 return handle_infiniband_guid(dev, ivt, guid_type);
2422 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb)
2424 const struct net_device_ops *ops = dev->netdev_ops;
2427 if (tb[IFLA_VF_MAC]) {
2428 struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]);
2430 if (ivm->vf >= INT_MAX)
2433 if (ops->ndo_set_vf_mac)
2434 err = ops->ndo_set_vf_mac(dev, ivm->vf,
2440 if (tb[IFLA_VF_VLAN]) {
2441 struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]);
2443 if (ivv->vf >= INT_MAX)
2446 if (ops->ndo_set_vf_vlan)
2447 err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan,
2449 htons(ETH_P_8021Q));
2454 if (tb[IFLA_VF_VLAN_LIST]) {
2455 struct ifla_vf_vlan_info *ivvl[MAX_VLAN_LIST_LEN];
2456 struct nlattr *attr;
2460 if (!ops->ndo_set_vf_vlan)
2463 nla_for_each_nested(attr, tb[IFLA_VF_VLAN_LIST], rem) {
2464 if (nla_type(attr) != IFLA_VF_VLAN_INFO ||
2465 nla_len(attr) < NLA_HDRLEN) {
2468 if (len >= MAX_VLAN_LIST_LEN)
2470 ivvl[len] = nla_data(attr);
2477 if (ivvl[0]->vf >= INT_MAX)
2479 err = ops->ndo_set_vf_vlan(dev, ivvl[0]->vf, ivvl[0]->vlan,
2480 ivvl[0]->qos, ivvl[0]->vlan_proto);
2485 if (tb[IFLA_VF_TX_RATE]) {
2486 struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]);
2487 struct ifla_vf_info ivf;
2489 if (ivt->vf >= INT_MAX)
2492 if (ops->ndo_get_vf_config)
2493 err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf);
2497 err = rtnl_set_vf_rate(dev, ivt->vf,
2498 ivf.min_tx_rate, ivt->rate);
2503 if (tb[IFLA_VF_RATE]) {
2504 struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]);
2506 if (ivt->vf >= INT_MAX)
2509 err = rtnl_set_vf_rate(dev, ivt->vf,
2510 ivt->min_tx_rate, ivt->max_tx_rate);
2515 if (tb[IFLA_VF_SPOOFCHK]) {
2516 struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]);
2518 if (ivs->vf >= INT_MAX)
2521 if (ops->ndo_set_vf_spoofchk)
2522 err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
2528 if (tb[IFLA_VF_LINK_STATE]) {
2529 struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]);
2531 if (ivl->vf >= INT_MAX)
2534 if (ops->ndo_set_vf_link_state)
2535 err = ops->ndo_set_vf_link_state(dev, ivl->vf,
2541 if (tb[IFLA_VF_RSS_QUERY_EN]) {
2542 struct ifla_vf_rss_query_en *ivrssq_en;
2545 ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]);
2546 if (ivrssq_en->vf >= INT_MAX)
2548 if (ops->ndo_set_vf_rss_query_en)
2549 err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf,
2550 ivrssq_en->setting);
2555 if (tb[IFLA_VF_TRUST]) {
2556 struct ifla_vf_trust *ivt = nla_data(tb[IFLA_VF_TRUST]);
2558 if (ivt->vf >= INT_MAX)
2561 if (ops->ndo_set_vf_trust)
2562 err = ops->ndo_set_vf_trust(dev, ivt->vf, ivt->setting);
2567 if (tb[IFLA_VF_IB_NODE_GUID]) {
2568 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_NODE_GUID]);
2570 if (ivt->vf >= INT_MAX)
2572 if (!ops->ndo_set_vf_guid)
2574 return handle_vf_guid(dev, ivt, IFLA_VF_IB_NODE_GUID);
2577 if (tb[IFLA_VF_IB_PORT_GUID]) {
2578 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_PORT_GUID]);
2580 if (ivt->vf >= INT_MAX)
2582 if (!ops->ndo_set_vf_guid)
2585 return handle_vf_guid(dev, ivt, IFLA_VF_IB_PORT_GUID);
2591 static int do_set_master(struct net_device *dev, int ifindex,
2592 struct netlink_ext_ack *extack)
2594 struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
2595 const struct net_device_ops *ops;
2599 if (upper_dev->ifindex == ifindex)
2601 ops = upper_dev->netdev_ops;
2602 if (ops->ndo_del_slave) {
2603 err = ops->ndo_del_slave(upper_dev, dev);
2612 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
2615 ops = upper_dev->netdev_ops;
2616 if (ops->ndo_add_slave) {
2617 err = ops->ndo_add_slave(upper_dev, dev, extack);
2627 static const struct nla_policy ifla_proto_down_reason_policy[IFLA_PROTO_DOWN_REASON_VALUE + 1] = {
2628 [IFLA_PROTO_DOWN_REASON_MASK] = { .type = NLA_U32 },
2629 [IFLA_PROTO_DOWN_REASON_VALUE] = { .type = NLA_U32 },
2632 static int do_set_proto_down(struct net_device *dev,
2633 struct nlattr *nl_proto_down,
2634 struct nlattr *nl_proto_down_reason,
2635 struct netlink_ext_ack *extack)
2637 struct nlattr *pdreason[IFLA_PROTO_DOWN_REASON_MAX + 1];
2638 unsigned long mask = 0;
2643 if (!(dev->priv_flags & IFF_CHANGE_PROTO_DOWN)) {
2644 NL_SET_ERR_MSG(extack, "Protodown not supported by device");
2648 if (nl_proto_down_reason) {
2649 err = nla_parse_nested_deprecated(pdreason,
2650 IFLA_PROTO_DOWN_REASON_MAX,
2651 nl_proto_down_reason,
2652 ifla_proto_down_reason_policy,
2657 if (!pdreason[IFLA_PROTO_DOWN_REASON_VALUE]) {
2658 NL_SET_ERR_MSG(extack, "Invalid protodown reason value");
2662 value = nla_get_u32(pdreason[IFLA_PROTO_DOWN_REASON_VALUE]);
2664 if (pdreason[IFLA_PROTO_DOWN_REASON_MASK])
2665 mask = nla_get_u32(pdreason[IFLA_PROTO_DOWN_REASON_MASK]);
2667 dev_change_proto_down_reason(dev, mask, value);
2670 if (nl_proto_down) {
2671 proto_down = nla_get_u8(nl_proto_down);
2673 /* Don't turn off protodown if there are active reasons */
2674 if (!proto_down && dev->proto_down_reason) {
2675 NL_SET_ERR_MSG(extack, "Cannot clear protodown, active reasons");
2678 err = dev_change_proto_down(dev,
2687 #define DO_SETLINK_MODIFIED 0x01
2688 /* notify flag means notify + modified. */
2689 #define DO_SETLINK_NOTIFY 0x03
2690 static int do_setlink(const struct sk_buff *skb,
2691 struct net_device *dev, struct ifinfomsg *ifm,
2692 struct netlink_ext_ack *extack,
2693 struct nlattr **tb, int status)
2695 const struct net_device_ops *ops = dev->netdev_ops;
2696 char ifname[IFNAMSIZ];
2699 err = validate_linkmsg(dev, tb, extack);
2703 if (tb[IFLA_IFNAME])
2704 nla_strscpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2708 if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD] || tb[IFLA_TARGET_NETNSID]) {
2709 const char *pat = ifname[0] ? ifname : NULL;
2713 net = rtnl_link_get_net_capable(skb, dev_net(dev),
2720 if (tb[IFLA_NEW_IFINDEX])
2721 new_ifindex = nla_get_s32(tb[IFLA_NEW_IFINDEX]);
2725 err = __dev_change_net_namespace(dev, net, pat, new_ifindex);
2729 status |= DO_SETLINK_MODIFIED;
2733 struct rtnl_link_ifmap *u_map;
2736 if (!ops->ndo_set_config) {
2741 if (!netif_device_present(dev)) {
2746 u_map = nla_data(tb[IFLA_MAP]);
2747 k_map.mem_start = (unsigned long) u_map->mem_start;
2748 k_map.mem_end = (unsigned long) u_map->mem_end;
2749 k_map.base_addr = (unsigned short) u_map->base_addr;
2750 k_map.irq = (unsigned char) u_map->irq;
2751 k_map.dma = (unsigned char) u_map->dma;
2752 k_map.port = (unsigned char) u_map->port;
2754 err = ops->ndo_set_config(dev, &k_map);
2758 status |= DO_SETLINK_NOTIFY;
2761 if (tb[IFLA_ADDRESS]) {
2762 struct sockaddr *sa;
2765 len = sizeof(sa_family_t) + max_t(size_t, dev->addr_len,
2767 sa = kmalloc(len, GFP_KERNEL);
2772 sa->sa_family = dev->type;
2773 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
2775 err = dev_set_mac_address_user(dev, sa, extack);
2779 status |= DO_SETLINK_MODIFIED;
2783 err = dev_set_mtu_ext(dev, nla_get_u32(tb[IFLA_MTU]), extack);
2786 status |= DO_SETLINK_MODIFIED;
2789 if (tb[IFLA_GROUP]) {
2790 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2791 status |= DO_SETLINK_NOTIFY;
2795 * Interface selected by interface index but interface
2796 * name provided implies that a name change has been
2799 if (ifm->ifi_index > 0 && ifname[0]) {
2800 err = dev_change_name(dev, ifname);
2803 status |= DO_SETLINK_MODIFIED;
2806 if (tb[IFLA_IFALIAS]) {
2807 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
2808 nla_len(tb[IFLA_IFALIAS]));
2811 status |= DO_SETLINK_NOTIFY;
2814 if (tb[IFLA_BROADCAST]) {
2815 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
2816 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
2819 if (tb[IFLA_MASTER]) {
2820 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]), extack);
2823 status |= DO_SETLINK_MODIFIED;
2826 if (ifm->ifi_flags || ifm->ifi_change) {
2827 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm),
2833 if (tb[IFLA_CARRIER]) {
2834 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
2837 status |= DO_SETLINK_MODIFIED;
2840 if (tb[IFLA_TXQLEN]) {
2841 unsigned int value = nla_get_u32(tb[IFLA_TXQLEN]);
2843 err = dev_change_tx_queue_len(dev, value);
2846 status |= DO_SETLINK_MODIFIED;
2849 if (tb[IFLA_GSO_MAX_SIZE]) {
2850 u32 max_size = nla_get_u32(tb[IFLA_GSO_MAX_SIZE]);
2852 if (max_size > dev->tso_max_size) {
2857 if (dev->gso_max_size ^ max_size) {
2858 netif_set_gso_max_size(dev, max_size);
2859 status |= DO_SETLINK_MODIFIED;
2863 if (tb[IFLA_GSO_MAX_SEGS]) {
2864 u32 max_segs = nla_get_u32(tb[IFLA_GSO_MAX_SEGS]);
2866 if (max_segs > GSO_MAX_SEGS || max_segs > dev->tso_max_segs) {
2871 if (dev->gso_max_segs ^ max_segs) {
2872 netif_set_gso_max_segs(dev, max_segs);
2873 status |= DO_SETLINK_MODIFIED;
2877 if (tb[IFLA_GRO_MAX_SIZE]) {
2878 u32 gro_max_size = nla_get_u32(tb[IFLA_GRO_MAX_SIZE]);
2880 if (dev->gro_max_size ^ gro_max_size) {
2881 netif_set_gro_max_size(dev, gro_max_size);
2882 status |= DO_SETLINK_MODIFIED;
2886 if (tb[IFLA_OPERSTATE])
2887 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2889 if (tb[IFLA_LINKMODE]) {
2890 unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
2892 write_lock(&dev_base_lock);
2893 if (dev->link_mode ^ value)
2894 status |= DO_SETLINK_NOTIFY;
2895 dev->link_mode = value;
2896 write_unlock(&dev_base_lock);
2899 if (tb[IFLA_VFINFO_LIST]) {
2900 struct nlattr *vfinfo[IFLA_VF_MAX + 1];
2901 struct nlattr *attr;
2904 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
2905 if (nla_type(attr) != IFLA_VF_INFO ||
2906 nla_len(attr) < NLA_HDRLEN) {
2910 err = nla_parse_nested_deprecated(vfinfo, IFLA_VF_MAX,
2916 err = do_setvfinfo(dev, vfinfo);
2919 status |= DO_SETLINK_NOTIFY;
2924 if (tb[IFLA_VF_PORTS]) {
2925 struct nlattr *port[IFLA_PORT_MAX+1];
2926 struct nlattr *attr;
2931 if (!ops->ndo_set_vf_port)
2934 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
2935 if (nla_type(attr) != IFLA_VF_PORT ||
2936 nla_len(attr) < NLA_HDRLEN) {
2940 err = nla_parse_nested_deprecated(port, IFLA_PORT_MAX,
2946 if (!port[IFLA_PORT_VF]) {
2950 vf = nla_get_u32(port[IFLA_PORT_VF]);
2951 err = ops->ndo_set_vf_port(dev, vf, port);
2954 status |= DO_SETLINK_NOTIFY;
2959 if (tb[IFLA_PORT_SELF]) {
2960 struct nlattr *port[IFLA_PORT_MAX+1];
2962 err = nla_parse_nested_deprecated(port, IFLA_PORT_MAX,
2964 ifla_port_policy, NULL);
2969 if (ops->ndo_set_vf_port)
2970 err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
2973 status |= DO_SETLINK_NOTIFY;
2976 if (tb[IFLA_AF_SPEC]) {
2980 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2981 const struct rtnl_af_ops *af_ops;
2983 BUG_ON(!(af_ops = rtnl_af_lookup(nla_type(af))));
2985 err = af_ops->set_link_af(dev, af, extack);
2989 status |= DO_SETLINK_NOTIFY;
2994 if (tb[IFLA_PROTO_DOWN] || tb[IFLA_PROTO_DOWN_REASON]) {
2995 err = do_set_proto_down(dev, tb[IFLA_PROTO_DOWN],
2996 tb[IFLA_PROTO_DOWN_REASON], extack);
2999 status |= DO_SETLINK_NOTIFY;
3003 struct nlattr *xdp[IFLA_XDP_MAX + 1];
3006 err = nla_parse_nested_deprecated(xdp, IFLA_XDP_MAX,
3008 ifla_xdp_policy, NULL);
3012 if (xdp[IFLA_XDP_ATTACHED] || xdp[IFLA_XDP_PROG_ID]) {
3017 if (xdp[IFLA_XDP_FLAGS]) {
3018 xdp_flags = nla_get_u32(xdp[IFLA_XDP_FLAGS]);
3019 if (xdp_flags & ~XDP_FLAGS_MASK) {
3023 if (hweight32(xdp_flags & XDP_FLAGS_MODES) > 1) {
3029 if (xdp[IFLA_XDP_FD]) {
3030 int expected_fd = -1;
3032 if (xdp_flags & XDP_FLAGS_REPLACE) {
3033 if (!xdp[IFLA_XDP_EXPECTED_FD]) {
3038 nla_get_s32(xdp[IFLA_XDP_EXPECTED_FD]);
3041 err = dev_change_xdp_fd(dev, extack,
3042 nla_get_s32(xdp[IFLA_XDP_FD]),
3047 status |= DO_SETLINK_NOTIFY;
3052 if (status & DO_SETLINK_MODIFIED) {
3053 if ((status & DO_SETLINK_NOTIFY) == DO_SETLINK_NOTIFY)
3054 netdev_state_change(dev);
3057 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",
3064 static struct net_device *rtnl_dev_get(struct net *net,
3065 struct nlattr *tb[])
3067 char ifname[ALTIFNAMSIZ];
3069 if (tb[IFLA_IFNAME])
3070 nla_strscpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
3071 else if (tb[IFLA_ALT_IFNAME])
3072 nla_strscpy(ifname, tb[IFLA_ALT_IFNAME], ALTIFNAMSIZ);
3076 return __dev_get_by_name(net, ifname);
3079 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3080 struct netlink_ext_ack *extack)
3082 struct net *net = sock_net(skb->sk);
3083 struct ifinfomsg *ifm;
3084 struct net_device *dev;
3086 struct nlattr *tb[IFLA_MAX+1];
3088 err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3089 ifla_policy, extack);
3093 err = rtnl_ensure_unique_netns(tb, extack, false);
3098 ifm = nlmsg_data(nlh);
3099 if (ifm->ifi_index > 0)
3100 dev = __dev_get_by_index(net, ifm->ifi_index);
3101 else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3102 dev = rtnl_dev_get(net, tb);
3111 err = do_setlink(skb, dev, ifm, extack, tb, 0);
3116 static int rtnl_group_dellink(const struct net *net, int group)
3118 struct net_device *dev, *aux;
3119 LIST_HEAD(list_kill);
3125 for_each_netdev(net, dev) {
3126 if (dev->group == group) {
3127 const struct rtnl_link_ops *ops;
3130 ops = dev->rtnl_link_ops;
3131 if (!ops || !ops->dellink)
3139 for_each_netdev_safe(net, dev, aux) {
3140 if (dev->group == group) {
3141 const struct rtnl_link_ops *ops;
3143 ops = dev->rtnl_link_ops;
3144 ops->dellink(dev, &list_kill);
3147 unregister_netdevice_many(&list_kill);
3152 int rtnl_delete_link(struct net_device *dev, u32 portid, const struct nlmsghdr *nlh)
3154 const struct rtnl_link_ops *ops;
3155 LIST_HEAD(list_kill);
3157 ops = dev->rtnl_link_ops;
3158 if (!ops || !ops->dellink)
3161 ops->dellink(dev, &list_kill);
3162 unregister_netdevice_many_notify(&list_kill, portid, nlh);
3166 EXPORT_SYMBOL_GPL(rtnl_delete_link);
3168 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
3169 struct netlink_ext_ack *extack)
3171 struct net *net = sock_net(skb->sk);
3172 u32 portid = NETLINK_CB(skb).portid;
3173 struct net *tgt_net = net;
3174 struct net_device *dev = NULL;
3175 struct ifinfomsg *ifm;
3176 struct nlattr *tb[IFLA_MAX+1];
3180 err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3181 ifla_policy, extack);
3185 err = rtnl_ensure_unique_netns(tb, extack, true);
3189 if (tb[IFLA_TARGET_NETNSID]) {
3190 netnsid = nla_get_s32(tb[IFLA_TARGET_NETNSID]);
3191 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(skb).sk, netnsid);
3192 if (IS_ERR(tgt_net))
3193 return PTR_ERR(tgt_net);
3197 ifm = nlmsg_data(nlh);
3198 if (ifm->ifi_index > 0)
3199 dev = __dev_get_by_index(tgt_net, ifm->ifi_index);
3200 else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3201 dev = rtnl_dev_get(net, tb);
3202 else if (tb[IFLA_GROUP])
3203 err = rtnl_group_dellink(tgt_net, nla_get_u32(tb[IFLA_GROUP]));
3208 if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME] || ifm->ifi_index > 0)
3214 err = rtnl_delete_link(dev, portid, nlh);
3223 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm,
3224 u32 portid, const struct nlmsghdr *nlh)
3226 unsigned int old_flags;
3229 old_flags = dev->flags;
3230 if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
3231 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm),
3237 if (dev->rtnl_link_state == RTNL_LINK_INITIALIZED) {
3238 __dev_notify_flags(dev, old_flags, (old_flags ^ dev->flags), portid, nlh);
3240 dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
3241 __dev_notify_flags(dev, old_flags, ~0U, portid, nlh);
3245 EXPORT_SYMBOL(rtnl_configure_link);
3247 struct net_device *rtnl_create_link(struct net *net, const char *ifname,
3248 unsigned char name_assign_type,
3249 const struct rtnl_link_ops *ops,
3250 struct nlattr *tb[],
3251 struct netlink_ext_ack *extack)
3253 struct net_device *dev;
3254 unsigned int num_tx_queues = 1;
3255 unsigned int num_rx_queues = 1;
3257 if (tb[IFLA_NUM_TX_QUEUES])
3258 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
3259 else if (ops->get_num_tx_queues)
3260 num_tx_queues = ops->get_num_tx_queues();
3262 if (tb[IFLA_NUM_RX_QUEUES])
3263 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
3264 else if (ops->get_num_rx_queues)
3265 num_rx_queues = ops->get_num_rx_queues();
3267 if (num_tx_queues < 1 || num_tx_queues > 4096) {
3268 NL_SET_ERR_MSG(extack, "Invalid number of transmit queues");
3269 return ERR_PTR(-EINVAL);
3272 if (num_rx_queues < 1 || num_rx_queues > 4096) {
3273 NL_SET_ERR_MSG(extack, "Invalid number of receive queues");
3274 return ERR_PTR(-EINVAL);
3278 dev = ops->alloc(tb, ifname, name_assign_type,
3279 num_tx_queues, num_rx_queues);
3283 dev = alloc_netdev_mqs(ops->priv_size, ifname,
3284 name_assign_type, ops->setup,
3285 num_tx_queues, num_rx_queues);
3289 return ERR_PTR(-ENOMEM);
3291 dev_net_set(dev, net);
3292 dev->rtnl_link_ops = ops;
3293 dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
3296 u32 mtu = nla_get_u32(tb[IFLA_MTU]);
3299 err = dev_validate_mtu(dev, mtu, extack);
3302 return ERR_PTR(err);
3306 if (tb[IFLA_ADDRESS]) {
3307 __dev_addr_set(dev, nla_data(tb[IFLA_ADDRESS]),
3308 nla_len(tb[IFLA_ADDRESS]));
3309 dev->addr_assign_type = NET_ADDR_SET;
3311 if (tb[IFLA_BROADCAST])
3312 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
3313 nla_len(tb[IFLA_BROADCAST]));
3314 if (tb[IFLA_TXQLEN])
3315 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
3316 if (tb[IFLA_OPERSTATE])
3317 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
3318 if (tb[IFLA_LINKMODE])
3319 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
3321 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
3322 if (tb[IFLA_GSO_MAX_SIZE])
3323 netif_set_gso_max_size(dev, nla_get_u32(tb[IFLA_GSO_MAX_SIZE]));
3324 if (tb[IFLA_GSO_MAX_SEGS])
3325 netif_set_gso_max_segs(dev, nla_get_u32(tb[IFLA_GSO_MAX_SEGS]));
3326 if (tb[IFLA_GRO_MAX_SIZE])
3327 netif_set_gro_max_size(dev, nla_get_u32(tb[IFLA_GRO_MAX_SIZE]));
3331 EXPORT_SYMBOL(rtnl_create_link);
3333 static int rtnl_group_changelink(const struct sk_buff *skb,
3334 struct net *net, int group,
3335 struct ifinfomsg *ifm,
3336 struct netlink_ext_ack *extack,
3339 struct net_device *dev, *aux;
3342 for_each_netdev_safe(net, dev, aux) {
3343 if (dev->group == group) {
3344 err = do_setlink(skb, dev, ifm, extack, tb, 0);
3353 static int rtnl_newlink_create(struct sk_buff *skb, struct ifinfomsg *ifm,
3354 const struct rtnl_link_ops *ops,
3355 const struct nlmsghdr *nlh,
3356 struct nlattr **tb, struct nlattr **data,
3357 struct netlink_ext_ack *extack)
3359 unsigned char name_assign_type = NET_NAME_USER;
3360 struct net *net = sock_net(skb->sk);
3361 u32 portid = NETLINK_CB(skb).portid;
3362 struct net *dest_net, *link_net;
3363 struct net_device *dev;
3364 char ifname[IFNAMSIZ];
3367 if (!ops->alloc && !ops->setup)
3370 if (tb[IFLA_IFNAME]) {
3371 nla_strscpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
3373 snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
3374 name_assign_type = NET_NAME_ENUM;
3377 dest_net = rtnl_link_get_net_capable(skb, net, tb, CAP_NET_ADMIN);
3378 if (IS_ERR(dest_net))
3379 return PTR_ERR(dest_net);
3381 if (tb[IFLA_LINK_NETNSID]) {
3382 int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
3384 link_net = get_net_ns_by_id(dest_net, id);
3386 NL_SET_ERR_MSG(extack, "Unknown network namespace id");
3391 if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
3397 dev = rtnl_create_link(link_net ? : dest_net, ifname,
3398 name_assign_type, ops, tb, extack);
3404 dev->ifindex = ifm->ifi_index;
3407 err = ops->newlink(link_net ? : net, dev, tb, data, extack);
3409 err = register_netdevice(dev);
3415 err = rtnl_configure_link(dev, ifm, portid, nlh);
3417 goto out_unregister;
3419 err = dev_change_net_namespace(dev, dest_net, ifname);
3421 goto out_unregister;
3423 if (tb[IFLA_MASTER]) {
3424 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]), extack);
3426 goto out_unregister;
3435 LIST_HEAD(list_kill);
3437 ops->dellink(dev, &list_kill);
3438 unregister_netdevice_many(&list_kill);
3440 unregister_netdevice(dev);
3445 struct rtnl_newlink_tbs {
3446 struct nlattr *tb[IFLA_MAX + 1];
3447 struct nlattr *attr[RTNL_MAX_TYPE + 1];
3448 struct nlattr *slave_attr[RTNL_SLAVE_MAX_TYPE + 1];
3451 static int __rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3452 struct rtnl_newlink_tbs *tbs,
3453 struct netlink_ext_ack *extack)
3455 struct nlattr *linkinfo[IFLA_INFO_MAX + 1];
3456 struct nlattr ** const tb = tbs->tb;
3457 const struct rtnl_link_ops *m_ops;
3458 struct net_device *master_dev;
3459 struct net *net = sock_net(skb->sk);
3460 const struct rtnl_link_ops *ops;
3461 struct nlattr **slave_data;
3462 char kind[MODULE_NAME_LEN];
3463 struct net_device *dev;
3464 struct ifinfomsg *ifm;
3465 struct nlattr **data;
3466 bool link_specified;
3469 #ifdef CONFIG_MODULES
3472 err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3473 ifla_policy, extack);
3477 err = rtnl_ensure_unique_netns(tb, extack, false);
3481 ifm = nlmsg_data(nlh);
3482 if (ifm->ifi_index > 0) {
3483 link_specified = true;
3484 dev = __dev_get_by_index(net, ifm->ifi_index);
3485 } else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME]) {
3486 link_specified = true;
3487 dev = rtnl_dev_get(net, tb);
3489 link_specified = false;
3496 master_dev = netdev_master_upper_dev_get(dev);
3498 m_ops = master_dev->rtnl_link_ops;
3501 err = validate_linkmsg(dev, tb, extack);
3505 if (tb[IFLA_LINKINFO]) {
3506 err = nla_parse_nested_deprecated(linkinfo, IFLA_INFO_MAX,
3508 ifla_info_policy, NULL);
3512 memset(linkinfo, 0, sizeof(linkinfo));
3514 if (linkinfo[IFLA_INFO_KIND]) {
3515 nla_strscpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
3516 ops = rtnl_link_ops_get(kind);
3524 if (ops->maxtype > RTNL_MAX_TYPE)
3527 if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
3528 err = nla_parse_nested_deprecated(tbs->attr, ops->maxtype,
3529 linkinfo[IFLA_INFO_DATA],
3530 ops->policy, extack);
3535 if (ops->validate) {
3536 err = ops->validate(tb, data, extack);
3544 if (m_ops->slave_maxtype > RTNL_SLAVE_MAX_TYPE)
3547 if (m_ops->slave_maxtype &&
3548 linkinfo[IFLA_INFO_SLAVE_DATA]) {
3549 err = nla_parse_nested_deprecated(tbs->slave_attr,
3550 m_ops->slave_maxtype,
3551 linkinfo[IFLA_INFO_SLAVE_DATA],
3552 m_ops->slave_policy,
3556 slave_data = tbs->slave_attr;
3563 if (nlh->nlmsg_flags & NLM_F_EXCL)
3565 if (nlh->nlmsg_flags & NLM_F_REPLACE)
3568 if (linkinfo[IFLA_INFO_DATA]) {
3569 if (!ops || ops != dev->rtnl_link_ops ||
3573 err = ops->changelink(dev, tb, data, extack);
3576 status |= DO_SETLINK_NOTIFY;
3579 if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
3580 if (!m_ops || !m_ops->slave_changelink)
3583 err = m_ops->slave_changelink(master_dev, dev, tb,
3584 slave_data, extack);
3587 status |= DO_SETLINK_NOTIFY;
3590 return do_setlink(skb, dev, ifm, extack, tb, status);
3593 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
3594 /* No dev found and NLM_F_CREATE not set. Requested dev does not exist,
3595 * or it's for a group
3600 return rtnl_group_changelink(skb, net,
3601 nla_get_u32(tb[IFLA_GROUP]),
3606 if (tb[IFLA_MAP] || tb[IFLA_PROTINFO])
3610 #ifdef CONFIG_MODULES
3613 request_module("rtnl-link-%s", kind);
3615 ops = rtnl_link_ops_get(kind);
3620 NL_SET_ERR_MSG(extack, "Unknown device type");
3624 return rtnl_newlink_create(skb, ifm, ops, nlh, tb, data, extack);
3627 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3628 struct netlink_ext_ack *extack)
3630 struct rtnl_newlink_tbs *tbs;
3633 tbs = kmalloc(sizeof(*tbs), GFP_KERNEL);
3637 ret = __rtnl_newlink(skb, nlh, tbs, extack);
3642 static int rtnl_valid_getlink_req(struct sk_buff *skb,
3643 const struct nlmsghdr *nlh,
3645 struct netlink_ext_ack *extack)
3647 struct ifinfomsg *ifm;
3650 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
3651 NL_SET_ERR_MSG(extack, "Invalid header for get link");
3655 if (!netlink_strict_get_check(skb))
3656 return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3657 ifla_policy, extack);
3659 ifm = nlmsg_data(nlh);
3660 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
3662 NL_SET_ERR_MSG(extack, "Invalid values in header for get link request");
3666 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFLA_MAX,
3667 ifla_policy, extack);
3671 for (i = 0; i <= IFLA_MAX; i++) {
3677 case IFLA_ALT_IFNAME:
3679 case IFLA_TARGET_NETNSID:
3682 NL_SET_ERR_MSG(extack, "Unsupported attribute in get link request");
3690 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3691 struct netlink_ext_ack *extack)
3693 struct net *net = sock_net(skb->sk);
3694 struct net *tgt_net = net;
3695 struct ifinfomsg *ifm;
3696 struct nlattr *tb[IFLA_MAX+1];
3697 struct net_device *dev = NULL;
3698 struct sk_buff *nskb;
3701 u32 ext_filter_mask = 0;
3703 err = rtnl_valid_getlink_req(skb, nlh, tb, extack);
3707 err = rtnl_ensure_unique_netns(tb, extack, true);
3711 if (tb[IFLA_TARGET_NETNSID]) {
3712 netnsid = nla_get_s32(tb[IFLA_TARGET_NETNSID]);
3713 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(skb).sk, netnsid);
3714 if (IS_ERR(tgt_net))
3715 return PTR_ERR(tgt_net);
3718 if (tb[IFLA_EXT_MASK])
3719 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
3722 ifm = nlmsg_data(nlh);
3723 if (ifm->ifi_index > 0)
3724 dev = __dev_get_by_index(tgt_net, ifm->ifi_index);
3725 else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3726 dev = rtnl_dev_get(tgt_net, tb);
3735 nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
3739 err = rtnl_fill_ifinfo(nskb, dev, net,
3740 RTM_NEWLINK, NETLINK_CB(skb).portid,
3741 nlh->nlmsg_seq, 0, 0, ext_filter_mask,
3742 0, NULL, 0, netnsid, GFP_KERNEL);
3744 /* -EMSGSIZE implies BUG in if_nlmsg_size */
3745 WARN_ON(err == -EMSGSIZE);
3748 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
3756 static int rtnl_alt_ifname(int cmd, struct net_device *dev, struct nlattr *attr,
3757 bool *changed, struct netlink_ext_ack *extack)
3763 err = nla_validate(attr, attr->nla_len, IFLA_MAX, ifla_policy, extack);
3767 if (cmd == RTM_NEWLINKPROP) {
3768 size = rtnl_prop_list_size(dev);
3769 size += nla_total_size(ALTIFNAMSIZ);
3770 if (size >= U16_MAX) {
3771 NL_SET_ERR_MSG(extack,
3772 "effective property list too long");
3777 alt_ifname = nla_strdup(attr, GFP_KERNEL_ACCOUNT);
3781 if (cmd == RTM_NEWLINKPROP) {
3782 err = netdev_name_node_alt_create(dev, alt_ifname);
3785 } else if (cmd == RTM_DELLINKPROP) {
3786 err = netdev_name_node_alt_destroy(dev, alt_ifname);
3798 static int rtnl_linkprop(int cmd, struct sk_buff *skb, struct nlmsghdr *nlh,
3799 struct netlink_ext_ack *extack)
3801 struct net *net = sock_net(skb->sk);
3802 struct nlattr *tb[IFLA_MAX + 1];
3803 struct net_device *dev;
3804 struct ifinfomsg *ifm;
3805 bool changed = false;
3806 struct nlattr *attr;
3809 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
3813 err = rtnl_ensure_unique_netns(tb, extack, true);
3817 ifm = nlmsg_data(nlh);
3818 if (ifm->ifi_index > 0)
3819 dev = __dev_get_by_index(net, ifm->ifi_index);
3820 else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3821 dev = rtnl_dev_get(net, tb);
3828 if (!tb[IFLA_PROP_LIST])
3831 nla_for_each_nested(attr, tb[IFLA_PROP_LIST], rem) {
3832 switch (nla_type(attr)) {
3833 case IFLA_ALT_IFNAME:
3834 err = rtnl_alt_ifname(cmd, dev, attr, &changed, extack);
3842 netdev_state_change(dev);
3846 static int rtnl_newlinkprop(struct sk_buff *skb, struct nlmsghdr *nlh,
3847 struct netlink_ext_ack *extack)
3849 return rtnl_linkprop(RTM_NEWLINKPROP, skb, nlh, extack);
3852 static int rtnl_dellinkprop(struct sk_buff *skb, struct nlmsghdr *nlh,
3853 struct netlink_ext_ack *extack)
3855 return rtnl_linkprop(RTM_DELLINKPROP, skb, nlh, extack);
3858 static u32 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
3860 struct net *net = sock_net(skb->sk);
3861 size_t min_ifinfo_dump_size = 0;
3862 struct nlattr *tb[IFLA_MAX+1];
3863 u32 ext_filter_mask = 0;
3864 struct net_device *dev;
3867 /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
3868 hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
3869 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
3871 if (nlmsg_parse_deprecated(nlh, hdrlen, tb, IFLA_MAX, ifla_policy, NULL) >= 0) {
3872 if (tb[IFLA_EXT_MASK])
3873 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
3876 if (!ext_filter_mask)
3877 return NLMSG_GOODSIZE;
3879 * traverse the list of net devices and compute the minimum
3880 * buffer size based upon the filter mask.
3883 for_each_netdev_rcu(net, dev) {
3884 min_ifinfo_dump_size = max(min_ifinfo_dump_size,
3885 if_nlmsg_size(dev, ext_filter_mask));
3889 return nlmsg_total_size(min_ifinfo_dump_size);
3892 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
3895 int s_idx = cb->family;
3896 int type = cb->nlh->nlmsg_type - RTM_BASE;
3902 for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
3903 struct rtnl_link __rcu **tab;
3904 struct rtnl_link *link;
3905 rtnl_dumpit_func dumpit;
3907 if (idx < s_idx || idx == PF_PACKET)
3910 if (type < 0 || type >= RTM_NR_MSGTYPES)
3913 tab = rcu_dereference_rtnl(rtnl_msg_handlers[idx]);
3917 link = rcu_dereference_rtnl(tab[type]);
3921 dumpit = link->dumpit;
3926 memset(&cb->args[0], 0, sizeof(cb->args));
3930 ret = dumpit(skb, cb);
3936 return skb->len ? : ret;
3939 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
3940 unsigned int change,
3941 u32 event, gfp_t flags, int *new_nsid,
3942 int new_ifindex, u32 portid, u32 seq)
3944 struct net *net = dev_net(dev);
3945 struct sk_buff *skb;
3948 skb = nlmsg_new(if_nlmsg_size(dev, 0), flags);
3952 err = rtnl_fill_ifinfo(skb, dev, dev_net(dev),
3953 type, portid, seq, change, 0, 0, event,
3954 new_nsid, new_ifindex, -1, flags);
3956 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
3957 WARN_ON(err == -EMSGSIZE);
3964 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
3968 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags,
3969 u32 portid, const struct nlmsghdr *nlh)
3971 struct net *net = dev_net(dev);
3973 rtnl_notify(skb, net, portid, RTNLGRP_LINK, nlh, flags);
3976 static void rtmsg_ifinfo_event(int type, struct net_device *dev,
3977 unsigned int change, u32 event,
3978 gfp_t flags, int *new_nsid, int new_ifindex,
3979 u32 portid, const struct nlmsghdr *nlh)
3981 struct sk_buff *skb;
3983 if (dev->reg_state != NETREG_REGISTERED)
3986 skb = rtmsg_ifinfo_build_skb(type, dev, change, event, flags, new_nsid,
3987 new_ifindex, portid, nlmsg_seq(nlh));
3989 rtmsg_ifinfo_send(skb, dev, flags, portid, nlh);
3992 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
3993 gfp_t flags, u32 portid, const struct nlmsghdr *nlh)
3995 rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
3996 NULL, 0, portid, nlh);
3999 void rtmsg_ifinfo_newnet(int type, struct net_device *dev, unsigned int change,
4000 gfp_t flags, int *new_nsid, int new_ifindex)
4002 rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
4003 new_nsid, new_ifindex, 0, NULL);
4006 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
4007 struct net_device *dev,
4008 u8 *addr, u16 vid, u32 pid, u32 seq,
4009 int type, unsigned int flags,
4010 int nlflags, u16 ndm_state)
4012 struct nlmsghdr *nlh;
4015 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
4019 ndm = nlmsg_data(nlh);
4020 ndm->ndm_family = AF_BRIDGE;
4023 ndm->ndm_flags = flags;
4025 ndm->ndm_ifindex = dev->ifindex;
4026 ndm->ndm_state = ndm_state;
4028 if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
4029 goto nla_put_failure;
4031 if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
4032 goto nla_put_failure;
4034 nlmsg_end(skb, nlh);
4038 nlmsg_cancel(skb, nlh);
4042 static inline size_t rtnl_fdb_nlmsg_size(void)
4044 return NLMSG_ALIGN(sizeof(struct ndmsg)) +
4045 nla_total_size(ETH_ALEN) + /* NDA_LLADDR */
4046 nla_total_size(sizeof(u16)) + /* NDA_VLAN */
4050 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type,
4053 struct net *net = dev_net(dev);
4054 struct sk_buff *skb;
4057 skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
4061 err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
4062 0, 0, type, NTF_SELF, 0, ndm_state);
4068 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
4071 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
4075 * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
4077 int ndo_dflt_fdb_add(struct ndmsg *ndm,
4078 struct nlattr *tb[],
4079 struct net_device *dev,
4080 const unsigned char *addr, u16 vid,
4085 /* If aging addresses are supported device will need to
4086 * implement its own handler for this.
4088 if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
4089 netdev_info(dev, "default FDB implementation only supports local addresses\n");
4093 if (tb[NDA_FLAGS_EXT]) {
4094 netdev_info(dev, "invalid flags given to default FDB implementation\n");
4099 netdev_info(dev, "vlans aren't supported yet for dev_uc|mc_add()\n");
4103 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
4104 err = dev_uc_add_excl(dev, addr);
4105 else if (is_multicast_ether_addr(addr))
4106 err = dev_mc_add_excl(dev, addr);
4108 /* Only return duplicate errors if NLM_F_EXCL is set */
4109 if (err == -EEXIST && !(flags & NLM_F_EXCL))
4114 EXPORT_SYMBOL(ndo_dflt_fdb_add);
4116 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid,
4117 struct netlink_ext_ack *extack)
4122 if (nla_len(vlan_attr) != sizeof(u16)) {
4123 NL_SET_ERR_MSG(extack, "invalid vlan attribute size");
4127 vid = nla_get_u16(vlan_attr);
4129 if (!vid || vid >= VLAN_VID_MASK) {
4130 NL_SET_ERR_MSG(extack, "invalid vlan id");
4138 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh,
4139 struct netlink_ext_ack *extack)
4141 struct net *net = sock_net(skb->sk);
4143 struct nlattr *tb[NDA_MAX+1];
4144 struct net_device *dev;
4149 err = nlmsg_parse_deprecated(nlh, sizeof(*ndm), tb, NDA_MAX, NULL,
4154 ndm = nlmsg_data(nlh);
4155 if (ndm->ndm_ifindex == 0) {
4156 NL_SET_ERR_MSG(extack, "invalid ifindex");
4160 dev = __dev_get_by_index(net, ndm->ndm_ifindex);
4162 NL_SET_ERR_MSG(extack, "unknown ifindex");
4166 if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
4167 NL_SET_ERR_MSG(extack, "invalid address");
4171 if (dev->type != ARPHRD_ETHER) {
4172 NL_SET_ERR_MSG(extack, "FDB add only supported for Ethernet devices");
4176 addr = nla_data(tb[NDA_LLADDR]);
4178 err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
4184 /* Support fdb on master device the net/bridge default case */
4185 if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
4186 netif_is_bridge_port(dev)) {
4187 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4188 const struct net_device_ops *ops = br_dev->netdev_ops;
4190 err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
4191 nlh->nlmsg_flags, extack);
4195 ndm->ndm_flags &= ~NTF_MASTER;
4198 /* Embedded bridge, macvlan, and any other device support */
4199 if ((ndm->ndm_flags & NTF_SELF)) {
4200 if (dev->netdev_ops->ndo_fdb_add)
4201 err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
4206 err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
4210 rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH,
4212 ndm->ndm_flags &= ~NTF_SELF;
4220 * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
4222 int ndo_dflt_fdb_del(struct ndmsg *ndm,
4223 struct nlattr *tb[],
4224 struct net_device *dev,
4225 const unsigned char *addr, u16 vid)
4229 /* If aging addresses are supported device will need to
4230 * implement its own handler for this.
4232 if (!(ndm->ndm_state & NUD_PERMANENT)) {
4233 netdev_info(dev, "default FDB implementation only supports local addresses\n");
4237 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
4238 err = dev_uc_del(dev, addr);
4239 else if (is_multicast_ether_addr(addr))
4240 err = dev_mc_del(dev, addr);
4244 EXPORT_SYMBOL(ndo_dflt_fdb_del);
4246 static const struct nla_policy fdb_del_bulk_policy[NDA_MAX + 1] = {
4247 [NDA_VLAN] = { .type = NLA_U16 },
4248 [NDA_IFINDEX] = NLA_POLICY_MIN(NLA_S32, 1),
4249 [NDA_NDM_STATE_MASK] = { .type = NLA_U16 },
4250 [NDA_NDM_FLAGS_MASK] = { .type = NLA_U8 },
4253 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh,
4254 struct netlink_ext_ack *extack)
4256 bool del_bulk = !!(nlh->nlmsg_flags & NLM_F_BULK);
4257 struct net *net = sock_net(skb->sk);
4258 const struct net_device_ops *ops;
4260 struct nlattr *tb[NDA_MAX+1];
4261 struct net_device *dev;
4266 if (!netlink_capable(skb, CAP_NET_ADMIN))
4270 err = nlmsg_parse_deprecated(nlh, sizeof(*ndm), tb, NDA_MAX,
4273 err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX,
4274 fdb_del_bulk_policy, extack);
4279 ndm = nlmsg_data(nlh);
4280 if (ndm->ndm_ifindex == 0) {
4281 NL_SET_ERR_MSG(extack, "invalid ifindex");
4285 dev = __dev_get_by_index(net, ndm->ndm_ifindex);
4287 NL_SET_ERR_MSG(extack, "unknown ifindex");
4292 if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
4293 NL_SET_ERR_MSG(extack, "invalid address");
4296 addr = nla_data(tb[NDA_LLADDR]);
4299 if (dev->type != ARPHRD_ETHER) {
4300 NL_SET_ERR_MSG(extack, "FDB delete only supported for Ethernet devices");
4304 err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
4310 /* Support fdb on master device the net/bridge default case */
4311 if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
4312 netif_is_bridge_port(dev)) {
4313 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4315 ops = br_dev->netdev_ops;
4317 if (ops->ndo_fdb_del)
4318 err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid, extack);
4320 if (ops->ndo_fdb_del_bulk)
4321 err = ops->ndo_fdb_del_bulk(ndm, tb, dev, vid,
4328 ndm->ndm_flags &= ~NTF_MASTER;
4331 /* Embedded bridge, macvlan, and any other device support */
4332 if (ndm->ndm_flags & NTF_SELF) {
4333 ops = dev->netdev_ops;
4335 if (ops->ndo_fdb_del)
4336 err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid, extack);
4338 err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
4340 /* in case err was cleared by NTF_MASTER call */
4342 if (ops->ndo_fdb_del_bulk)
4343 err = ops->ndo_fdb_del_bulk(ndm, tb, dev, vid,
4349 rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH,
4351 ndm->ndm_flags &= ~NTF_SELF;
4358 static int nlmsg_populate_fdb(struct sk_buff *skb,
4359 struct netlink_callback *cb,
4360 struct net_device *dev,
4362 struct netdev_hw_addr_list *list)
4364 struct netdev_hw_addr *ha;
4368 portid = NETLINK_CB(cb->skb).portid;
4369 seq = cb->nlh->nlmsg_seq;
4371 list_for_each_entry(ha, &list->list, list) {
4372 if (*idx < cb->args[2])
4375 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
4377 RTM_NEWNEIGH, NTF_SELF,
4378 NLM_F_MULTI, NUD_PERMANENT);
4388 * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
4389 * @skb: socket buffer to store message in
4390 * @cb: netlink callback
4392 * @filter_dev: ignored
4393 * @idx: the number of FDB table entries dumped is added to *@idx
4395 * Default netdevice operation to dump the existing unicast address list.
4396 * Returns number of addresses from list put in skb.
4398 int ndo_dflt_fdb_dump(struct sk_buff *skb,
4399 struct netlink_callback *cb,
4400 struct net_device *dev,
4401 struct net_device *filter_dev,
4406 if (dev->type != ARPHRD_ETHER)
4409 netif_addr_lock_bh(dev);
4410 err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->uc);
4413 err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->mc);
4415 netif_addr_unlock_bh(dev);
4418 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
4420 static int valid_fdb_dump_strict(const struct nlmsghdr *nlh,
4421 int *br_idx, int *brport_idx,
4422 struct netlink_ext_ack *extack)
4424 struct nlattr *tb[NDA_MAX + 1];
4428 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ndm))) {
4429 NL_SET_ERR_MSG(extack, "Invalid header for fdb dump request");
4433 ndm = nlmsg_data(nlh);
4434 if (ndm->ndm_pad1 || ndm->ndm_pad2 || ndm->ndm_state ||
4435 ndm->ndm_flags || ndm->ndm_type) {
4436 NL_SET_ERR_MSG(extack, "Invalid values in header for fdb dump request");
4440 err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct ndmsg), tb,
4441 NDA_MAX, NULL, extack);
4445 *brport_idx = ndm->ndm_ifindex;
4446 for (i = 0; i <= NDA_MAX; ++i) {
4452 if (nla_len(tb[i]) != sizeof(u32)) {
4453 NL_SET_ERR_MSG(extack, "Invalid IFINDEX attribute in fdb dump request");
4456 *brport_idx = nla_get_u32(tb[NDA_IFINDEX]);
4459 if (nla_len(tb[i]) != sizeof(u32)) {
4460 NL_SET_ERR_MSG(extack, "Invalid MASTER attribute in fdb dump request");
4463 *br_idx = nla_get_u32(tb[NDA_MASTER]);
4466 NL_SET_ERR_MSG(extack, "Unsupported attribute in fdb dump request");
4474 static int valid_fdb_dump_legacy(const struct nlmsghdr *nlh,
4475 int *br_idx, int *brport_idx,
4476 struct netlink_ext_ack *extack)
4478 struct nlattr *tb[IFLA_MAX+1];
4481 /* A hack to preserve kernel<->userspace interface.
4482 * Before Linux v4.12 this code accepted ndmsg since iproute2 v3.3.0.
4483 * However, ndmsg is shorter than ifinfomsg thus nlmsg_parse() bails.
4484 * So, check for ndmsg with an optional u32 attribute (not used here).
4485 * Fortunately these sizes don't conflict with the size of ifinfomsg
4486 * with an optional attribute.
4488 if (nlmsg_len(nlh) != sizeof(struct ndmsg) &&
4489 (nlmsg_len(nlh) != sizeof(struct ndmsg) +
4490 nla_attr_size(sizeof(u32)))) {
4491 struct ifinfomsg *ifm;
4493 err = nlmsg_parse_deprecated(nlh, sizeof(struct ifinfomsg),
4494 tb, IFLA_MAX, ifla_policy,
4498 } else if (err == 0) {
4499 if (tb[IFLA_MASTER])
4500 *br_idx = nla_get_u32(tb[IFLA_MASTER]);
4503 ifm = nlmsg_data(nlh);
4504 *brport_idx = ifm->ifi_index;
4509 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
4511 struct net_device *dev;
4512 struct net_device *br_dev = NULL;
4513 const struct net_device_ops *ops = NULL;
4514 const struct net_device_ops *cops = NULL;
4515 struct net *net = sock_net(skb->sk);
4516 struct hlist_head *head;
4524 if (cb->strict_check)
4525 err = valid_fdb_dump_strict(cb->nlh, &br_idx, &brport_idx,
4528 err = valid_fdb_dump_legacy(cb->nlh, &br_idx, &brport_idx,
4534 br_dev = __dev_get_by_index(net, br_idx);
4538 ops = br_dev->netdev_ops;
4542 s_idx = cb->args[1];
4544 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4546 head = &net->dev_index_head[h];
4547 hlist_for_each_entry(dev, head, index_hlist) {
4549 if (brport_idx && (dev->ifindex != brport_idx))
4552 if (!br_idx) { /* user did not specify a specific bridge */
4553 if (netif_is_bridge_port(dev)) {
4554 br_dev = netdev_master_upper_dev_get(dev);
4555 cops = br_dev->netdev_ops;
4558 if (dev != br_dev &&
4559 !netif_is_bridge_port(dev))
4562 if (br_dev != netdev_master_upper_dev_get(dev) &&
4563 !netif_is_bridge_master(dev))
4571 if (netif_is_bridge_port(dev)) {
4572 if (cops && cops->ndo_fdb_dump) {
4573 err = cops->ndo_fdb_dump(skb, cb,
4576 if (err == -EMSGSIZE)
4581 if (dev->netdev_ops->ndo_fdb_dump)
4582 err = dev->netdev_ops->ndo_fdb_dump(skb, cb,
4586 err = ndo_dflt_fdb_dump(skb, cb, dev, NULL,
4588 if (err == -EMSGSIZE)
4593 /* reset fdb offset to 0 for rest of the interfaces */
4609 static int valid_fdb_get_strict(const struct nlmsghdr *nlh,
4610 struct nlattr **tb, u8 *ndm_flags,
4611 int *br_idx, int *brport_idx, u8 **addr,
4612 u16 *vid, struct netlink_ext_ack *extack)
4617 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ndm))) {
4618 NL_SET_ERR_MSG(extack, "Invalid header for fdb get request");
4622 ndm = nlmsg_data(nlh);
4623 if (ndm->ndm_pad1 || ndm->ndm_pad2 || ndm->ndm_state ||
4625 NL_SET_ERR_MSG(extack, "Invalid values in header for fdb get request");
4629 if (ndm->ndm_flags & ~(NTF_MASTER | NTF_SELF)) {
4630 NL_SET_ERR_MSG(extack, "Invalid flags in header for fdb get request");
4634 err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct ndmsg), tb,
4635 NDA_MAX, nda_policy, extack);
4639 *ndm_flags = ndm->ndm_flags;
4640 *brport_idx = ndm->ndm_ifindex;
4641 for (i = 0; i <= NDA_MAX; ++i) {
4647 *br_idx = nla_get_u32(tb[i]);
4650 if (nla_len(tb[i]) != ETH_ALEN) {
4651 NL_SET_ERR_MSG(extack, "Invalid address in fdb get request");
4654 *addr = nla_data(tb[i]);
4657 err = fdb_vid_parse(tb[i], vid, extack);
4664 NL_SET_ERR_MSG(extack, "Unsupported attribute in fdb get request");
4672 static int rtnl_fdb_get(struct sk_buff *in_skb, struct nlmsghdr *nlh,
4673 struct netlink_ext_ack *extack)
4675 struct net_device *dev = NULL, *br_dev = NULL;
4676 const struct net_device_ops *ops = NULL;
4677 struct net *net = sock_net(in_skb->sk);
4678 struct nlattr *tb[NDA_MAX + 1];
4679 struct sk_buff *skb;
4687 err = valid_fdb_get_strict(nlh, tb, &ndm_flags, &br_idx,
4688 &brport_idx, &addr, &vid, extack);
4693 NL_SET_ERR_MSG(extack, "Missing lookup address for fdb get request");
4698 dev = __dev_get_by_index(net, brport_idx);
4700 NL_SET_ERR_MSG(extack, "Unknown device ifindex");
4707 NL_SET_ERR_MSG(extack, "Master and device are mutually exclusive");
4711 br_dev = __dev_get_by_index(net, br_idx);
4713 NL_SET_ERR_MSG(extack, "Invalid master ifindex");
4716 ops = br_dev->netdev_ops;
4720 if (!ndm_flags || (ndm_flags & NTF_MASTER)) {
4721 if (!netif_is_bridge_port(dev)) {
4722 NL_SET_ERR_MSG(extack, "Device is not a bridge port");
4725 br_dev = netdev_master_upper_dev_get(dev);
4727 NL_SET_ERR_MSG(extack, "Master of device not found");
4730 ops = br_dev->netdev_ops;
4732 if (!(ndm_flags & NTF_SELF)) {
4733 NL_SET_ERR_MSG(extack, "Missing NTF_SELF");
4736 ops = dev->netdev_ops;
4740 if (!br_dev && !dev) {
4741 NL_SET_ERR_MSG(extack, "No device specified");
4745 if (!ops || !ops->ndo_fdb_get) {
4746 NL_SET_ERR_MSG(extack, "Fdb get operation not supported by device");
4750 skb = nlmsg_new(NLMSG_GOODSIZE, GFP_KERNEL);
4756 err = ops->ndo_fdb_get(skb, tb, dev, addr, vid,
4757 NETLINK_CB(in_skb).portid,
4758 nlh->nlmsg_seq, extack);
4762 return rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4768 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
4769 unsigned int attrnum, unsigned int flag)
4772 return nla_put_u8(skb, attrnum, !!(flags & flag));
4776 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
4777 struct net_device *dev, u16 mode,
4778 u32 flags, u32 mask, int nlflags,
4780 int (*vlan_fill)(struct sk_buff *skb,
4781 struct net_device *dev,
4784 struct nlmsghdr *nlh;
4785 struct ifinfomsg *ifm;
4786 struct nlattr *br_afspec;
4787 struct nlattr *protinfo;
4788 u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
4789 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4792 nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
4796 ifm = nlmsg_data(nlh);
4797 ifm->ifi_family = AF_BRIDGE;
4799 ifm->ifi_type = dev->type;
4800 ifm->ifi_index = dev->ifindex;
4801 ifm->ifi_flags = dev_get_flags(dev);
4802 ifm->ifi_change = 0;
4805 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4806 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4807 nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
4809 nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
4811 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4812 (dev->ifindex != dev_get_iflink(dev) &&
4813 nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
4814 goto nla_put_failure;
4816 br_afspec = nla_nest_start_noflag(skb, IFLA_AF_SPEC);
4818 goto nla_put_failure;
4820 if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
4821 nla_nest_cancel(skb, br_afspec);
4822 goto nla_put_failure;
4825 if (mode != BRIDGE_MODE_UNDEF) {
4826 if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
4827 nla_nest_cancel(skb, br_afspec);
4828 goto nla_put_failure;
4832 err = vlan_fill(skb, dev, filter_mask);
4834 nla_nest_cancel(skb, br_afspec);
4835 goto nla_put_failure;
4838 nla_nest_end(skb, br_afspec);
4840 protinfo = nla_nest_start(skb, IFLA_PROTINFO);
4842 goto nla_put_failure;
4844 if (brport_nla_put_flag(skb, flags, mask,
4845 IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
4846 brport_nla_put_flag(skb, flags, mask,
4847 IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
4848 brport_nla_put_flag(skb, flags, mask,
4849 IFLA_BRPORT_FAST_LEAVE,
4850 BR_MULTICAST_FAST_LEAVE) ||
4851 brport_nla_put_flag(skb, flags, mask,
4852 IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
4853 brport_nla_put_flag(skb, flags, mask,
4854 IFLA_BRPORT_LEARNING, BR_LEARNING) ||
4855 brport_nla_put_flag(skb, flags, mask,
4856 IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
4857 brport_nla_put_flag(skb, flags, mask,
4858 IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
4859 brport_nla_put_flag(skb, flags, mask,
4860 IFLA_BRPORT_PROXYARP, BR_PROXYARP) ||
4861 brport_nla_put_flag(skb, flags, mask,
4862 IFLA_BRPORT_MCAST_FLOOD, BR_MCAST_FLOOD) ||
4863 brport_nla_put_flag(skb, flags, mask,
4864 IFLA_BRPORT_BCAST_FLOOD, BR_BCAST_FLOOD)) {
4865 nla_nest_cancel(skb, protinfo);
4866 goto nla_put_failure;
4869 nla_nest_end(skb, protinfo);
4871 nlmsg_end(skb, nlh);
4874 nlmsg_cancel(skb, nlh);
4875 return err ? err : -EMSGSIZE;
4877 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink);
4879 static int valid_bridge_getlink_req(const struct nlmsghdr *nlh,
4880 bool strict_check, u32 *filter_mask,
4881 struct netlink_ext_ack *extack)
4883 struct nlattr *tb[IFLA_MAX+1];
4887 struct ifinfomsg *ifm;
4889 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
4890 NL_SET_ERR_MSG(extack, "Invalid header for bridge link dump");
4894 ifm = nlmsg_data(nlh);
4895 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
4896 ifm->ifi_change || ifm->ifi_index) {
4897 NL_SET_ERR_MSG(extack, "Invalid values in header for bridge link dump request");
4901 err = nlmsg_parse_deprecated_strict(nlh,
4902 sizeof(struct ifinfomsg),
4903 tb, IFLA_MAX, ifla_policy,
4906 err = nlmsg_parse_deprecated(nlh, sizeof(struct ifinfomsg),
4907 tb, IFLA_MAX, ifla_policy,
4913 /* new attributes should only be added with strict checking */
4914 for (i = 0; i <= IFLA_MAX; ++i) {
4920 *filter_mask = nla_get_u32(tb[i]);
4924 NL_SET_ERR_MSG(extack, "Unsupported attribute in bridge link dump request");
4933 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
4935 const struct nlmsghdr *nlh = cb->nlh;
4936 struct net *net = sock_net(skb->sk);
4937 struct net_device *dev;
4939 u32 portid = NETLINK_CB(cb->skb).portid;
4940 u32 seq = nlh->nlmsg_seq;
4941 u32 filter_mask = 0;
4944 err = valid_bridge_getlink_req(nlh, cb->strict_check, &filter_mask,
4946 if (err < 0 && cb->strict_check)
4950 for_each_netdev_rcu(net, dev) {
4951 const struct net_device_ops *ops = dev->netdev_ops;
4952 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4954 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
4955 if (idx >= cb->args[0]) {
4956 err = br_dev->netdev_ops->ndo_bridge_getlink(
4957 skb, portid, seq, dev,
4958 filter_mask, NLM_F_MULTI);
4959 if (err < 0 && err != -EOPNOTSUPP) {
4960 if (likely(skb->len))
4969 if (ops->ndo_bridge_getlink) {
4970 if (idx >= cb->args[0]) {
4971 err = ops->ndo_bridge_getlink(skb, portid,
4975 if (err < 0 && err != -EOPNOTSUPP) {
4976 if (likely(skb->len))
4993 static inline size_t bridge_nlmsg_size(void)
4995 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4996 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
4997 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
4998 + nla_total_size(sizeof(u32)) /* IFLA_MASTER */
4999 + nla_total_size(sizeof(u32)) /* IFLA_MTU */
5000 + nla_total_size(sizeof(u32)) /* IFLA_LINK */
5001 + nla_total_size(sizeof(u32)) /* IFLA_OPERSTATE */
5002 + nla_total_size(sizeof(u8)) /* IFLA_PROTINFO */
5003 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
5004 + nla_total_size(sizeof(u16)) /* IFLA_BRIDGE_FLAGS */
5005 + nla_total_size(sizeof(u16)); /* IFLA_BRIDGE_MODE */
5008 static int rtnl_bridge_notify(struct net_device *dev)
5010 struct net *net = dev_net(dev);
5011 struct sk_buff *skb;
5012 int err = -EOPNOTSUPP;
5014 if (!dev->netdev_ops->ndo_bridge_getlink)
5017 skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
5023 err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
5027 /* Notification info is only filled for bridge ports, not the bridge
5028 * device itself. Therefore, a zero notification length is valid and
5029 * should not result in an error.
5034 rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
5037 WARN_ON(err == -EMSGSIZE);
5040 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
5044 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
5045 struct netlink_ext_ack *extack)
5047 struct net *net = sock_net(skb->sk);
5048 struct ifinfomsg *ifm;
5049 struct net_device *dev;
5050 struct nlattr *br_spec, *attr = NULL;
5051 int rem, err = -EOPNOTSUPP;
5053 bool have_flags = false;
5055 if (nlmsg_len(nlh) < sizeof(*ifm))
5058 ifm = nlmsg_data(nlh);
5059 if (ifm->ifi_family != AF_BRIDGE)
5060 return -EPFNOSUPPORT;
5062 dev = __dev_get_by_index(net, ifm->ifi_index);
5064 NL_SET_ERR_MSG(extack, "unknown ifindex");
5068 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
5070 nla_for_each_nested(attr, br_spec, rem) {
5071 if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
5072 if (nla_len(attr) < sizeof(flags))
5076 flags = nla_get_u16(attr);
5082 if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
5083 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
5085 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
5090 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags,
5095 flags &= ~BRIDGE_FLAGS_MASTER;
5098 if ((flags & BRIDGE_FLAGS_SELF)) {
5099 if (!dev->netdev_ops->ndo_bridge_setlink)
5102 err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
5106 flags &= ~BRIDGE_FLAGS_SELF;
5108 /* Generate event to notify upper layer of bridge
5111 err = rtnl_bridge_notify(dev);
5116 memcpy(nla_data(attr), &flags, sizeof(flags));
5121 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
5122 struct netlink_ext_ack *extack)
5124 struct net *net = sock_net(skb->sk);
5125 struct ifinfomsg *ifm;
5126 struct net_device *dev;
5127 struct nlattr *br_spec, *attr = NULL;
5128 int rem, err = -EOPNOTSUPP;
5130 bool have_flags = false;
5132 if (nlmsg_len(nlh) < sizeof(*ifm))
5135 ifm = nlmsg_data(nlh);
5136 if (ifm->ifi_family != AF_BRIDGE)
5137 return -EPFNOSUPPORT;
5139 dev = __dev_get_by_index(net, ifm->ifi_index);
5141 NL_SET_ERR_MSG(extack, "unknown ifindex");
5145 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
5147 nla_for_each_nested(attr, br_spec, rem) {
5148 if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
5149 if (nla_len(attr) < sizeof(flags))
5153 flags = nla_get_u16(attr);
5159 if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
5160 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
5162 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
5167 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
5171 flags &= ~BRIDGE_FLAGS_MASTER;
5174 if ((flags & BRIDGE_FLAGS_SELF)) {
5175 if (!dev->netdev_ops->ndo_bridge_dellink)
5178 err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
5182 flags &= ~BRIDGE_FLAGS_SELF;
5184 /* Generate event to notify upper layer of bridge
5187 err = rtnl_bridge_notify(dev);
5192 memcpy(nla_data(attr), &flags, sizeof(flags));
5197 static bool stats_attr_valid(unsigned int mask, int attrid, int idxattr)
5199 return (mask & IFLA_STATS_FILTER_BIT(attrid)) &&
5200 (!idxattr || idxattr == attrid);
5204 rtnl_offload_xstats_have_ndo(const struct net_device *dev, int attr_id)
5206 return dev->netdev_ops &&
5207 dev->netdev_ops->ndo_has_offload_stats &&
5208 dev->netdev_ops->ndo_get_offload_stats &&
5209 dev->netdev_ops->ndo_has_offload_stats(dev, attr_id);
5213 rtnl_offload_xstats_get_size_ndo(const struct net_device *dev, int attr_id)
5215 return rtnl_offload_xstats_have_ndo(dev, attr_id) ?
5216 sizeof(struct rtnl_link_stats64) : 0;
5220 rtnl_offload_xstats_fill_ndo(struct net_device *dev, int attr_id,
5221 struct sk_buff *skb)
5223 unsigned int size = rtnl_offload_xstats_get_size_ndo(dev, attr_id);
5224 struct nlattr *attr = NULL;
5231 attr = nla_reserve_64bit(skb, attr_id, size,
5232 IFLA_OFFLOAD_XSTATS_UNSPEC);
5236 attr_data = nla_data(attr);
5237 memset(attr_data, 0, size);
5239 err = dev->netdev_ops->ndo_get_offload_stats(attr_id, dev, attr_data);
5247 rtnl_offload_xstats_get_size_stats(const struct net_device *dev,
5248 enum netdev_offload_xstats_type type)
5250 bool enabled = netdev_offload_xstats_enabled(dev, type);
5252 return enabled ? sizeof(struct rtnl_hw_stats64) : 0;
5255 struct rtnl_offload_xstats_request_used {
5261 rtnl_offload_xstats_get_stats(struct net_device *dev,
5262 enum netdev_offload_xstats_type type,
5263 struct rtnl_offload_xstats_request_used *ru,
5264 struct rtnl_hw_stats64 *stats,
5265 struct netlink_ext_ack *extack)
5271 request = netdev_offload_xstats_enabled(dev, type);
5277 err = netdev_offload_xstats_get(dev, type, stats, &used, extack);
5283 ru->request = request;
5290 rtnl_offload_xstats_fill_hw_s_info_one(struct sk_buff *skb, int attr_id,
5291 struct rtnl_offload_xstats_request_used *ru)
5293 struct nlattr *nest;
5295 nest = nla_nest_start(skb, attr_id);
5299 if (nla_put_u8(skb, IFLA_OFFLOAD_XSTATS_HW_S_INFO_REQUEST, ru->request))
5300 goto nla_put_failure;
5302 if (nla_put_u8(skb, IFLA_OFFLOAD_XSTATS_HW_S_INFO_USED, ru->used))
5303 goto nla_put_failure;
5305 nla_nest_end(skb, nest);
5309 nla_nest_cancel(skb, nest);
5314 rtnl_offload_xstats_fill_hw_s_info(struct sk_buff *skb, struct net_device *dev,
5315 struct netlink_ext_ack *extack)
5317 enum netdev_offload_xstats_type t_l3 = NETDEV_OFFLOAD_XSTATS_TYPE_L3;
5318 struct rtnl_offload_xstats_request_used ru_l3;
5319 struct nlattr *nest;
5322 err = rtnl_offload_xstats_get_stats(dev, t_l3, &ru_l3, NULL, extack);
5326 nest = nla_nest_start(skb, IFLA_OFFLOAD_XSTATS_HW_S_INFO);
5330 if (rtnl_offload_xstats_fill_hw_s_info_one(skb,
5331 IFLA_OFFLOAD_XSTATS_L3_STATS,
5333 goto nla_put_failure;
5335 nla_nest_end(skb, nest);
5339 nla_nest_cancel(skb, nest);
5343 static int rtnl_offload_xstats_fill(struct sk_buff *skb, struct net_device *dev,
5344 int *prividx, u32 off_filter_mask,
5345 struct netlink_ext_ack *extack)
5347 enum netdev_offload_xstats_type t_l3 = NETDEV_OFFLOAD_XSTATS_TYPE_L3;
5348 int attr_id_hw_s_info = IFLA_OFFLOAD_XSTATS_HW_S_INFO;
5349 int attr_id_l3_stats = IFLA_OFFLOAD_XSTATS_L3_STATS;
5350 int attr_id_cpu_hit = IFLA_OFFLOAD_XSTATS_CPU_HIT;
5351 bool have_data = false;
5354 if (*prividx <= attr_id_cpu_hit &&
5356 IFLA_STATS_FILTER_BIT(attr_id_cpu_hit))) {
5357 err = rtnl_offload_xstats_fill_ndo(dev, attr_id_cpu_hit, skb);
5360 } else if (err != -ENODATA) {
5361 *prividx = attr_id_cpu_hit;
5366 if (*prividx <= attr_id_hw_s_info &&
5367 (off_filter_mask & IFLA_STATS_FILTER_BIT(attr_id_hw_s_info))) {
5368 *prividx = attr_id_hw_s_info;
5370 err = rtnl_offload_xstats_fill_hw_s_info(skb, dev, extack);
5378 if (*prividx <= attr_id_l3_stats &&
5379 (off_filter_mask & IFLA_STATS_FILTER_BIT(attr_id_l3_stats))) {
5380 unsigned int size_l3;
5381 struct nlattr *attr;
5383 *prividx = attr_id_l3_stats;
5385 size_l3 = rtnl_offload_xstats_get_size_stats(dev, t_l3);
5388 attr = nla_reserve_64bit(skb, attr_id_l3_stats, size_l3,
5389 IFLA_OFFLOAD_XSTATS_UNSPEC);
5393 err = rtnl_offload_xstats_get_stats(dev, t_l3, NULL,
5394 nla_data(attr), extack);
5411 rtnl_offload_xstats_get_size_hw_s_info_one(const struct net_device *dev,
5412 enum netdev_offload_xstats_type type)
5414 bool enabled = netdev_offload_xstats_enabled(dev, type);
5416 return nla_total_size(0) +
5417 /* IFLA_OFFLOAD_XSTATS_HW_S_INFO_REQUEST */
5418 nla_total_size(sizeof(u8)) +
5419 /* IFLA_OFFLOAD_XSTATS_HW_S_INFO_USED */
5420 (enabled ? nla_total_size(sizeof(u8)) : 0) +
5425 rtnl_offload_xstats_get_size_hw_s_info(const struct net_device *dev)
5427 enum netdev_offload_xstats_type t_l3 = NETDEV_OFFLOAD_XSTATS_TYPE_L3;
5429 return nla_total_size(0) +
5430 /* IFLA_OFFLOAD_XSTATS_L3_STATS */
5431 rtnl_offload_xstats_get_size_hw_s_info_one(dev, t_l3) +
5435 static int rtnl_offload_xstats_get_size(const struct net_device *dev,
5436 u32 off_filter_mask)
5438 enum netdev_offload_xstats_type t_l3 = NETDEV_OFFLOAD_XSTATS_TYPE_L3;
5439 int attr_id_cpu_hit = IFLA_OFFLOAD_XSTATS_CPU_HIT;
5443 if (off_filter_mask &
5444 IFLA_STATS_FILTER_BIT(attr_id_cpu_hit)) {
5445 size = rtnl_offload_xstats_get_size_ndo(dev, attr_id_cpu_hit);
5446 nla_size += nla_total_size_64bit(size);
5449 if (off_filter_mask &
5450 IFLA_STATS_FILTER_BIT(IFLA_OFFLOAD_XSTATS_HW_S_INFO))
5451 nla_size += rtnl_offload_xstats_get_size_hw_s_info(dev);
5453 if (off_filter_mask &
5454 IFLA_STATS_FILTER_BIT(IFLA_OFFLOAD_XSTATS_L3_STATS)) {
5455 size = rtnl_offload_xstats_get_size_stats(dev, t_l3);
5456 nla_size += nla_total_size_64bit(size);
5460 nla_size += nla_total_size(0);
5465 struct rtnl_stats_dump_filters {
5466 /* mask[0] filters outer attributes. Then individual nests have their
5467 * filtering mask at the index of the nested attribute.
5469 u32 mask[IFLA_STATS_MAX + 1];
5472 static int rtnl_fill_statsinfo(struct sk_buff *skb, struct net_device *dev,
5473 int type, u32 pid, u32 seq, u32 change,
5475 const struct rtnl_stats_dump_filters *filters,
5476 int *idxattr, int *prividx,
5477 struct netlink_ext_ack *extack)
5479 unsigned int filter_mask = filters->mask[0];
5480 struct if_stats_msg *ifsm;
5481 struct nlmsghdr *nlh;
5482 struct nlattr *attr;
5483 int s_prividx = *prividx;
5488 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifsm), flags);
5492 ifsm = nlmsg_data(nlh);
5493 ifsm->family = PF_UNSPEC;
5496 ifsm->ifindex = dev->ifindex;
5497 ifsm->filter_mask = filter_mask;
5499 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, *idxattr)) {
5500 struct rtnl_link_stats64 *sp;
5502 attr = nla_reserve_64bit(skb, IFLA_STATS_LINK_64,
5503 sizeof(struct rtnl_link_stats64),
5507 goto nla_put_failure;
5510 sp = nla_data(attr);
5511 dev_get_stats(dev, sp);
5514 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, *idxattr)) {
5515 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
5517 if (ops && ops->fill_linkxstats) {
5518 *idxattr = IFLA_STATS_LINK_XSTATS;
5519 attr = nla_nest_start_noflag(skb,
5520 IFLA_STATS_LINK_XSTATS);
5523 goto nla_put_failure;
5526 err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
5527 nla_nest_end(skb, attr);
5529 goto nla_put_failure;
5534 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE,
5536 const struct rtnl_link_ops *ops = NULL;
5537 const struct net_device *master;
5539 master = netdev_master_upper_dev_get(dev);
5541 ops = master->rtnl_link_ops;
5542 if (ops && ops->fill_linkxstats) {
5543 *idxattr = IFLA_STATS_LINK_XSTATS_SLAVE;
5544 attr = nla_nest_start_noflag(skb,
5545 IFLA_STATS_LINK_XSTATS_SLAVE);
5548 goto nla_put_failure;
5551 err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
5552 nla_nest_end(skb, attr);
5554 goto nla_put_failure;
5559 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS,
5561 u32 off_filter_mask;
5563 off_filter_mask = filters->mask[IFLA_STATS_LINK_OFFLOAD_XSTATS];
5564 *idxattr = IFLA_STATS_LINK_OFFLOAD_XSTATS;
5565 attr = nla_nest_start_noflag(skb,
5566 IFLA_STATS_LINK_OFFLOAD_XSTATS);
5569 goto nla_put_failure;
5572 err = rtnl_offload_xstats_fill(skb, dev, prividx,
5573 off_filter_mask, extack);
5574 if (err == -ENODATA)
5575 nla_nest_cancel(skb, attr);
5577 nla_nest_end(skb, attr);
5579 if (err && err != -ENODATA)
5580 goto nla_put_failure;
5584 if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, *idxattr)) {
5585 struct rtnl_af_ops *af_ops;
5587 *idxattr = IFLA_STATS_AF_SPEC;
5588 attr = nla_nest_start_noflag(skb, IFLA_STATS_AF_SPEC);
5591 goto nla_put_failure;
5595 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
5596 if (af_ops->fill_stats_af) {
5599 af = nla_nest_start_noflag(skb,
5604 goto nla_put_failure;
5606 err = af_ops->fill_stats_af(skb, dev);
5608 if (err == -ENODATA) {
5609 nla_nest_cancel(skb, af);
5610 } else if (err < 0) {
5612 goto nla_put_failure;
5615 nla_nest_end(skb, af);
5620 nla_nest_end(skb, attr);
5625 nlmsg_end(skb, nlh);
5630 /* not a multi message or no progress mean a real error */
5631 if (!(flags & NLM_F_MULTI) || s_prividx == *prividx)
5632 nlmsg_cancel(skb, nlh);
5634 nlmsg_end(skb, nlh);
5639 static size_t if_nlmsg_stats_size(const struct net_device *dev,
5640 const struct rtnl_stats_dump_filters *filters)
5642 size_t size = NLMSG_ALIGN(sizeof(struct if_stats_msg));
5643 unsigned int filter_mask = filters->mask[0];
5645 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, 0))
5646 size += nla_total_size_64bit(sizeof(struct rtnl_link_stats64));
5648 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, 0)) {
5649 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
5650 int attr = IFLA_STATS_LINK_XSTATS;
5652 if (ops && ops->get_linkxstats_size) {
5653 size += nla_total_size(ops->get_linkxstats_size(dev,
5655 /* for IFLA_STATS_LINK_XSTATS */
5656 size += nla_total_size(0);
5660 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE, 0)) {
5661 struct net_device *_dev = (struct net_device *)dev;
5662 const struct rtnl_link_ops *ops = NULL;
5663 const struct net_device *master;
5665 /* netdev_master_upper_dev_get can't take const */
5666 master = netdev_master_upper_dev_get(_dev);
5668 ops = master->rtnl_link_ops;
5669 if (ops && ops->get_linkxstats_size) {
5670 int attr = IFLA_STATS_LINK_XSTATS_SLAVE;
5672 size += nla_total_size(ops->get_linkxstats_size(dev,
5674 /* for IFLA_STATS_LINK_XSTATS_SLAVE */
5675 size += nla_total_size(0);
5679 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS, 0)) {
5680 u32 off_filter_mask;
5682 off_filter_mask = filters->mask[IFLA_STATS_LINK_OFFLOAD_XSTATS];
5683 size += rtnl_offload_xstats_get_size(dev, off_filter_mask);
5686 if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, 0)) {
5687 struct rtnl_af_ops *af_ops;
5689 /* for IFLA_STATS_AF_SPEC */
5690 size += nla_total_size(0);
5693 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
5694 if (af_ops->get_stats_af_size) {
5695 size += nla_total_size(
5696 af_ops->get_stats_af_size(dev));
5699 size += nla_total_size(0);
5708 #define RTNL_STATS_OFFLOAD_XSTATS_VALID ((1 << __IFLA_OFFLOAD_XSTATS_MAX) - 1)
5710 static const struct nla_policy
5711 rtnl_stats_get_policy_filters[IFLA_STATS_MAX + 1] = {
5712 [IFLA_STATS_LINK_OFFLOAD_XSTATS] =
5713 NLA_POLICY_MASK(NLA_U32, RTNL_STATS_OFFLOAD_XSTATS_VALID),
5716 static const struct nla_policy
5717 rtnl_stats_get_policy[IFLA_STATS_GETSET_MAX + 1] = {
5718 [IFLA_STATS_GET_FILTERS] =
5719 NLA_POLICY_NESTED(rtnl_stats_get_policy_filters),
5722 static const struct nla_policy
5723 ifla_stats_set_policy[IFLA_STATS_GETSET_MAX + 1] = {
5724 [IFLA_STATS_SET_OFFLOAD_XSTATS_L3_STATS] = NLA_POLICY_MAX(NLA_U8, 1),
5727 static int rtnl_stats_get_parse_filters(struct nlattr *ifla_filters,
5728 struct rtnl_stats_dump_filters *filters,
5729 struct netlink_ext_ack *extack)
5731 struct nlattr *tb[IFLA_STATS_MAX + 1];
5735 err = nla_parse_nested(tb, IFLA_STATS_MAX, ifla_filters,
5736 rtnl_stats_get_policy_filters, extack);
5740 for (at = 1; at <= IFLA_STATS_MAX; at++) {
5742 if (!(filters->mask[0] & IFLA_STATS_FILTER_BIT(at))) {
5743 NL_SET_ERR_MSG(extack, "Filtered attribute not enabled in filter_mask");
5746 filters->mask[at] = nla_get_u32(tb[at]);
5753 static int rtnl_stats_get_parse(const struct nlmsghdr *nlh,
5755 struct rtnl_stats_dump_filters *filters,
5756 struct netlink_ext_ack *extack)
5758 struct nlattr *tb[IFLA_STATS_GETSET_MAX + 1];
5762 filters->mask[0] = filter_mask;
5763 for (i = 1; i < ARRAY_SIZE(filters->mask); i++)
5764 filters->mask[i] = -1U;
5766 err = nlmsg_parse(nlh, sizeof(struct if_stats_msg), tb,
5767 IFLA_STATS_GETSET_MAX, rtnl_stats_get_policy, extack);
5771 if (tb[IFLA_STATS_GET_FILTERS]) {
5772 err = rtnl_stats_get_parse_filters(tb[IFLA_STATS_GET_FILTERS],
5781 static int rtnl_valid_stats_req(const struct nlmsghdr *nlh, bool strict_check,
5782 bool is_dump, struct netlink_ext_ack *extack)
5784 struct if_stats_msg *ifsm;
5786 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifsm))) {
5787 NL_SET_ERR_MSG(extack, "Invalid header for stats dump");
5794 ifsm = nlmsg_data(nlh);
5796 /* only requests using strict checks can pass data to influence
5797 * the dump. The legacy exception is filter_mask.
5799 if (ifsm->pad1 || ifsm->pad2 || (is_dump && ifsm->ifindex)) {
5800 NL_SET_ERR_MSG(extack, "Invalid values in header for stats dump request");
5803 if (ifsm->filter_mask >= IFLA_STATS_FILTER_BIT(IFLA_STATS_MAX + 1)) {
5804 NL_SET_ERR_MSG(extack, "Invalid stats requested through filter mask");
5811 static int rtnl_stats_get(struct sk_buff *skb, struct nlmsghdr *nlh,
5812 struct netlink_ext_ack *extack)
5814 struct rtnl_stats_dump_filters filters;
5815 struct net *net = sock_net(skb->sk);
5816 struct net_device *dev = NULL;
5817 int idxattr = 0, prividx = 0;
5818 struct if_stats_msg *ifsm;
5819 struct sk_buff *nskb;
5822 err = rtnl_valid_stats_req(nlh, netlink_strict_get_check(skb),
5827 ifsm = nlmsg_data(nlh);
5828 if (ifsm->ifindex > 0)
5829 dev = __dev_get_by_index(net, ifsm->ifindex);
5836 if (!ifsm->filter_mask) {
5837 NL_SET_ERR_MSG(extack, "Filter mask must be set for stats get");
5841 err = rtnl_stats_get_parse(nlh, ifsm->filter_mask, &filters, extack);
5845 nskb = nlmsg_new(if_nlmsg_stats_size(dev, &filters), GFP_KERNEL);
5849 err = rtnl_fill_statsinfo(nskb, dev, RTM_NEWSTATS,
5850 NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
5851 0, &filters, &idxattr, &prividx, extack);
5853 /* -EMSGSIZE implies BUG in if_nlmsg_stats_size */
5854 WARN_ON(err == -EMSGSIZE);
5857 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
5863 static int rtnl_stats_dump(struct sk_buff *skb, struct netlink_callback *cb)
5865 struct netlink_ext_ack *extack = cb->extack;
5866 int h, s_h, err, s_idx, s_idxattr, s_prividx;
5867 struct rtnl_stats_dump_filters filters;
5868 struct net *net = sock_net(skb->sk);
5869 unsigned int flags = NLM_F_MULTI;
5870 struct if_stats_msg *ifsm;
5871 struct hlist_head *head;
5872 struct net_device *dev;
5876 s_idx = cb->args[1];
5877 s_idxattr = cb->args[2];
5878 s_prividx = cb->args[3];
5880 cb->seq = net->dev_base_seq;
5882 err = rtnl_valid_stats_req(cb->nlh, cb->strict_check, true, extack);
5886 ifsm = nlmsg_data(cb->nlh);
5887 if (!ifsm->filter_mask) {
5888 NL_SET_ERR_MSG(extack, "Filter mask must be set for stats dump");
5892 err = rtnl_stats_get_parse(cb->nlh, ifsm->filter_mask, &filters,
5897 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5899 head = &net->dev_index_head[h];
5900 hlist_for_each_entry(dev, head, index_hlist) {
5903 err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS,
5904 NETLINK_CB(cb->skb).portid,
5905 cb->nlh->nlmsg_seq, 0,
5907 &s_idxattr, &s_prividx,
5909 /* If we ran out of room on the first message,
5912 WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
5918 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
5924 cb->args[3] = s_prividx;
5925 cb->args[2] = s_idxattr;
5932 void rtnl_offload_xstats_notify(struct net_device *dev)
5934 struct rtnl_stats_dump_filters response_filters = {};
5935 struct net *net = dev_net(dev);
5936 int idxattr = 0, prividx = 0;
5937 struct sk_buff *skb;
5942 response_filters.mask[0] |=
5943 IFLA_STATS_FILTER_BIT(IFLA_STATS_LINK_OFFLOAD_XSTATS);
5944 response_filters.mask[IFLA_STATS_LINK_OFFLOAD_XSTATS] |=
5945 IFLA_STATS_FILTER_BIT(IFLA_OFFLOAD_XSTATS_HW_S_INFO);
5947 skb = nlmsg_new(if_nlmsg_stats_size(dev, &response_filters),
5952 err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS, 0, 0, 0, 0,
5953 &response_filters, &idxattr, &prividx, NULL);
5959 rtnl_notify(skb, net, 0, RTNLGRP_STATS, NULL, GFP_KERNEL);
5963 rtnl_set_sk_err(net, RTNLGRP_STATS, err);
5965 EXPORT_SYMBOL(rtnl_offload_xstats_notify);
5967 static int rtnl_stats_set(struct sk_buff *skb, struct nlmsghdr *nlh,
5968 struct netlink_ext_ack *extack)
5970 enum netdev_offload_xstats_type t_l3 = NETDEV_OFFLOAD_XSTATS_TYPE_L3;
5971 struct rtnl_stats_dump_filters response_filters = {};
5972 struct nlattr *tb[IFLA_STATS_GETSET_MAX + 1];
5973 struct net *net = sock_net(skb->sk);
5974 struct net_device *dev = NULL;
5975 struct if_stats_msg *ifsm;
5976 bool notify = false;
5979 err = rtnl_valid_stats_req(nlh, netlink_strict_get_check(skb),
5984 ifsm = nlmsg_data(nlh);
5985 if (ifsm->family != AF_UNSPEC) {
5986 NL_SET_ERR_MSG(extack, "Address family should be AF_UNSPEC");
5990 if (ifsm->ifindex > 0)
5991 dev = __dev_get_by_index(net, ifsm->ifindex);
5998 if (ifsm->filter_mask) {
5999 NL_SET_ERR_MSG(extack, "Filter mask must be 0 for stats set");
6003 err = nlmsg_parse(nlh, sizeof(*ifsm), tb, IFLA_STATS_GETSET_MAX,
6004 ifla_stats_set_policy, extack);
6008 if (tb[IFLA_STATS_SET_OFFLOAD_XSTATS_L3_STATS]) {
6009 u8 req = nla_get_u8(tb[IFLA_STATS_SET_OFFLOAD_XSTATS_L3_STATS]);
6012 err = netdev_offload_xstats_enable(dev, t_l3, extack);
6014 err = netdev_offload_xstats_disable(dev, t_l3);
6018 else if (err != -EALREADY)
6021 response_filters.mask[0] |=
6022 IFLA_STATS_FILTER_BIT(IFLA_STATS_LINK_OFFLOAD_XSTATS);
6023 response_filters.mask[IFLA_STATS_LINK_OFFLOAD_XSTATS] |=
6024 IFLA_STATS_FILTER_BIT(IFLA_OFFLOAD_XSTATS_HW_S_INFO);
6028 rtnl_offload_xstats_notify(dev);
6033 /* Process one rtnetlink message. */
6035 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh,
6036 struct netlink_ext_ack *extack)
6038 struct net *net = sock_net(skb->sk);
6039 struct rtnl_link *link;
6040 enum rtnl_kinds kind;
6041 struct module *owner;
6042 int err = -EOPNOTSUPP;
6043 rtnl_doit_func doit;
6048 type = nlh->nlmsg_type;
6054 /* All the messages must have at least 1 byte length */
6055 if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
6058 family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
6059 kind = rtnl_msgtype_kind(type);
6061 if (kind != RTNL_KIND_GET && !netlink_net_capable(skb, CAP_NET_ADMIN))
6065 if (kind == RTNL_KIND_GET && (nlh->nlmsg_flags & NLM_F_DUMP)) {
6067 rtnl_dumpit_func dumpit;
6068 u32 min_dump_alloc = 0;
6070 link = rtnl_get_link(family, type);
6071 if (!link || !link->dumpit) {
6073 link = rtnl_get_link(family, type);
6074 if (!link || !link->dumpit)
6077 owner = link->owner;
6078 dumpit = link->dumpit;
6080 if (type == RTM_GETLINK - RTM_BASE)
6081 min_dump_alloc = rtnl_calcit(skb, nlh);
6084 /* need to do this before rcu_read_unlock() */
6085 if (!try_module_get(owner))
6086 err = -EPROTONOSUPPORT;
6092 struct netlink_dump_control c = {
6094 .min_dump_alloc = min_dump_alloc,
6097 err = netlink_dump_start(rtnl, skb, nlh, &c);
6098 /* netlink_dump_start() will keep a reference on
6099 * module if dump is still in progress.
6106 link = rtnl_get_link(family, type);
6107 if (!link || !link->doit) {
6109 link = rtnl_get_link(PF_UNSPEC, type);
6110 if (!link || !link->doit)
6114 owner = link->owner;
6115 if (!try_module_get(owner)) {
6116 err = -EPROTONOSUPPORT;
6120 flags = link->flags;
6121 if (kind == RTNL_KIND_DEL && (nlh->nlmsg_flags & NLM_F_BULK) &&
6122 !(flags & RTNL_FLAG_BULK_DEL_SUPPORTED)) {
6123 NL_SET_ERR_MSG(extack, "Bulk delete is not supported");
6128 if (flags & RTNL_FLAG_DOIT_UNLOCKED) {
6132 err = doit(skb, nlh, extack);
6139 link = rtnl_get_link(family, type);
6140 if (link && link->doit)
6141 err = link->doit(skb, nlh, extack);
6157 static void rtnetlink_rcv(struct sk_buff *skb)
6159 netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
6162 static int rtnetlink_bind(struct net *net, int group)
6165 case RTNLGRP_IPV4_MROUTE_R:
6166 case RTNLGRP_IPV6_MROUTE_R:
6167 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
6174 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
6176 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
6180 case NETDEV_CHANGEMTU:
6181 case NETDEV_CHANGEADDR:
6182 case NETDEV_CHANGENAME:
6183 case NETDEV_FEAT_CHANGE:
6184 case NETDEV_BONDING_FAILOVER:
6185 case NETDEV_POST_TYPE_CHANGE:
6186 case NETDEV_NOTIFY_PEERS:
6187 case NETDEV_CHANGEUPPER:
6188 case NETDEV_RESEND_IGMP:
6189 case NETDEV_CHANGEINFODATA:
6190 case NETDEV_CHANGELOWERSTATE:
6191 case NETDEV_CHANGE_TX_QUEUE_LEN:
6192 rtmsg_ifinfo_event(RTM_NEWLINK, dev, 0, rtnl_get_event(event),
6193 GFP_KERNEL, NULL, 0, 0, NULL);
6201 static struct notifier_block rtnetlink_dev_notifier = {
6202 .notifier_call = rtnetlink_event,
6206 static int __net_init rtnetlink_net_init(struct net *net)
6209 struct netlink_kernel_cfg cfg = {
6210 .groups = RTNLGRP_MAX,
6211 .input = rtnetlink_rcv,
6212 .cb_mutex = &rtnl_mutex,
6213 .flags = NL_CFG_F_NONROOT_RECV,
6214 .bind = rtnetlink_bind,
6217 sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
6224 static void __net_exit rtnetlink_net_exit(struct net *net)
6226 netlink_kernel_release(net->rtnl);
6230 static struct pernet_operations rtnetlink_net_ops = {
6231 .init = rtnetlink_net_init,
6232 .exit = rtnetlink_net_exit,
6235 void __init rtnetlink_init(void)
6237 if (register_pernet_subsys(&rtnetlink_net_ops))
6238 panic("rtnetlink_init: cannot initialize rtnetlink\n");
6240 register_netdevice_notifier(&rtnetlink_dev_notifier);
6242 rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
6243 rtnl_dump_ifinfo, 0);
6244 rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, 0);
6245 rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, 0);
6246 rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, 0);
6248 rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, 0);
6249 rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, 0);
6250 rtnl_register(PF_UNSPEC, RTM_GETNETCONF, NULL, rtnl_dump_all, 0);
6252 rtnl_register(PF_UNSPEC, RTM_NEWLINKPROP, rtnl_newlinkprop, NULL, 0);
6253 rtnl_register(PF_UNSPEC, RTM_DELLINKPROP, rtnl_dellinkprop, NULL, 0);
6255 rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, 0);
6256 rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL,
6257 RTNL_FLAG_BULK_DEL_SUPPORTED);
6258 rtnl_register(PF_BRIDGE, RTM_GETNEIGH, rtnl_fdb_get, rtnl_fdb_dump, 0);
6260 rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, 0);
6261 rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, 0);
6262 rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, 0);
6264 rtnl_register(PF_UNSPEC, RTM_GETSTATS, rtnl_stats_get, rtnl_stats_dump,
6266 rtnl_register(PF_UNSPEC, RTM_SETSTATS, rtnl_stats_set, NULL, 0);