]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blob - net/core/rtnetlink.c
projects / mirror_ubuntu-bionic-kernel.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
UBUNTU: Ubuntu-4.15.0-96.97
[mirror_ubuntu-bionic-kernel.git] / net / core / rtnetlink.c
1 /*
2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
5 *
6 * Routing netlink socket interface: protocol independent part.
7 *
8 * Authors: Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
9 *
10 * This program is free software; you can redistribute it and/or
11 * modify it under the terms of the GNU General Public License
12 * as published by the Free Software Foundation; either version
13 * 2 of the License, or (at your option) any later version.
14 *
15 * Fixes:
16 * Vitaly E. Lavrov RTA_OK arithmetics was wrong.
17 */
18
19 #include <linux/bitops.h>
20 #include <linux/errno.h>
21 #include <linux/module.h>
22 #include <linux/types.h>
23 #include <linux/socket.h>
24 #include <linux/kernel.h>
25 #include <linux/timer.h>
26 #include <linux/string.h>
27 #include <linux/sockios.h>
28 #include <linux/net.h>
29 #include <linux/fcntl.h>
30 #include <linux/mm.h>
31 #include <linux/slab.h>
32 #include <linux/interrupt.h>
33 #include <linux/capability.h>
34 #include <linux/skbuff.h>
35 #include <linux/init.h>
36 #include <linux/security.h>
37 #include <linux/mutex.h>
38 #include <linux/if_addr.h>
39 #include <linux/if_bridge.h>
40 #include <linux/if_vlan.h>
41 #include <linux/pci.h>
42 #include <linux/etherdevice.h>
43 #include <linux/bpf.h>
44
45 #include <linux/uaccess.h>
46
47 #include <linux/inet.h>
48 #include <linux/netdevice.h>
49 #include <net/switchdev.h>
50 #include <net/ip.h>
51 #include <net/protocol.h>
52 #include <net/arp.h>
53 #include <net/route.h>
54 #include <net/udp.h>
55 #include <net/tcp.h>
56 #include <net/sock.h>
57 #include <net/pkt_sched.h>
58 #include <net/fib_rules.h>
59 #include <net/rtnetlink.h>
60 #include <net/net_namespace.h>
61
62 struct rtnl_link {
63 rtnl_doit_func doit;
64 rtnl_dumpit_func dumpit;
65 unsigned int flags;
66 };
67
68 static DEFINE_MUTEX(rtnl_mutex);
69
70 void rtnl_lock(void)
71 {
72 mutex_lock(&rtnl_mutex);
73 }
74 EXPORT_SYMBOL(rtnl_lock);
75
76 static struct sk_buff *defer_kfree_skb_list;
77 void rtnl_kfree_skbs(struct sk_buff *head, struct sk_buff *tail)
78 {
79 if (head && tail) {
80 tail->next = defer_kfree_skb_list;
81 defer_kfree_skb_list = head;
82 }
83 }
84 EXPORT_SYMBOL(rtnl_kfree_skbs);
85
86 void __rtnl_unlock(void)
87 {
88 struct sk_buff *head = defer_kfree_skb_list;
89
90 defer_kfree_skb_list = NULL;
91
92 mutex_unlock(&rtnl_mutex);
93
94 while (head) {
95 struct sk_buff *next = head->next;
96
97 kfree_skb(head);
98 cond_resched();
99 head = next;
100 }
101 }
102
103 void rtnl_unlock(void)
104 {
105 /* This fellow will unlock it for us. */
106 netdev_run_todo();
107 }
108 EXPORT_SYMBOL(rtnl_unlock);
109
110 int rtnl_trylock(void)
111 {
112 return mutex_trylock(&rtnl_mutex);
113 }
114 EXPORT_SYMBOL(rtnl_trylock);
115
116 int rtnl_is_locked(void)
117 {
118 return mutex_is_locked(&rtnl_mutex);
119 }
120 EXPORT_SYMBOL(rtnl_is_locked);
121
122 #ifdef CONFIG_PROVE_LOCKING
123 bool lockdep_rtnl_is_held(void)
124 {
125 return lockdep_is_held(&rtnl_mutex);
126 }
127 EXPORT_SYMBOL(lockdep_rtnl_is_held);
128 #endif /* #ifdef CONFIG_PROVE_LOCKING */
129
130 static struct rtnl_link __rcu *rtnl_msg_handlers[RTNL_FAMILY_MAX + 1];
131 static refcount_t rtnl_msg_handlers_ref[RTNL_FAMILY_MAX + 1];
132
133 static inline int rtm_msgindex(int msgtype)
134 {
135 int msgindex = msgtype - RTM_BASE;
136
137 /*
138 * msgindex < 0 implies someone tried to register a netlink
139 * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
140 * the message type has not been added to linux/rtnetlink.h
141 */
142 BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
143
144 return msgindex;
145 }
146
147 /**
148 * __rtnl_register - Register a rtnetlink message type
149 * @protocol: Protocol family or PF_UNSPEC
150 * @msgtype: rtnetlink message type
151 * @doit: Function pointer called for each request message
152 * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
153 * @flags: rtnl_link_flags to modifiy behaviour of doit/dumpit functions
154 *
155 * Registers the specified function pointers (at least one of them has
156 * to be non-NULL) to be called whenever a request message for the
157 * specified protocol family and message type is received.
158 *
159 * The special protocol family PF_UNSPEC may be used to define fallback
160 * function pointers for the case when no entry for the specific protocol
161 * family exists.
162 *
163 * Returns 0 on success or a negative error code.
164 */
165 int __rtnl_register(int protocol, int msgtype,
166 rtnl_doit_func doit, rtnl_dumpit_func dumpit,
167 unsigned int flags)
168 {
169 struct rtnl_link *tab;
170 int msgindex;
171
172 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
173 msgindex = rtm_msgindex(msgtype);
174
175 tab = rcu_dereference_raw(rtnl_msg_handlers[protocol]);
176 if (tab == NULL) {
177 tab = kcalloc(RTM_NR_MSGTYPES, sizeof(*tab), GFP_KERNEL);
178 if (tab == NULL)
179 return -ENOBUFS;
180
181 rcu_assign_pointer(rtnl_msg_handlers[protocol], tab);
182 }
183
184 if (doit)
185 tab[msgindex].doit = doit;
186 if (dumpit)
187 tab[msgindex].dumpit = dumpit;
188 tab[msgindex].flags |= flags;
189
190 return 0;
191 }
192 EXPORT_SYMBOL_GPL(__rtnl_register);
193
194 /**
195 * rtnl_register - Register a rtnetlink message type
196 *
197 * Identical to __rtnl_register() but panics on failure. This is useful
198 * as failure of this function is very unlikely, it can only happen due
199 * to lack of memory when allocating the chain to store all message
200 * handlers for a protocol. Meant for use in init functions where lack
201 * of memory implies no sense in continuing.
202 */
203 void rtnl_register(int protocol, int msgtype,
204 rtnl_doit_func doit, rtnl_dumpit_func dumpit,
205 unsigned int flags)
206 {
207 if (__rtnl_register(protocol, msgtype, doit, dumpit, flags) < 0)
208 panic("Unable to register rtnetlink message handler, "
209 "protocol = %d, message type = %d\n",
210 protocol, msgtype);
211 }
212 EXPORT_SYMBOL_GPL(rtnl_register);
213
214 /**
215 * rtnl_unregister - Unregister a rtnetlink message type
216 * @protocol: Protocol family or PF_UNSPEC
217 * @msgtype: rtnetlink message type
218 *
219 * Returns 0 on success or a negative error code.
220 */
221 int rtnl_unregister(int protocol, int msgtype)
222 {
223 struct rtnl_link *handlers;
224 int msgindex;
225
226 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
227 msgindex = rtm_msgindex(msgtype);
228
229 rtnl_lock();
230 handlers = rtnl_dereference(rtnl_msg_handlers[protocol]);
231 if (!handlers) {
232 rtnl_unlock();
233 return -ENOENT;
234 }
235
236 handlers[msgindex].doit = NULL;
237 handlers[msgindex].dumpit = NULL;
238 handlers[msgindex].flags = 0;
239 rtnl_unlock();
240
241 return 0;
242 }
243 EXPORT_SYMBOL_GPL(rtnl_unregister);
244
245 /**
246 * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
247 * @protocol : Protocol family or PF_UNSPEC
248 *
249 * Identical to calling rtnl_unregster() for all registered message types
250 * of a certain protocol family.
251 */
252 void rtnl_unregister_all(int protocol)
253 {
254 struct rtnl_link *handlers;
255
256 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
257
258 rtnl_lock();
259 handlers = rtnl_dereference(rtnl_msg_handlers[protocol]);
260 RCU_INIT_POINTER(rtnl_msg_handlers[protocol], NULL);
261 rtnl_unlock();
262
263 synchronize_net();
264
265 while (refcount_read(&rtnl_msg_handlers_ref[protocol]) > 1)
266 schedule();
267 kfree(handlers);
268 }
269 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
270
271 static LIST_HEAD(link_ops);
272
273 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
274 {
275 const struct rtnl_link_ops *ops;
276
277 list_for_each_entry(ops, &link_ops, list) {
278 if (!strcmp(ops->kind, kind))
279 return ops;
280 }
281 return NULL;
282 }
283
284 /**
285 * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
286 * @ops: struct rtnl_link_ops * to register
287 *
288 * The caller must hold the rtnl_mutex. This function should be used
289 * by drivers that create devices during module initialization. It
290 * must be called before registering the devices.
291 *
292 * Returns 0 on success or a negative error code.
293 */
294 int __rtnl_link_register(struct rtnl_link_ops *ops)
295 {
296 if (rtnl_link_ops_get(ops->kind))
297 return -EEXIST;
298
299 /* The check for setup is here because if ops
300 * does not have that filled up, it is not possible
301 * to use the ops for creating device. So do not
302 * fill up dellink as well. That disables rtnl_dellink.
303 */
304 if (ops->setup && !ops->dellink)
305 ops->dellink = unregister_netdevice_queue;
306
307 list_add_tail(&ops->list, &link_ops);
308 return 0;
309 }
310 EXPORT_SYMBOL_GPL(__rtnl_link_register);
311
312 /**
313 * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
314 * @ops: struct rtnl_link_ops * to register
315 *
316 * Returns 0 on success or a negative error code.
317 */
318 int rtnl_link_register(struct rtnl_link_ops *ops)
319 {
320 int err;
321
322 rtnl_lock();
323 err = __rtnl_link_register(ops);
324 rtnl_unlock();
325 return err;
326 }
327 EXPORT_SYMBOL_GPL(rtnl_link_register);
328
329 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
330 {
331 struct net_device *dev;
332 LIST_HEAD(list_kill);
333
334 for_each_netdev(net, dev) {
335 if (dev->rtnl_link_ops == ops)
336 ops->dellink(dev, &list_kill);
337 }
338 unregister_netdevice_many(&list_kill);
339 }
340
341 /**
342 * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
343 * @ops: struct rtnl_link_ops * to unregister
344 *
345 * The caller must hold the rtnl_mutex.
346 */
347 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
348 {
349 struct net *net;
350
351 for_each_net(net) {
352 __rtnl_kill_links(net, ops);
353 }
354 list_del(&ops->list);
355 }
356 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
357
358 /* Return with the rtnl_lock held when there are no network
359 * devices unregistering in any network namespace.
360 */
361 static void rtnl_lock_unregistering_all(void)
362 {
363 struct net *net;
364 bool unregistering;
365 DEFINE_WAIT_FUNC(wait, woken_wake_function);
366
367 add_wait_queue(&netdev_unregistering_wq, &wait);
368 for (;;) {
369 unregistering = false;
370 rtnl_lock();
371 for_each_net(net) {
372 if (net->dev_unreg_count > 0) {
373 unregistering = true;
374 break;
375 }
376 }
377 if (!unregistering)
378 break;
379 __rtnl_unlock();
380
381 wait_woken(&wait, TASK_UNINTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
382 }
383 remove_wait_queue(&netdev_unregistering_wq, &wait);
384 }
385
386 /**
387 * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
388 * @ops: struct rtnl_link_ops * to unregister
389 */
390 void rtnl_link_unregister(struct rtnl_link_ops *ops)
391 {
392 /* Close the race with cleanup_net() */
393 mutex_lock(&net_mutex);
394 rtnl_lock_unregistering_all();
395 __rtnl_link_unregister(ops);
396 rtnl_unlock();
397 mutex_unlock(&net_mutex);
398 }
399 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
400
401 static size_t rtnl_link_get_slave_info_data_size(const struct net_device *dev)
402 {
403 struct net_device *master_dev;
404 const struct rtnl_link_ops *ops;
405 size_t size = 0;
406
407 rcu_read_lock();
408
409 master_dev = netdev_master_upper_dev_get_rcu((struct net_device *)dev);
410 if (!master_dev)
411 goto out;
412
413 ops = master_dev->rtnl_link_ops;
414 if (!ops || !ops->get_slave_size)
415 goto out;
416 /* IFLA_INFO_SLAVE_DATA + nested data */
417 size = nla_total_size(sizeof(struct nlattr)) +
418 ops->get_slave_size(master_dev, dev);
419
420 out:
421 rcu_read_unlock();
422 return size;
423 }
424
425 static size_t rtnl_link_get_size(const struct net_device *dev)
426 {
427 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
428 size_t size;
429
430 if (!ops)
431 return 0;
432
433 size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
434 nla_total_size(strlen(ops->kind) + 1); /* IFLA_INFO_KIND */
435
436 if (ops->get_size)
437 /* IFLA_INFO_DATA + nested data */
438 size += nla_total_size(sizeof(struct nlattr)) +
439 ops->get_size(dev);
440
441 if (ops->get_xstats_size)
442 /* IFLA_INFO_XSTATS */
443 size += nla_total_size(ops->get_xstats_size(dev));
444
445 size += rtnl_link_get_slave_info_data_size(dev);
446
447 return size;
448 }
449
450 static LIST_HEAD(rtnl_af_ops);
451
452 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
453 {
454 const struct rtnl_af_ops *ops;
455
456 list_for_each_entry_rcu(ops, &rtnl_af_ops, list) {
457 if (ops->family == family)
458 return ops;
459 }
460
461 return NULL;
462 }
463
464 /**
465 * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
466 * @ops: struct rtnl_af_ops * to register
467 *
468 * Returns 0 on success or a negative error code.
469 */
470 void rtnl_af_register(struct rtnl_af_ops *ops)
471 {
472 rtnl_lock();
473 list_add_tail_rcu(&ops->list, &rtnl_af_ops);
474 rtnl_unlock();
475 }
476 EXPORT_SYMBOL_GPL(rtnl_af_register);
477
478 /**
479 * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
480 * @ops: struct rtnl_af_ops * to unregister
481 */
482 void rtnl_af_unregister(struct rtnl_af_ops *ops)
483 {
484 rtnl_lock();
485 list_del_rcu(&ops->list);
486 rtnl_unlock();
487
488 synchronize_rcu();
489 }
490 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
491
492 static size_t rtnl_link_get_af_size(const struct net_device *dev,
493 u32 ext_filter_mask)
494 {
495 struct rtnl_af_ops *af_ops;
496 size_t size;
497
498 /* IFLA_AF_SPEC */
499 size = nla_total_size(sizeof(struct nlattr));
500
501 rcu_read_lock();
502 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
503 if (af_ops->get_link_af_size) {
504 /* AF_* + nested data */
505 size += nla_total_size(sizeof(struct nlattr)) +
506 af_ops->get_link_af_size(dev, ext_filter_mask);
507 }
508 }
509 rcu_read_unlock();
510
511 return size;
512 }
513
514 static bool rtnl_have_link_slave_info(const struct net_device *dev)
515 {
516 struct net_device *master_dev;
517 bool ret = false;
518
519 rcu_read_lock();
520
521 master_dev = netdev_master_upper_dev_get_rcu((struct net_device *)dev);
522 if (master_dev && master_dev->rtnl_link_ops)
523 ret = true;
524 rcu_read_unlock();
525 return ret;
526 }
527
528 static int rtnl_link_slave_info_fill(struct sk_buff *skb,
529 const struct net_device *dev)
530 {
531 struct net_device *master_dev;
532 const struct rtnl_link_ops *ops;
533 struct nlattr *slave_data;
534 int err;
535
536 master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
537 if (!master_dev)
538 return 0;
539 ops = master_dev->rtnl_link_ops;
540 if (!ops)
541 return 0;
542 if (nla_put_string(skb, IFLA_INFO_SLAVE_KIND, ops->kind) < 0)
543 return -EMSGSIZE;
544 if (ops->fill_slave_info) {
545 slave_data = nla_nest_start(skb, IFLA_INFO_SLAVE_DATA);
546 if (!slave_data)
547 return -EMSGSIZE;
548 err = ops->fill_slave_info(skb, master_dev, dev);
549 if (err < 0)
550 goto err_cancel_slave_data;
551 nla_nest_end(skb, slave_data);
552 }
553 return 0;
554
555 err_cancel_slave_data:
556 nla_nest_cancel(skb, slave_data);
557 return err;
558 }
559
560 static int rtnl_link_info_fill(struct sk_buff *skb,
561 const struct net_device *dev)
562 {
563 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
564 struct nlattr *data;
565 int err;
566
567 if (!ops)
568 return 0;
569 if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
570 return -EMSGSIZE;
571 if (ops->fill_xstats) {
572 err = ops->fill_xstats(skb, dev);
573 if (err < 0)
574 return err;
575 }
576 if (ops->fill_info) {
577 data = nla_nest_start(skb, IFLA_INFO_DATA);
578 if (data == NULL)
579 return -EMSGSIZE;
580 err = ops->fill_info(skb, dev);
581 if (err < 0)
582 goto err_cancel_data;
583 nla_nest_end(skb, data);
584 }
585 return 0;
586
587 err_cancel_data:
588 nla_nest_cancel(skb, data);
589 return err;
590 }
591
592 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
593 {
594 struct nlattr *linkinfo;
595 int err = -EMSGSIZE;
596
597 linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
598 if (linkinfo == NULL)
599 goto out;
600
601 err = rtnl_link_info_fill(skb, dev);
602 if (err < 0)
603 goto err_cancel_link;
604
605 err = rtnl_link_slave_info_fill(skb, dev);
606 if (err < 0)
607 goto err_cancel_link;
608
609 nla_nest_end(skb, linkinfo);
610 return 0;
611
612 err_cancel_link:
613 nla_nest_cancel(skb, linkinfo);
614 out:
615 return err;
616 }
617
618 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
619 {
620 struct sock *rtnl = net->rtnl;
621 int err = 0;
622
623 NETLINK_CB(skb).dst_group = group;
624 if (echo)
625 refcount_inc(&skb->users);
626 netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
627 if (echo)
628 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
629 return err;
630 }
631
632 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
633 {
634 struct sock *rtnl = net->rtnl;
635
636 return nlmsg_unicast(rtnl, skb, pid);
637 }
638 EXPORT_SYMBOL(rtnl_unicast);
639
640 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
641 struct nlmsghdr *nlh, gfp_t flags)
642 {
643 struct sock *rtnl = net->rtnl;
644 int report = 0;
645
646 if (nlh)
647 report = nlmsg_report(nlh);
648
649 nlmsg_notify(rtnl, skb, pid, group, report, flags);
650 }
651 EXPORT_SYMBOL(rtnl_notify);
652
653 void rtnl_set_sk_err(struct net *net, u32 group, int error)
654 {
655 struct sock *rtnl = net->rtnl;
656
657 netlink_set_err(rtnl, 0, group, error);
658 }
659 EXPORT_SYMBOL(rtnl_set_sk_err);
660
661 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
662 {
663 struct nlattr *mx;
664 int i, valid = 0;
665
666 mx = nla_nest_start(skb, RTA_METRICS);
667 if (mx == NULL)
668 return -ENOBUFS;
669
670 for (i = 0; i < RTAX_MAX; i++) {
671 if (metrics[i]) {
672 if (i == RTAX_CC_ALGO - 1) {
673 char tmp[TCP_CA_NAME_MAX], *name;
674
675 name = tcp_ca_get_name_by_key(metrics[i], tmp);
676 if (!name)
677 continue;
678 if (nla_put_string(skb, i + 1, name))
679 goto nla_put_failure;
680 } else if (i == RTAX_FEATURES - 1) {
681 u32 user_features = metrics[i] & RTAX_FEATURE_MASK;
682
683 if (!user_features)
684 continue;
685 BUILD_BUG_ON(RTAX_FEATURE_MASK & DST_FEATURE_MASK);
686 if (nla_put_u32(skb, i + 1, user_features))
687 goto nla_put_failure;
688 } else {
689 if (nla_put_u32(skb, i + 1, metrics[i]))
690 goto nla_put_failure;
691 }
692 valid++;
693 }
694 }
695
696 if (!valid) {
697 nla_nest_cancel(skb, mx);
698 return 0;
699 }
700
701 return nla_nest_end(skb, mx);
702
703 nla_put_failure:
704 nla_nest_cancel(skb, mx);
705 return -EMSGSIZE;
706 }
707 EXPORT_SYMBOL(rtnetlink_put_metrics);
708
709 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
710 long expires, u32 error)
711 {
712 struct rta_cacheinfo ci = {
713 .rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse),
714 .rta_used = dst->__use,
715 .rta_clntref = atomic_read(&(dst->__refcnt)),
716 .rta_error = error,
717 .rta_id = id,
718 };
719
720 if (expires) {
721 unsigned long clock;
722
723 clock = jiffies_to_clock_t(abs(expires));
724 clock = min_t(unsigned long, clock, INT_MAX);
725 ci.rta_expires = (expires > 0) ? clock : -clock;
726 }
727 return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
728 }
729 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
730
731 static void set_operstate(struct net_device *dev, unsigned char transition)
732 {
733 unsigned char operstate = dev->operstate;
734
735 switch (transition) {
736 case IF_OPER_UP:
737 if ((operstate == IF_OPER_DORMANT ||
738 operstate == IF_OPER_UNKNOWN) &&
739 !netif_dormant(dev))
740 operstate = IF_OPER_UP;
741 break;
742
743 case IF_OPER_DORMANT:
744 if (operstate == IF_OPER_UP ||
745 operstate == IF_OPER_UNKNOWN)
746 operstate = IF_OPER_DORMANT;
747 break;
748 }
749
750 if (dev->operstate != operstate) {
751 write_lock_bh(&dev_base_lock);
752 dev->operstate = operstate;
753 write_unlock_bh(&dev_base_lock);
754 netdev_state_change(dev);
755 }
756 }
757
758 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
759 {
760 return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
761 (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
762 }
763
764 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
765 const struct ifinfomsg *ifm)
766 {
767 unsigned int flags = ifm->ifi_flags;
768
769 /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
770 if (ifm->ifi_change)
771 flags = (flags & ifm->ifi_change) |
772 (rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
773
774 return flags;
775 }
776
777 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
778 const struct rtnl_link_stats64 *b)
779 {
780 a->rx_packets = b->rx_packets;
781 a->tx_packets = b->tx_packets;
782 a->rx_bytes = b->rx_bytes;
783 a->tx_bytes = b->tx_bytes;
784 a->rx_errors = b->rx_errors;
785 a->tx_errors = b->tx_errors;
786 a->rx_dropped = b->rx_dropped;
787 a->tx_dropped = b->tx_dropped;
788
789 a->multicast = b->multicast;
790 a->collisions = b->collisions;
791
792 a->rx_length_errors = b->rx_length_errors;
793 a->rx_over_errors = b->rx_over_errors;
794 a->rx_crc_errors = b->rx_crc_errors;
795 a->rx_frame_errors = b->rx_frame_errors;
796 a->rx_fifo_errors = b->rx_fifo_errors;
797 a->rx_missed_errors = b->rx_missed_errors;
798
799 a->tx_aborted_errors = b->tx_aborted_errors;
800 a->tx_carrier_errors = b->tx_carrier_errors;
801 a->tx_fifo_errors = b->tx_fifo_errors;
802 a->tx_heartbeat_errors = b->tx_heartbeat_errors;
803 a->tx_window_errors = b->tx_window_errors;
804
805 a->rx_compressed = b->rx_compressed;
806 a->tx_compressed = b->tx_compressed;
807
808 a->rx_nohandler = b->rx_nohandler;
809 }
810
811 /* All VF info */
812 static inline int rtnl_vfinfo_size(const struct net_device *dev,
813 u32 ext_filter_mask)
814 {
815 if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF)) {
816 int num_vfs = dev_num_vf(dev->dev.parent);
817 size_t size = nla_total_size(0);
818 size += num_vfs *
819 (nla_total_size(0) +
820 nla_total_size(sizeof(struct ifla_vf_mac)) +
821 nla_total_size(sizeof(struct ifla_vf_vlan)) +
822 nla_total_size(0) + /* nest IFLA_VF_VLAN_LIST */
823 nla_total_size(MAX_VLAN_LIST_LEN *
824 sizeof(struct ifla_vf_vlan_info)) +
825 nla_total_size(sizeof(struct ifla_vf_spoofchk)) +
826 nla_total_size(sizeof(struct ifla_vf_tx_rate)) +
827 nla_total_size(sizeof(struct ifla_vf_rate)) +
828 nla_total_size(sizeof(struct ifla_vf_link_state)) +
829 nla_total_size(sizeof(struct ifla_vf_rss_query_en)) +
830 nla_total_size(0) + /* nest IFLA_VF_STATS */
831 /* IFLA_VF_STATS_RX_PACKETS */
832 nla_total_size_64bit(sizeof(__u64)) +
833 /* IFLA_VF_STATS_TX_PACKETS */
834 nla_total_size_64bit(sizeof(__u64)) +
835 /* IFLA_VF_STATS_RX_BYTES */
836 nla_total_size_64bit(sizeof(__u64)) +
837 /* IFLA_VF_STATS_TX_BYTES */
838 nla_total_size_64bit(sizeof(__u64)) +
839 /* IFLA_VF_STATS_BROADCAST */
840 nla_total_size_64bit(sizeof(__u64)) +
841 /* IFLA_VF_STATS_MULTICAST */
842 nla_total_size_64bit(sizeof(__u64)) +
843 nla_total_size(sizeof(struct ifla_vf_trust)));
844 return size;
845 } else
846 return 0;
847 }
848
849 static size_t rtnl_port_size(const struct net_device *dev,
850 u32 ext_filter_mask)
851 {
852 size_t port_size = nla_total_size(4) /* PORT_VF */
853 + nla_total_size(PORT_PROFILE_MAX) /* PORT_PROFILE */
854 + nla_total_size(PORT_UUID_MAX) /* PORT_INSTANCE_UUID */
855 + nla_total_size(PORT_UUID_MAX) /* PORT_HOST_UUID */
856 + nla_total_size(1) /* PROT_VDP_REQUEST */
857 + nla_total_size(2); /* PORT_VDP_RESPONSE */
858 size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
859 size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
860 + port_size;
861 size_t port_self_size = nla_total_size(sizeof(struct nlattr))
862 + port_size;
863
864 if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
865 !(ext_filter_mask & RTEXT_FILTER_VF))
866 return 0;
867 if (dev_num_vf(dev->dev.parent))
868 return port_self_size + vf_ports_size +
869 vf_port_size * dev_num_vf(dev->dev.parent);
870 else
871 return port_self_size;
872 }
873
874 static size_t rtnl_xdp_size(void)
875 {
876 size_t xdp_size = nla_total_size(0) + /* nest IFLA_XDP */
877 nla_total_size(1) + /* XDP_ATTACHED */
878 nla_total_size(4); /* XDP_PROG_ID */
879
880 return xdp_size;
881 }
882
883 static noinline size_t if_nlmsg_size(const struct net_device *dev,
884 u32 ext_filter_mask)
885 {
886 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
887 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
888 + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
889 + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
890 + nla_total_size_64bit(sizeof(struct rtnl_link_ifmap))
891 + nla_total_size(sizeof(struct rtnl_link_stats))
892 + nla_total_size_64bit(sizeof(struct rtnl_link_stats64))
893 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
894 + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
895 + nla_total_size(4) /* IFLA_TXQLEN */
896 + nla_total_size(4) /* IFLA_WEIGHT */
897 + nla_total_size(4) /* IFLA_MTU */
898 + nla_total_size(4) /* IFLA_LINK */
899 + nla_total_size(4) /* IFLA_MASTER */
900 + nla_total_size(1) /* IFLA_CARRIER */
901 + nla_total_size(4) /* IFLA_PROMISCUITY */
902 + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
903 + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
904 + nla_total_size(4) /* IFLA_GSO_MAX_SEGS */
905 + nla_total_size(4) /* IFLA_GSO_MAX_SIZE */
906 + nla_total_size(1) /* IFLA_OPERSTATE */
907 + nla_total_size(1) /* IFLA_LINKMODE */
908 + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
909 + nla_total_size(4) /* IFLA_LINK_NETNSID */
910 + nla_total_size(4) /* IFLA_GROUP */
911 + nla_total_size(ext_filter_mask
912 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
913 + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
914 + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
915 + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
916 + rtnl_link_get_af_size(dev, ext_filter_mask) /* IFLA_AF_SPEC */
917 + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */
918 + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_SWITCH_ID */
919 + nla_total_size(IFNAMSIZ) /* IFLA_PHYS_PORT_NAME */
920 + rtnl_xdp_size() /* IFLA_XDP */
921 + nla_total_size(4) /* IFLA_EVENT */
922 + nla_total_size(4) /* IFLA_NEW_NETNSID */
923 + nla_total_size(4) /* IFLA_NEW_IFINDEX */
924 + nla_total_size(1) /* IFLA_PROTO_DOWN */
925 + nla_total_size(4) /* IFLA_IF_NETNSID */
926 + nla_total_size(4) /* IFLA_CARRIER_UP_COUNT */
927 + nla_total_size(4) /* IFLA_CARRIER_DOWN_COUNT */
928 + 0;
929 }
930
931 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
932 {
933 struct nlattr *vf_ports;
934 struct nlattr *vf_port;
935 int vf;
936 int err;
937
938 vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
939 if (!vf_ports)
940 return -EMSGSIZE;
941
942 for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
943 vf_port = nla_nest_start(skb, IFLA_VF_PORT);
944 if (!vf_port)
945 goto nla_put_failure;
946 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
947 goto nla_put_failure;
948 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
949 if (err == -EMSGSIZE)
950 goto nla_put_failure;
951 if (err) {
952 nla_nest_cancel(skb, vf_port);
953 continue;
954 }
955 nla_nest_end(skb, vf_port);
956 }
957
958 nla_nest_end(skb, vf_ports);
959
960 return 0;
961
962 nla_put_failure:
963 nla_nest_cancel(skb, vf_ports);
964 return -EMSGSIZE;
965 }
966
967 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
968 {
969 struct nlattr *port_self;
970 int err;
971
972 port_self = nla_nest_start(skb, IFLA_PORT_SELF);
973 if (!port_self)
974 return -EMSGSIZE;
975
976 err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
977 if (err) {
978 nla_nest_cancel(skb, port_self);
979 return (err == -EMSGSIZE) ? err : 0;
980 }
981
982 nla_nest_end(skb, port_self);
983
984 return 0;
985 }
986
987 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
988 u32 ext_filter_mask)
989 {
990 int err;
991
992 if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
993 !(ext_filter_mask & RTEXT_FILTER_VF))
994 return 0;
995
996 err = rtnl_port_self_fill(skb, dev);
997 if (err)
998 return err;
999
1000 if (dev_num_vf(dev->dev.parent)) {
1001 err = rtnl_vf_ports_fill(skb, dev);
1002 if (err)
1003 return err;
1004 }
1005
1006 return 0;
1007 }
1008
1009 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
1010 {
1011 int err;
1012 struct netdev_phys_item_id ppid;
1013
1014 err = dev_get_phys_port_id(dev, &ppid);
1015 if (err) {
1016 if (err == -EOPNOTSUPP)
1017 return 0;
1018 return err;
1019 }
1020
1021 if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
1022 return -EMSGSIZE;
1023
1024 return 0;
1025 }
1026
1027 static int rtnl_phys_port_name_fill(struct sk_buff *skb, struct net_device *dev)
1028 {
1029 char name[IFNAMSIZ];
1030 int err;
1031
1032 err = dev_get_phys_port_name(dev, name, sizeof(name));
1033 if (err) {
1034 if (err == -EOPNOTSUPP)
1035 return 0;
1036 return err;
1037 }
1038
1039 if (nla_put_string(skb, IFLA_PHYS_PORT_NAME, name))
1040 return -EMSGSIZE;
1041
1042 return 0;
1043 }
1044
1045 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev)
1046 {
1047 int err;
1048 struct switchdev_attr attr = {
1049 .orig_dev = dev,
1050 .id = SWITCHDEV_ATTR_ID_PORT_PARENT_ID,
1051 .flags = SWITCHDEV_F_NO_RECURSE,
1052 };
1053
1054 err = switchdev_port_attr_get(dev, &attr);
1055 if (err) {
1056 if (err == -EOPNOTSUPP)
1057 return 0;
1058 return err;
1059 }
1060
1061 if (nla_put(skb, IFLA_PHYS_SWITCH_ID, attr.u.ppid.id_len,
1062 attr.u.ppid.id))
1063 return -EMSGSIZE;
1064
1065 return 0;
1066 }
1067
1068 static noinline_for_stack int rtnl_fill_stats(struct sk_buff *skb,
1069 struct net_device *dev)
1070 {
1071 struct rtnl_link_stats64 *sp;
1072 struct nlattr *attr;
1073
1074 attr = nla_reserve_64bit(skb, IFLA_STATS64,
1075 sizeof(struct rtnl_link_stats64), IFLA_PAD);
1076 if (!attr)
1077 return -EMSGSIZE;
1078
1079 sp = nla_data(attr);
1080 dev_get_stats(dev, sp);
1081
1082 attr = nla_reserve(skb, IFLA_STATS,
1083 sizeof(struct rtnl_link_stats));
1084 if (!attr)
1085 return -EMSGSIZE;
1086
1087 copy_rtnl_link_stats(nla_data(attr), sp);
1088
1089 return 0;
1090 }
1091
1092 static noinline_for_stack int rtnl_fill_vfinfo(struct sk_buff *skb,
1093 struct net_device *dev,
1094 int vfs_num,
1095 struct nlattr *vfinfo)
1096 {
1097 struct ifla_vf_rss_query_en vf_rss_query_en;
1098 struct nlattr *vf, *vfstats, *vfvlanlist;
1099 struct ifla_vf_link_state vf_linkstate;
1100 struct ifla_vf_vlan_info vf_vlan_info;
1101 struct ifla_vf_spoofchk vf_spoofchk;
1102 struct ifla_vf_tx_rate vf_tx_rate;
1103 struct ifla_vf_stats vf_stats;
1104 struct ifla_vf_trust vf_trust;
1105 struct ifla_vf_vlan vf_vlan;
1106 struct ifla_vf_rate vf_rate;
1107 struct ifla_vf_mac vf_mac;
1108 struct ifla_vf_info ivi;
1109
1110 memset(&ivi, 0, sizeof(ivi));
1111
1112 /* Not all SR-IOV capable drivers support the
1113 * spoofcheck and "RSS query enable" query. Preset to
1114 * -1 so the user space tool can detect that the driver
1115 * didn't report anything.
1116 */
1117 ivi.spoofchk = -1;
1118 ivi.rss_query_en = -1;
1119 ivi.trusted = -1;
1120 /* The default value for VF link state is "auto"
1121 * IFLA_VF_LINK_STATE_AUTO which equals zero
1122 */
1123 ivi.linkstate = 0;
1124 /* VLAN Protocol by default is 802.1Q */
1125 ivi.vlan_proto = htons(ETH_P_8021Q);
1126 if (dev->netdev_ops->ndo_get_vf_config(dev, vfs_num, &ivi))
1127 return 0;
1128
1129 memset(&vf_vlan_info, 0, sizeof(vf_vlan_info));
1130
1131 vf_mac.vf =
1132 vf_vlan.vf =
1133 vf_vlan_info.vf =
1134 vf_rate.vf =
1135 vf_tx_rate.vf =
1136 vf_spoofchk.vf =
1137 vf_linkstate.vf =
1138 vf_rss_query_en.vf =
1139 vf_trust.vf = ivi.vf;
1140
1141 memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1142 vf_vlan.vlan = ivi.vlan;
1143 vf_vlan.qos = ivi.qos;
1144 vf_vlan_info.vlan = ivi.vlan;
1145 vf_vlan_info.qos = ivi.qos;
1146 vf_vlan_info.vlan_proto = ivi.vlan_proto;
1147 vf_tx_rate.rate = ivi.max_tx_rate;
1148 vf_rate.min_tx_rate = ivi.min_tx_rate;
1149 vf_rate.max_tx_rate = ivi.max_tx_rate;
1150 vf_spoofchk.setting = ivi.spoofchk;
1151 vf_linkstate.link_state = ivi.linkstate;
1152 vf_rss_query_en.setting = ivi.rss_query_en;
1153 vf_trust.setting = ivi.trusted;
1154 vf = nla_nest_start(skb, IFLA_VF_INFO);
1155 if (!vf)
1156 goto nla_put_vfinfo_failure;
1157 if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1158 nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1159 nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1160 &vf_rate) ||
1161 nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1162 &vf_tx_rate) ||
1163 nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1164 &vf_spoofchk) ||
1165 nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1166 &vf_linkstate) ||
1167 nla_put(skb, IFLA_VF_RSS_QUERY_EN,
1168 sizeof(vf_rss_query_en),
1169 &vf_rss_query_en) ||
1170 nla_put(skb, IFLA_VF_TRUST,
1171 sizeof(vf_trust), &vf_trust))
1172 goto nla_put_vf_failure;
1173 vfvlanlist = nla_nest_start(skb, IFLA_VF_VLAN_LIST);
1174 if (!vfvlanlist)
1175 goto nla_put_vf_failure;
1176 if (nla_put(skb, IFLA_VF_VLAN_INFO, sizeof(vf_vlan_info),
1177 &vf_vlan_info)) {
1178 nla_nest_cancel(skb, vfvlanlist);
1179 goto nla_put_vf_failure;
1180 }
1181 nla_nest_end(skb, vfvlanlist);
1182 memset(&vf_stats, 0, sizeof(vf_stats));
1183 if (dev->netdev_ops->ndo_get_vf_stats)
1184 dev->netdev_ops->ndo_get_vf_stats(dev, vfs_num,
1185 &vf_stats);
1186 vfstats = nla_nest_start(skb, IFLA_VF_STATS);
1187 if (!vfstats)
1188 goto nla_put_vf_failure;
1189 if (nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_PACKETS,
1190 vf_stats.rx_packets, IFLA_VF_STATS_PAD) ||
1191 nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_PACKETS,
1192 vf_stats.tx_packets, IFLA_VF_STATS_PAD) ||
1193 nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_BYTES,
1194 vf_stats.rx_bytes, IFLA_VF_STATS_PAD) ||
1195 nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_BYTES,
1196 vf_stats.tx_bytes, IFLA_VF_STATS_PAD) ||
1197 nla_put_u64_64bit(skb, IFLA_VF_STATS_BROADCAST,
1198 vf_stats.broadcast, IFLA_VF_STATS_PAD) ||
1199 nla_put_u64_64bit(skb, IFLA_VF_STATS_MULTICAST,
1200 vf_stats.multicast, IFLA_VF_STATS_PAD)) {
1201 nla_nest_cancel(skb, vfstats);
1202 goto nla_put_vf_failure;
1203 }
1204 nla_nest_end(skb, vfstats);
1205 nla_nest_end(skb, vf);
1206 return 0;
1207
1208 nla_put_vf_failure:
1209 nla_nest_cancel(skb, vf);
1210 nla_put_vfinfo_failure:
1211 nla_nest_cancel(skb, vfinfo);
1212 return -EMSGSIZE;
1213 }
1214
1215 static noinline_for_stack int rtnl_fill_vf(struct sk_buff *skb,
1216 struct net_device *dev,
1217 u32 ext_filter_mask)
1218 {
1219 struct nlattr *vfinfo;
1220 int i, num_vfs;
1221
1222 if (!dev->dev.parent || ((ext_filter_mask & RTEXT_FILTER_VF) == 0))
1223 return 0;
1224
1225 num_vfs = dev_num_vf(dev->dev.parent);
1226 if (nla_put_u32(skb, IFLA_NUM_VF, num_vfs))
1227 return -EMSGSIZE;
1228
1229 if (!dev->netdev_ops->ndo_get_vf_config)
1230 return 0;
1231
1232 vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
1233 if (!vfinfo)
1234 return -EMSGSIZE;
1235
1236 for (i = 0; i < num_vfs; i++) {
1237 if (rtnl_fill_vfinfo(skb, dev, i, vfinfo))
1238 return -EMSGSIZE;
1239 }
1240
1241 nla_nest_end(skb, vfinfo);
1242 return 0;
1243 }
1244
1245 static int rtnl_fill_link_ifmap(struct sk_buff *skb, struct net_device *dev)
1246 {
1247 struct rtnl_link_ifmap map;
1248
1249 memset(&map, 0, sizeof(map));
1250 map.mem_start = dev->mem_start;
1251 map.mem_end = dev->mem_end;
1252 map.base_addr = dev->base_addr;
1253 map.irq = dev->irq;
1254 map.dma = dev->dma;
1255 map.port = dev->if_port;
1256
1257 if (nla_put_64bit(skb, IFLA_MAP, sizeof(map), &map, IFLA_PAD))
1258 return -EMSGSIZE;
1259
1260 return 0;
1261 }
1262
1263 static u8 rtnl_xdp_attached_mode(struct net_device *dev, u32 *prog_id)
1264 {
1265 const struct net_device_ops *ops = dev->netdev_ops;
1266 const struct bpf_prog *generic_xdp_prog;
1267
1268 ASSERT_RTNL();
1269
1270 *prog_id = 0;
1271 generic_xdp_prog = rtnl_dereference(dev->xdp_prog);
1272 if (generic_xdp_prog) {
1273 *prog_id = generic_xdp_prog->aux->id;
1274 return XDP_ATTACHED_SKB;
1275 }
1276 if (!ops->ndo_bpf)
1277 return XDP_ATTACHED_NONE;
1278
1279 return __dev_xdp_attached(dev, ops->ndo_bpf, prog_id);
1280 }
1281
1282 static int rtnl_xdp_fill(struct sk_buff *skb, struct net_device *dev)
1283 {
1284 struct nlattr *xdp;
1285 u32 prog_id;
1286 int err;
1287
1288 xdp = nla_nest_start(skb, IFLA_XDP);
1289 if (!xdp)
1290 return -EMSGSIZE;
1291
1292 err = nla_put_u8(skb, IFLA_XDP_ATTACHED,
1293 rtnl_xdp_attached_mode(dev, &prog_id));
1294 if (err)
1295 goto err_cancel;
1296
1297 if (prog_id) {
1298 err = nla_put_u32(skb, IFLA_XDP_PROG_ID, prog_id);
1299 if (err)
1300 goto err_cancel;
1301 }
1302
1303 nla_nest_end(skb, xdp);
1304 return 0;
1305
1306 err_cancel:
1307 nla_nest_cancel(skb, xdp);
1308 return err;
1309 }
1310
1311 static u32 rtnl_get_event(unsigned long event)
1312 {
1313 u32 rtnl_event_type = IFLA_EVENT_NONE;
1314
1315 switch (event) {
1316 case NETDEV_REBOOT:
1317 rtnl_event_type = IFLA_EVENT_REBOOT;
1318 break;
1319 case NETDEV_FEAT_CHANGE:
1320 rtnl_event_type = IFLA_EVENT_FEATURES;
1321 break;
1322 case NETDEV_BONDING_FAILOVER:
1323 rtnl_event_type = IFLA_EVENT_BONDING_FAILOVER;
1324 break;
1325 case NETDEV_NOTIFY_PEERS:
1326 rtnl_event_type = IFLA_EVENT_NOTIFY_PEERS;
1327 break;
1328 case NETDEV_RESEND_IGMP:
1329 rtnl_event_type = IFLA_EVENT_IGMP_RESEND;
1330 break;
1331 case NETDEV_CHANGEINFODATA:
1332 rtnl_event_type = IFLA_EVENT_BONDING_OPTIONS;
1333 break;
1334 default:
1335 break;
1336 }
1337
1338 return rtnl_event_type;
1339 }
1340
1341 static int put_master_ifindex(struct sk_buff *skb, struct net_device *dev)
1342 {
1343 const struct net_device *upper_dev;
1344 int ret = 0;
1345
1346 rcu_read_lock();
1347
1348 upper_dev = netdev_master_upper_dev_get_rcu(dev);
1349 if (upper_dev)
1350 ret = nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex);
1351
1352 rcu_read_unlock();
1353 return ret;
1354 }
1355
1356 static int nla_put_iflink(struct sk_buff *skb, const struct net_device *dev,
1357 bool force)
1358 {
1359 int ifindex = dev_get_iflink(dev);
1360
1361 if (force || dev->ifindex != ifindex)
1362 return nla_put_u32(skb, IFLA_LINK, ifindex);
1363
1364 return 0;
1365 }
1366
1367 static noinline_for_stack int nla_put_ifalias(struct sk_buff *skb,
1368 struct net_device *dev)
1369 {
1370 char buf[IFALIASZ];
1371 int ret;
1372
1373 ret = dev_get_alias(dev, buf, sizeof(buf));
1374 return ret > 0 ? nla_put_string(skb, IFLA_IFALIAS, buf) : 0;
1375 }
1376
1377 static int rtnl_fill_link_netnsid(struct sk_buff *skb,
1378 const struct net_device *dev,
1379 struct net *src_net, gfp_t gfp)
1380 {
1381 bool put_iflink = false;
1382
1383 if (dev->rtnl_link_ops && dev->rtnl_link_ops->get_link_net) {
1384 struct net *link_net = dev->rtnl_link_ops->get_link_net(dev);
1385
1386 if (!net_eq(dev_net(dev), link_net)) {
1387 int id = peernet2id_alloc(src_net, link_net, gfp);
1388
1389 if (nla_put_s32(skb, IFLA_LINK_NETNSID, id))
1390 return -EMSGSIZE;
1391
1392 put_iflink = true;
1393 }
1394 }
1395
1396 return nla_put_iflink(skb, dev, put_iflink);
1397 }
1398
1399 static int rtnl_fill_link_af(struct sk_buff *skb,
1400 const struct net_device *dev,
1401 u32 ext_filter_mask)
1402 {
1403 const struct rtnl_af_ops *af_ops;
1404 struct nlattr *af_spec;
1405
1406 af_spec = nla_nest_start(skb, IFLA_AF_SPEC);
1407 if (!af_spec)
1408 return -EMSGSIZE;
1409
1410 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
1411 struct nlattr *af;
1412 int err;
1413
1414 if (!af_ops->fill_link_af)
1415 continue;
1416
1417 af = nla_nest_start(skb, af_ops->family);
1418 if (!af)
1419 return -EMSGSIZE;
1420
1421 err = af_ops->fill_link_af(skb, dev, ext_filter_mask);
1422 /*
1423 * Caller may return ENODATA to indicate that there
1424 * was no data to be dumped. This is not an error, it
1425 * means we should trim the attribute header and
1426 * continue.
1427 */
1428 if (err == -ENODATA)
1429 nla_nest_cancel(skb, af);
1430 else if (err < 0)
1431 return -EMSGSIZE;
1432
1433 nla_nest_end(skb, af);
1434 }
1435
1436 nla_nest_end(skb, af_spec);
1437 return 0;
1438 }
1439
1440 static int rtnl_fill_ifinfo(struct sk_buff *skb,
1441 struct net_device *dev, struct net *src_net,
1442 int type, u32 pid, u32 seq, u32 change,
1443 unsigned int flags, u32 ext_filter_mask,
1444 u32 event, int *new_nsid, int new_ifindex,
1445 int tgt_netnsid, gfp_t gfp)
1446 {
1447 struct ifinfomsg *ifm;
1448 struct nlmsghdr *nlh;
1449
1450 ASSERT_RTNL();
1451 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
1452 if (nlh == NULL)
1453 return -EMSGSIZE;
1454
1455 ifm = nlmsg_data(nlh);
1456 ifm->ifi_family = AF_UNSPEC;
1457 ifm->__ifi_pad = 0;
1458 ifm->ifi_type = dev->type;
1459 ifm->ifi_index = dev->ifindex;
1460 ifm->ifi_flags = dev_get_flags(dev);
1461 ifm->ifi_change = change;
1462
1463 if (tgt_netnsid >= 0 && nla_put_s32(skb, IFLA_IF_NETNSID, tgt_netnsid))
1464 goto nla_put_failure;
1465
1466 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
1467 nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
1468 nla_put_u8(skb, IFLA_OPERSTATE,
1469 netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
1470 nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
1471 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
1472 nla_put_u32(skb, IFLA_GROUP, dev->group) ||
1473 nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
1474 nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
1475 nla_put_u32(skb, IFLA_GSO_MAX_SEGS, dev->gso_max_segs) ||
1476 nla_put_u32(skb, IFLA_GSO_MAX_SIZE, dev->gso_max_size) ||
1477 #ifdef CONFIG_RPS
1478 nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1479 #endif
1480 put_master_ifindex(skb, dev) ||
1481 nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1482 (dev->qdisc &&
1483 nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
1484 nla_put_ifalias(skb, dev) ||
1485 nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1486 atomic_read(&dev->carrier_up_count) +
1487 atomic_read(&dev->carrier_down_count)) ||
1488 nla_put_u8(skb, IFLA_PROTO_DOWN, dev->proto_down) ||
1489 nla_put_u32(skb, IFLA_CARRIER_UP_COUNT,
1490 atomic_read(&dev->carrier_up_count)) ||
1491 nla_put_u32(skb, IFLA_CARRIER_DOWN_COUNT,
1492 atomic_read(&dev->carrier_down_count)))
1493 goto nla_put_failure;
1494
1495 if (event != IFLA_EVENT_NONE) {
1496 if (nla_put_u32(skb, IFLA_EVENT, event))
1497 goto nla_put_failure;
1498 }
1499
1500 if (rtnl_fill_link_ifmap(skb, dev))
1501 goto nla_put_failure;
1502
1503 if (dev->addr_len) {
1504 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1505 nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1506 goto nla_put_failure;
1507 }
1508
1509 if (rtnl_phys_port_id_fill(skb, dev))
1510 goto nla_put_failure;
1511
1512 if (rtnl_phys_port_name_fill(skb, dev))
1513 goto nla_put_failure;
1514
1515 if (rtnl_phys_switch_id_fill(skb, dev))
1516 goto nla_put_failure;
1517
1518 if (rtnl_fill_stats(skb, dev))
1519 goto nla_put_failure;
1520
1521 if (rtnl_fill_vf(skb, dev, ext_filter_mask))
1522 goto nla_put_failure;
1523
1524 if (rtnl_port_fill(skb, dev, ext_filter_mask))
1525 goto nla_put_failure;
1526
1527 if (rtnl_xdp_fill(skb, dev))
1528 goto nla_put_failure;
1529
1530 if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1531 if (rtnl_link_fill(skb, dev) < 0)
1532 goto nla_put_failure;
1533 }
1534
1535 if (rtnl_fill_link_netnsid(skb, dev, src_net, gfp))
1536 goto nla_put_failure;
1537
1538 if (new_nsid &&
1539 nla_put_s32(skb, IFLA_NEW_NETNSID, *new_nsid) < 0)
1540 goto nla_put_failure;
1541 if (new_ifindex &&
1542 nla_put_s32(skb, IFLA_NEW_IFINDEX, new_ifindex) < 0)
1543 goto nla_put_failure;
1544
1545
1546 rcu_read_lock();
1547 if (rtnl_fill_link_af(skb, dev, ext_filter_mask))
1548 goto nla_put_failure_rcu;
1549 rcu_read_unlock();
1550
1551 nlmsg_end(skb, nlh);
1552 return 0;
1553
1554 nla_put_failure_rcu:
1555 rcu_read_unlock();
1556 nla_put_failure:
1557 nlmsg_cancel(skb, nlh);
1558 return -EMSGSIZE;
1559 }
1560
1561 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1562 [IFLA_IFNAME] = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1563 [IFLA_ADDRESS] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1564 [IFLA_BROADCAST] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1565 [IFLA_MAP] = { .len = sizeof(struct rtnl_link_ifmap) },
1566 [IFLA_MTU] = { .type = NLA_U32 },
1567 [IFLA_LINK] = { .type = NLA_U32 },
1568 [IFLA_MASTER] = { .type = NLA_U32 },
1569 [IFLA_CARRIER] = { .type = NLA_U8 },
1570 [IFLA_TXQLEN] = { .type = NLA_U32 },
1571 [IFLA_WEIGHT] = { .type = NLA_U32 },
1572 [IFLA_OPERSTATE] = { .type = NLA_U8 },
1573 [IFLA_LINKMODE] = { .type = NLA_U8 },
1574 [IFLA_LINKINFO] = { .type = NLA_NESTED },
1575 [IFLA_NET_NS_PID] = { .type = NLA_U32 },
1576 [IFLA_NET_NS_FD] = { .type = NLA_U32 },
1577 /* IFLA_IFALIAS is a string, but policy is set to NLA_BINARY to
1578 * allow 0-length string (needed to remove an alias).
1579 */
1580 [IFLA_IFALIAS] = { .type = NLA_BINARY, .len = IFALIASZ - 1 },
1581 [IFLA_VFINFO_LIST] = {. type = NLA_NESTED },
1582 [IFLA_VF_PORTS] = { .type = NLA_NESTED },
1583 [IFLA_PORT_SELF] = { .type = NLA_NESTED },
1584 [IFLA_AF_SPEC] = { .type = NLA_NESTED },
1585 [IFLA_EXT_MASK] = { .type = NLA_U32 },
1586 [IFLA_PROMISCUITY] = { .type = NLA_U32 },
1587 [IFLA_NUM_TX_QUEUES] = { .type = NLA_U32 },
1588 [IFLA_NUM_RX_QUEUES] = { .type = NLA_U32 },
1589 [IFLA_PHYS_PORT_ID] = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1590 [IFLA_CARRIER_CHANGES] = { .type = NLA_U32 }, /* ignored */
1591 [IFLA_PHYS_SWITCH_ID] = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1592 [IFLA_LINK_NETNSID] = { .type = NLA_S32 },
1593 [IFLA_PROTO_DOWN] = { .type = NLA_U8 },
1594 [IFLA_XDP] = { .type = NLA_NESTED },
1595 [IFLA_EVENT] = { .type = NLA_U32 },
1596 [IFLA_GROUP] = { .type = NLA_U32 },
1597 [IFLA_IF_NETNSID] = { .type = NLA_S32 },
1598 [IFLA_CARRIER_UP_COUNT] = { .type = NLA_U32 },
1599 [IFLA_CARRIER_DOWN_COUNT] = { .type = NLA_U32 },
1600 };
1601
1602 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1603 [IFLA_INFO_KIND] = { .type = NLA_STRING },
1604 [IFLA_INFO_DATA] = { .type = NLA_NESTED },
1605 [IFLA_INFO_SLAVE_KIND] = { .type = NLA_STRING },
1606 [IFLA_INFO_SLAVE_DATA] = { .type = NLA_NESTED },
1607 };
1608
1609 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1610 [IFLA_VF_MAC] = { .len = sizeof(struct ifla_vf_mac) },
1611 [IFLA_VF_VLAN] = { .len = sizeof(struct ifla_vf_vlan) },
1612 [IFLA_VF_VLAN_LIST] = { .type = NLA_NESTED },
1613 [IFLA_VF_TX_RATE] = { .len = sizeof(struct ifla_vf_tx_rate) },
1614 [IFLA_VF_SPOOFCHK] = { .len = sizeof(struct ifla_vf_spoofchk) },
1615 [IFLA_VF_RATE] = { .len = sizeof(struct ifla_vf_rate) },
1616 [IFLA_VF_LINK_STATE] = { .len = sizeof(struct ifla_vf_link_state) },
1617 [IFLA_VF_RSS_QUERY_EN] = { .len = sizeof(struct ifla_vf_rss_query_en) },
1618 [IFLA_VF_STATS] = { .type = NLA_NESTED },
1619 [IFLA_VF_TRUST] = { .len = sizeof(struct ifla_vf_trust) },
1620 [IFLA_VF_IB_NODE_GUID] = { .len = sizeof(struct ifla_vf_guid) },
1621 [IFLA_VF_IB_PORT_GUID] = { .len = sizeof(struct ifla_vf_guid) },
1622 };
1623
1624 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1625 [IFLA_PORT_VF] = { .type = NLA_U32 },
1626 [IFLA_PORT_PROFILE] = { .type = NLA_STRING,
1627 .len = PORT_PROFILE_MAX },
1628 [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1629 .len = PORT_UUID_MAX },
1630 [IFLA_PORT_HOST_UUID] = { .type = NLA_STRING,
1631 .len = PORT_UUID_MAX },
1632 [IFLA_PORT_REQUEST] = { .type = NLA_U8, },
1633 [IFLA_PORT_RESPONSE] = { .type = NLA_U16, },
1634
1635 /* Unused, but we need to keep it here since user space could
1636 * fill it. It's also broken with regard to NLA_BINARY use in
1637 * combination with structs.
1638 */
1639 [IFLA_PORT_VSI_TYPE] = { .type = NLA_BINARY,
1640 .len = sizeof(struct ifla_port_vsi) },
1641 };
1642
1643 static const struct nla_policy ifla_xdp_policy[IFLA_XDP_MAX + 1] = {
1644 [IFLA_XDP_FD] = { .type = NLA_S32 },
1645 [IFLA_XDP_ATTACHED] = { .type = NLA_U8 },
1646 [IFLA_XDP_FLAGS] = { .type = NLA_U32 },
1647 [IFLA_XDP_PROG_ID] = { .type = NLA_U32 },
1648 };
1649
1650 static const struct rtnl_link_ops *linkinfo_to_kind_ops(const struct nlattr *nla)
1651 {
1652 const struct rtnl_link_ops *ops = NULL;
1653 struct nlattr *linfo[IFLA_INFO_MAX + 1];
1654
1655 if (nla_parse_nested(linfo, IFLA_INFO_MAX, nla,
1656 ifla_info_policy, NULL) < 0)
1657 return NULL;
1658
1659 if (linfo[IFLA_INFO_KIND]) {
1660 char kind[MODULE_NAME_LEN];
1661
1662 nla_strlcpy(kind, linfo[IFLA_INFO_KIND], sizeof(kind));
1663 ops = rtnl_link_ops_get(kind);
1664 }
1665
1666 return ops;
1667 }
1668
1669 static bool link_master_filtered(struct net_device *dev, int master_idx)
1670 {
1671 struct net_device *master;
1672
1673 if (!master_idx)
1674 return false;
1675
1676 master = netdev_master_upper_dev_get(dev);
1677 if (!master || master->ifindex != master_idx)
1678 return true;
1679
1680 return false;
1681 }
1682
1683 static bool link_kind_filtered(const struct net_device *dev,
1684 const struct rtnl_link_ops *kind_ops)
1685 {
1686 if (kind_ops && dev->rtnl_link_ops != kind_ops)
1687 return true;
1688
1689 return false;
1690 }
1691
1692 static bool link_dump_filtered(struct net_device *dev,
1693 int master_idx,
1694 const struct rtnl_link_ops *kind_ops)
1695 {
1696 if (link_master_filtered(dev, master_idx) ||
1697 link_kind_filtered(dev, kind_ops))
1698 return true;
1699
1700 return false;
1701 }
1702
1703 static struct net *get_target_net(struct sock *sk, int netnsid)
1704 {
1705 struct net *net;
1706
1707 net = get_net_ns_by_id(sock_net(sk), netnsid);
1708 if (!net)
1709 return ERR_PTR(-EINVAL);
1710
1711 /* For now, the caller is required to have CAP_NET_ADMIN in
1712 * the user namespace owning the target net ns.
1713 */
1714 if (!sk_ns_capable(sk, net->user_ns, CAP_NET_ADMIN)) {
1715 put_net(net);
1716 return ERR_PTR(-EACCES);
1717 }
1718 return net;
1719 }
1720
1721 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1722 {
1723 struct net *net = sock_net(skb->sk);
1724 struct net *tgt_net = net;
1725 int h, s_h;
1726 int idx = 0, s_idx;
1727 struct net_device *dev;
1728 struct hlist_head *head;
1729 struct nlattr *tb[IFLA_MAX+1];
1730 u32 ext_filter_mask = 0;
1731 const struct rtnl_link_ops *kind_ops = NULL;
1732 unsigned int flags = NLM_F_MULTI;
1733 int master_idx = 0;
1734 int netnsid = -1;
1735 int err;
1736 int hdrlen;
1737
1738 s_h = cb->args[0];
1739 s_idx = cb->args[1];
1740
1741 /* A hack to preserve kernel<->userspace interface.
1742 * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
1743 * However, before Linux v3.9 the code here assumed rtgenmsg and that's
1744 * what iproute2 < v3.9.0 used.
1745 * We can detect the old iproute2. Even including the IFLA_EXT_MASK
1746 * attribute, its netlink message is shorter than struct ifinfomsg.
1747 */
1748 hdrlen = nlmsg_len(cb->nlh) < sizeof(struct ifinfomsg) ?
1749 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
1750
1751 if (nlmsg_parse(cb->nlh, hdrlen, tb, IFLA_MAX,
1752 ifla_policy, NULL) >= 0) {
1753 if (tb[IFLA_IF_NETNSID]) {
1754 netnsid = nla_get_s32(tb[IFLA_IF_NETNSID]);
1755 tgt_net = get_target_net(skb->sk, netnsid);
1756 if (IS_ERR(tgt_net))
1757 return PTR_ERR(tgt_net);
1758 }
1759
1760 if (tb[IFLA_EXT_MASK])
1761 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1762
1763 if (tb[IFLA_MASTER])
1764 master_idx = nla_get_u32(tb[IFLA_MASTER]);
1765
1766 if (tb[IFLA_LINKINFO])
1767 kind_ops = linkinfo_to_kind_ops(tb[IFLA_LINKINFO]);
1768
1769 if (master_idx || kind_ops)
1770 flags |= NLM_F_DUMP_FILTERED;
1771 }
1772
1773 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1774 idx = 0;
1775 head = &tgt_net->dev_index_head[h];
1776 hlist_for_each_entry(dev, head, index_hlist) {
1777 if (link_dump_filtered(dev, master_idx, kind_ops))
1778 goto cont;
1779 if (idx < s_idx)
1780 goto cont;
1781 err = rtnl_fill_ifinfo(skb, dev, net,
1782 RTM_NEWLINK,
1783 NETLINK_CB(cb->skb).portid,
1784 cb->nlh->nlmsg_seq, 0,
1785 flags,
1786 ext_filter_mask, 0, NULL, 0,
1787 netnsid, GFP_KERNEL);
1788
1789 if (err < 0) {
1790 if (likely(skb->len))
1791 goto out;
1792
1793 goto out_err;
1794 }
1795 cont:
1796 idx++;
1797 }
1798 }
1799 out:
1800 err = skb->len;
1801 out_err:
1802 cb->args[1] = idx;
1803 cb->args[0] = h;
1804 cb->seq = net->dev_base_seq;
1805 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1806 if (netnsid >= 0)
1807 put_net(tgt_net);
1808
1809 return err;
1810 }
1811
1812 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len,
1813 struct netlink_ext_ack *exterr)
1814 {
1815 return nla_parse(tb, IFLA_MAX, head, len, ifla_policy, exterr);
1816 }
1817 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
1818
1819 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1820 {
1821 struct net *net;
1822 /* Examine the link attributes and figure out which
1823 * network namespace we are talking about.
1824 */
1825 if (tb[IFLA_NET_NS_PID])
1826 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1827 else if (tb[IFLA_NET_NS_FD])
1828 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1829 else
1830 net = get_net(src_net);
1831 return net;
1832 }
1833 EXPORT_SYMBOL(rtnl_link_get_net);
1834
1835 /* Figure out which network namespace we are talking about by
1836 * examining the link attributes in the following order:
1837 *
1838 * 1. IFLA_NET_NS_PID
1839 * 2. IFLA_NET_NS_FD
1840 * 3. IFLA_IF_NETNSID
1841 */
1842 static struct net *rtnl_link_get_net_by_nlattr(struct net *src_net,
1843 struct nlattr *tb[])
1844 {
1845 struct net *net;
1846
1847 if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD])
1848 return rtnl_link_get_net(src_net, tb);
1849
1850 if (!tb[IFLA_IF_NETNSID])
1851 return get_net(src_net);
1852
1853 net = get_net_ns_by_id(src_net, nla_get_u32(tb[IFLA_IF_NETNSID]));
1854 if (!net)
1855 return ERR_PTR(-EINVAL);
1856
1857 return net;
1858 }
1859
1860 static struct net *rtnl_link_get_net_capable(const struct sk_buff *skb,
1861 struct net *src_net,
1862 struct nlattr *tb[], int cap)
1863 {
1864 struct net *net;
1865
1866 net = rtnl_link_get_net_by_nlattr(src_net, tb);
1867 if (IS_ERR(net))
1868 return net;
1869
1870 if (!netlink_ns_capable(skb, net->user_ns, cap)) {
1871 put_net(net);
1872 return ERR_PTR(-EPERM);
1873 }
1874
1875 return net;
1876 }
1877
1878 /* Verify that rtnetlink requests do not pass additional properties
1879 * potentially referring to different network namespaces.
1880 */
1881 static int rtnl_ensure_unique_netns(struct nlattr *tb[],
1882 struct netlink_ext_ack *extack,
1883 bool netns_id_only)
1884 {
1885
1886 if (netns_id_only) {
1887 if (!tb[IFLA_NET_NS_PID] && !tb[IFLA_NET_NS_FD])
1888 return 0;
1889
1890 NL_SET_ERR_MSG(extack, "specified netns attribute not supported");
1891 return -EOPNOTSUPP;
1892 }
1893
1894 if (tb[IFLA_IF_NETNSID] && (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]))
1895 goto invalid_attr;
1896
1897 if (tb[IFLA_NET_NS_PID] && (tb[IFLA_IF_NETNSID] || tb[IFLA_NET_NS_FD]))
1898 goto invalid_attr;
1899
1900 if (tb[IFLA_NET_NS_FD] && (tb[IFLA_IF_NETNSID] || tb[IFLA_NET_NS_PID]))
1901 goto invalid_attr;
1902
1903 return 0;
1904
1905 invalid_attr:
1906 NL_SET_ERR_MSG(extack, "multiple netns identifying attributes specified");
1907 return -EINVAL;
1908 }
1909
1910 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1911 {
1912 if (dev) {
1913 if (tb[IFLA_ADDRESS] &&
1914 nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1915 return -EINVAL;
1916
1917 if (tb[IFLA_BROADCAST] &&
1918 nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1919 return -EINVAL;
1920 }
1921
1922 if (tb[IFLA_AF_SPEC]) {
1923 struct nlattr *af;
1924 int rem, err;
1925
1926 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1927 const struct rtnl_af_ops *af_ops;
1928
1929 rcu_read_lock();
1930 af_ops = rtnl_af_lookup(nla_type(af));
1931 if (!af_ops) {
1932 rcu_read_unlock();
1933 return -EAFNOSUPPORT;
1934 }
1935
1936 if (!af_ops->set_link_af) {
1937 rcu_read_unlock();
1938 return -EOPNOTSUPP;
1939 }
1940
1941 if (af_ops->validate_link_af) {
1942 err = af_ops->validate_link_af(dev, af);
1943 if (err < 0) {
1944 rcu_read_unlock();
1945 return err;
1946 }
1947 }
1948
1949 rcu_read_unlock();
1950 }
1951 }
1952
1953 return 0;
1954 }
1955
1956 static int handle_infiniband_guid(struct net_device *dev, struct ifla_vf_guid *ivt,
1957 int guid_type)
1958 {
1959 const struct net_device_ops *ops = dev->netdev_ops;
1960
1961 return ops->ndo_set_vf_guid(dev, ivt->vf, ivt->guid, guid_type);
1962 }
1963
1964 static int handle_vf_guid(struct net_device *dev, struct ifla_vf_guid *ivt, int guid_type)
1965 {
1966 if (dev->type != ARPHRD_INFINIBAND)
1967 return -EOPNOTSUPP;
1968
1969 return handle_infiniband_guid(dev, ivt, guid_type);
1970 }
1971
1972 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb)
1973 {
1974 const struct net_device_ops *ops = dev->netdev_ops;
1975 int err = -EINVAL;
1976
1977 if (tb[IFLA_VF_MAC]) {
1978 struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]);
1979
1980 if (ivm->vf >= INT_MAX)
1981 return -EINVAL;
1982 err = -EOPNOTSUPP;
1983 if (ops->ndo_set_vf_mac)
1984 err = ops->ndo_set_vf_mac(dev, ivm->vf,
1985 ivm->mac);
1986 if (err < 0)
1987 return err;
1988 }
1989
1990 if (tb[IFLA_VF_VLAN]) {
1991 struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]);
1992
1993 if (ivv->vf >= INT_MAX)
1994 return -EINVAL;
1995 err = -EOPNOTSUPP;
1996 if (ops->ndo_set_vf_vlan)
1997 err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan,
1998 ivv->qos,
1999 htons(ETH_P_8021Q));
2000 if (err < 0)
2001 return err;
2002 }
2003
2004 if (tb[IFLA_VF_VLAN_LIST]) {
2005 struct ifla_vf_vlan_info *ivvl[MAX_VLAN_LIST_LEN];
2006 struct nlattr *attr;
2007 int rem, len = 0;
2008
2009 err = -EOPNOTSUPP;
2010 if (!ops->ndo_set_vf_vlan)
2011 return err;
2012
2013 nla_for_each_nested(attr, tb[IFLA_VF_VLAN_LIST], rem) {
2014 if (nla_type(attr) != IFLA_VF_VLAN_INFO ||
2015 nla_len(attr) < NLA_HDRLEN) {
2016 return -EINVAL;
2017 }
2018 if (len >= MAX_VLAN_LIST_LEN)
2019 return -EOPNOTSUPP;
2020 ivvl[len] = nla_data(attr);
2021
2022 len++;
2023 }
2024 if (len == 0)
2025 return -EINVAL;
2026
2027 if (ivvl[0]->vf >= INT_MAX)
2028 return -EINVAL;
2029 err = ops->ndo_set_vf_vlan(dev, ivvl[0]->vf, ivvl[0]->vlan,
2030 ivvl[0]->qos, ivvl[0]->vlan_proto);
2031 if (err < 0)
2032 return err;
2033 }
2034
2035 if (tb[IFLA_VF_TX_RATE]) {
2036 struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]);
2037 struct ifla_vf_info ivf;
2038
2039 if (ivt->vf >= INT_MAX)
2040 return -EINVAL;
2041 err = -EOPNOTSUPP;
2042 if (ops->ndo_get_vf_config)
2043 err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf);
2044 if (err < 0)
2045 return err;
2046
2047 err = -EOPNOTSUPP;
2048 if (ops->ndo_set_vf_rate)
2049 err = ops->ndo_set_vf_rate(dev, ivt->vf,
2050 ivf.min_tx_rate,
2051 ivt->rate);
2052 if (err < 0)
2053 return err;
2054 }
2055
2056 if (tb[IFLA_VF_RATE]) {
2057 struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]);
2058
2059 if (ivt->vf >= INT_MAX)
2060 return -EINVAL;
2061 err = -EOPNOTSUPP;
2062 if (ops->ndo_set_vf_rate)
2063 err = ops->ndo_set_vf_rate(dev, ivt->vf,
2064 ivt->min_tx_rate,
2065 ivt->max_tx_rate);
2066 if (err < 0)
2067 return err;
2068 }
2069
2070 if (tb[IFLA_VF_SPOOFCHK]) {
2071 struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]);
2072
2073 if (ivs->vf >= INT_MAX)
2074 return -EINVAL;
2075 err = -EOPNOTSUPP;
2076 if (ops->ndo_set_vf_spoofchk)
2077 err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
2078 ivs->setting);
2079 if (err < 0)
2080 return err;
2081 }
2082
2083 if (tb[IFLA_VF_LINK_STATE]) {
2084 struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]);
2085
2086 if (ivl->vf >= INT_MAX)
2087 return -EINVAL;
2088 err = -EOPNOTSUPP;
2089 if (ops->ndo_set_vf_link_state)
2090 err = ops->ndo_set_vf_link_state(dev, ivl->vf,
2091 ivl->link_state);
2092 if (err < 0)
2093 return err;
2094 }
2095
2096 if (tb[IFLA_VF_RSS_QUERY_EN]) {
2097 struct ifla_vf_rss_query_en *ivrssq_en;
2098
2099 err = -EOPNOTSUPP;
2100 ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]);
2101 if (ivrssq_en->vf >= INT_MAX)
2102 return -EINVAL;
2103 if (ops->ndo_set_vf_rss_query_en)
2104 err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf,
2105 ivrssq_en->setting);
2106 if (err < 0)
2107 return err;
2108 }
2109
2110 if (tb[IFLA_VF_TRUST]) {
2111 struct ifla_vf_trust *ivt = nla_data(tb[IFLA_VF_TRUST]);
2112
2113 if (ivt->vf >= INT_MAX)
2114 return -EINVAL;
2115 err = -EOPNOTSUPP;
2116 if (ops->ndo_set_vf_trust)
2117 err = ops->ndo_set_vf_trust(dev, ivt->vf, ivt->setting);
2118 if (err < 0)
2119 return err;
2120 }
2121
2122 if (tb[IFLA_VF_IB_NODE_GUID]) {
2123 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_NODE_GUID]);
2124
2125 if (ivt->vf >= INT_MAX)
2126 return -EINVAL;
2127 if (!ops->ndo_set_vf_guid)
2128 return -EOPNOTSUPP;
2129 return handle_vf_guid(dev, ivt, IFLA_VF_IB_NODE_GUID);
2130 }
2131
2132 if (tb[IFLA_VF_IB_PORT_GUID]) {
2133 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_PORT_GUID]);
2134
2135 if (ivt->vf >= INT_MAX)
2136 return -EINVAL;
2137 if (!ops->ndo_set_vf_guid)
2138 return -EOPNOTSUPP;
2139
2140 return handle_vf_guid(dev, ivt, IFLA_VF_IB_PORT_GUID);
2141 }
2142
2143 return err;
2144 }
2145
2146 static int do_set_master(struct net_device *dev, int ifindex,
2147 struct netlink_ext_ack *extack)
2148 {
2149 struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
2150 const struct net_device_ops *ops;
2151 int err;
2152
2153 if (upper_dev) {
2154 if (upper_dev->ifindex == ifindex)
2155 return 0;
2156 ops = upper_dev->netdev_ops;
2157 if (ops->ndo_del_slave) {
2158 err = ops->ndo_del_slave(upper_dev, dev);
2159 if (err)
2160 return err;
2161 } else {
2162 return -EOPNOTSUPP;
2163 }
2164 }
2165
2166 if (ifindex) {
2167 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
2168 if (!upper_dev)
2169 return -EINVAL;
2170 ops = upper_dev->netdev_ops;
2171 if (ops->ndo_add_slave) {
2172 err = ops->ndo_add_slave(upper_dev, dev, extack);
2173 if (err)
2174 return err;
2175 } else {
2176 return -EOPNOTSUPP;
2177 }
2178 }
2179 return 0;
2180 }
2181
2182 #define DO_SETLINK_MODIFIED 0x01
2183 /* notify flag means notify + modified. */
2184 #define DO_SETLINK_NOTIFY 0x03
2185 static int do_setlink(const struct sk_buff *skb,
2186 struct net_device *dev, struct ifinfomsg *ifm,
2187 struct netlink_ext_ack *extack,
2188 struct nlattr **tb, char *ifname, int status)
2189 {
2190 const struct net_device_ops *ops = dev->netdev_ops;
2191 int err;
2192
2193 err = validate_linkmsg(dev, tb);
2194 if (err < 0)
2195 return err;
2196
2197 if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD] || tb[IFLA_IF_NETNSID]) {
2198 struct net *net = rtnl_link_get_net_capable(skb, dev_net(dev),
2199 tb, CAP_NET_ADMIN);
2200 if (IS_ERR(net)) {
2201 err = PTR_ERR(net);
2202 goto errout;
2203 }
2204
2205 err = dev_change_net_namespace(dev, net, ifname);
2206 put_net(net);
2207 if (err)
2208 goto errout;
2209 status |= DO_SETLINK_MODIFIED;
2210 }
2211
2212 if (tb[IFLA_MAP]) {
2213 struct rtnl_link_ifmap *u_map;
2214 struct ifmap k_map;
2215
2216 if (!ops->ndo_set_config) {
2217 err = -EOPNOTSUPP;
2218 goto errout;
2219 }
2220
2221 if (!netif_device_present(dev)) {
2222 err = -ENODEV;
2223 goto errout;
2224 }
2225
2226 u_map = nla_data(tb[IFLA_MAP]);
2227 k_map.mem_start = (unsigned long) u_map->mem_start;
2228 k_map.mem_end = (unsigned long) u_map->mem_end;
2229 k_map.base_addr = (unsigned short) u_map->base_addr;
2230 k_map.irq = (unsigned char) u_map->irq;
2231 k_map.dma = (unsigned char) u_map->dma;
2232 k_map.port = (unsigned char) u_map->port;
2233
2234 err = ops->ndo_set_config(dev, &k_map);
2235 if (err < 0)
2236 goto errout;
2237
2238 status |= DO_SETLINK_NOTIFY;
2239 }
2240
2241 if (tb[IFLA_ADDRESS]) {
2242 struct sockaddr *sa;
2243 int len;
2244
2245 len = sizeof(sa_family_t) + max_t(size_t, dev->addr_len,
2246 sizeof(*sa));
2247 sa = kmalloc(len, GFP_KERNEL);
2248 if (!sa) {
2249 err = -ENOMEM;
2250 goto errout;
2251 }
2252 sa->sa_family = dev->type;
2253 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
2254 dev->addr_len);
2255 err = dev_set_mac_address(dev, sa);
2256 kfree(sa);
2257 if (err)
2258 goto errout;
2259 status |= DO_SETLINK_MODIFIED;
2260 }
2261
2262 if (tb[IFLA_MTU]) {
2263 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
2264 if (err < 0)
2265 goto errout;
2266 status |= DO_SETLINK_MODIFIED;
2267 }
2268
2269 if (tb[IFLA_GROUP]) {
2270 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2271 status |= DO_SETLINK_NOTIFY;
2272 }
2273
2274 /*
2275 * Interface selected by interface index but interface
2276 * name provided implies that a name change has been
2277 * requested.
2278 */
2279 if (ifm->ifi_index > 0 && ifname[0]) {
2280 err = dev_change_name(dev, ifname);
2281 if (err < 0)
2282 goto errout;
2283 status |= DO_SETLINK_MODIFIED;
2284 }
2285
2286 if (tb[IFLA_IFALIAS]) {
2287 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
2288 nla_len(tb[IFLA_IFALIAS]));
2289 if (err < 0)
2290 goto errout;
2291 status |= DO_SETLINK_NOTIFY;
2292 }
2293
2294 if (tb[IFLA_BROADCAST]) {
2295 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
2296 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
2297 }
2298
2299 if (ifm->ifi_flags || ifm->ifi_change) {
2300 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
2301 if (err < 0)
2302 goto errout;
2303 }
2304
2305 if (tb[IFLA_MASTER]) {
2306 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]), extack);
2307 if (err)
2308 goto errout;
2309 status |= DO_SETLINK_MODIFIED;
2310 }
2311
2312 if (tb[IFLA_CARRIER]) {
2313 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
2314 if (err)
2315 goto errout;
2316 status |= DO_SETLINK_MODIFIED;
2317 }
2318
2319 if (tb[IFLA_TXQLEN]) {
2320 unsigned int value = nla_get_u32(tb[IFLA_TXQLEN]);
2321 unsigned int orig_len = dev->tx_queue_len;
2322
2323 if (dev->tx_queue_len ^ value) {
2324 dev->tx_queue_len = value;
2325 err = call_netdevice_notifiers(
2326 NETDEV_CHANGE_TX_QUEUE_LEN, dev);
2327 err = notifier_to_errno(err);
2328 if (err) {
2329 dev->tx_queue_len = orig_len;
2330 goto errout;
2331 }
2332 status |= DO_SETLINK_MODIFIED;
2333 }
2334 }
2335
2336 if (tb[IFLA_OPERSTATE])
2337 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2338
2339 if (tb[IFLA_LINKMODE]) {
2340 unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
2341
2342 write_lock_bh(&dev_base_lock);
2343 if (dev->link_mode ^ value)
2344 status |= DO_SETLINK_NOTIFY;
2345 dev->link_mode = value;
2346 write_unlock_bh(&dev_base_lock);
2347 }
2348
2349 if (tb[IFLA_VFINFO_LIST]) {
2350 struct nlattr *vfinfo[IFLA_VF_MAX + 1];
2351 struct nlattr *attr;
2352 int rem;
2353
2354 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
2355 if (nla_type(attr) != IFLA_VF_INFO ||
2356 nla_len(attr) < NLA_HDRLEN) {
2357 err = -EINVAL;
2358 goto errout;
2359 }
2360 err = nla_parse_nested(vfinfo, IFLA_VF_MAX, attr,
2361 ifla_vf_policy, NULL);
2362 if (err < 0)
2363 goto errout;
2364 err = do_setvfinfo(dev, vfinfo);
2365 if (err < 0)
2366 goto errout;
2367 status |= DO_SETLINK_NOTIFY;
2368 }
2369 }
2370 err = 0;
2371
2372 if (tb[IFLA_VF_PORTS]) {
2373 struct nlattr *port[IFLA_PORT_MAX+1];
2374 struct nlattr *attr;
2375 int vf;
2376 int rem;
2377
2378 err = -EOPNOTSUPP;
2379 if (!ops->ndo_set_vf_port)
2380 goto errout;
2381
2382 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
2383 if (nla_type(attr) != IFLA_VF_PORT ||
2384 nla_len(attr) < NLA_HDRLEN) {
2385 err = -EINVAL;
2386 goto errout;
2387 }
2388 err = nla_parse_nested(port, IFLA_PORT_MAX, attr,
2389 ifla_port_policy, NULL);
2390 if (err < 0)
2391 goto errout;
2392 if (!port[IFLA_PORT_VF]) {
2393 err = -EOPNOTSUPP;
2394 goto errout;
2395 }
2396 vf = nla_get_u32(port[IFLA_PORT_VF]);
2397 err = ops->ndo_set_vf_port(dev, vf, port);
2398 if (err < 0)
2399 goto errout;
2400 status |= DO_SETLINK_NOTIFY;
2401 }
2402 }
2403 err = 0;
2404
2405 if (tb[IFLA_PORT_SELF]) {
2406 struct nlattr *port[IFLA_PORT_MAX+1];
2407
2408 err = nla_parse_nested(port, IFLA_PORT_MAX,
2409 tb[IFLA_PORT_SELF], ifla_port_policy,
2410 NULL);
2411 if (err < 0)
2412 goto errout;
2413
2414 err = -EOPNOTSUPP;
2415 if (ops->ndo_set_vf_port)
2416 err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
2417 if (err < 0)
2418 goto errout;
2419 status |= DO_SETLINK_NOTIFY;
2420 }
2421
2422 if (tb[IFLA_AF_SPEC]) {
2423 struct nlattr *af;
2424 int rem;
2425
2426 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2427 const struct rtnl_af_ops *af_ops;
2428
2429 rcu_read_lock();
2430
2431 BUG_ON(!(af_ops = rtnl_af_lookup(nla_type(af))));
2432
2433 err = af_ops->set_link_af(dev, af);
2434 if (err < 0) {
2435 rcu_read_unlock();
2436 goto errout;
2437 }
2438
2439 rcu_read_unlock();
2440 status |= DO_SETLINK_NOTIFY;
2441 }
2442 }
2443 err = 0;
2444
2445 if (tb[IFLA_PROTO_DOWN]) {
2446 err = dev_change_proto_down(dev,
2447 nla_get_u8(tb[IFLA_PROTO_DOWN]));
2448 if (err)
2449 goto errout;
2450 status |= DO_SETLINK_NOTIFY;
2451 }
2452
2453 if (tb[IFLA_XDP]) {
2454 struct nlattr *xdp[IFLA_XDP_MAX + 1];
2455 u32 xdp_flags = 0;
2456
2457 err = nla_parse_nested(xdp, IFLA_XDP_MAX, tb[IFLA_XDP],
2458 ifla_xdp_policy, NULL);
2459 if (err < 0)
2460 goto errout;
2461
2462 if (xdp[IFLA_XDP_ATTACHED] || xdp[IFLA_XDP_PROG_ID]) {
2463 err = -EINVAL;
2464 goto errout;
2465 }
2466
2467 if (xdp[IFLA_XDP_FLAGS]) {
2468 xdp_flags = nla_get_u32(xdp[IFLA_XDP_FLAGS]);
2469 if (xdp_flags & ~XDP_FLAGS_MASK) {
2470 err = -EINVAL;
2471 goto errout;
2472 }
2473 if (hweight32(xdp_flags & XDP_FLAGS_MODES) > 1) {
2474 err = -EINVAL;
2475 goto errout;
2476 }
2477 }
2478
2479 if (xdp[IFLA_XDP_FD]) {
2480 err = dev_change_xdp_fd(dev, extack,
2481 nla_get_s32(xdp[IFLA_XDP_FD]),
2482 xdp_flags);
2483 if (err)
2484 goto errout;
2485 status |= DO_SETLINK_NOTIFY;
2486 }
2487 }
2488
2489 errout:
2490 if (status & DO_SETLINK_MODIFIED) {
2491 if ((status & DO_SETLINK_NOTIFY) == DO_SETLINK_NOTIFY)
2492 netdev_state_change(dev);
2493
2494 if (err < 0)
2495 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",
2496 dev->name);
2497 }
2498
2499 return err;
2500 }
2501
2502 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
2503 struct netlink_ext_ack *extack)
2504 {
2505 struct net *net = sock_net(skb->sk);
2506 struct ifinfomsg *ifm;
2507 struct net_device *dev;
2508 int err;
2509 struct nlattr *tb[IFLA_MAX+1];
2510 char ifname[IFNAMSIZ];
2511
2512 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy,
2513 extack);
2514 if (err < 0)
2515 goto errout;
2516
2517 err = rtnl_ensure_unique_netns(tb, extack, false);
2518 if (err < 0)
2519 goto errout;
2520
2521 if (tb[IFLA_IFNAME])
2522 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2523 else
2524 ifname[0] = '\0';
2525
2526 err = -EINVAL;
2527 ifm = nlmsg_data(nlh);
2528 if (ifm->ifi_index > 0)
2529 dev = __dev_get_by_index(net, ifm->ifi_index);
2530 else if (tb[IFLA_IFNAME])
2531 dev = __dev_get_by_name(net, ifname);
2532 else
2533 goto errout;
2534
2535 if (dev == NULL) {
2536 err = -ENODEV;
2537 goto errout;
2538 }
2539
2540 err = do_setlink(skb, dev, ifm, extack, tb, ifname, 0);
2541 errout:
2542 return err;
2543 }
2544
2545 static int rtnl_group_dellink(const struct net *net, int group)
2546 {
2547 struct net_device *dev, *aux;
2548 LIST_HEAD(list_kill);
2549 bool found = false;
2550
2551 if (!group)
2552 return -EPERM;
2553
2554 for_each_netdev(net, dev) {
2555 if (dev->group == group) {
2556 const struct rtnl_link_ops *ops;
2557
2558 found = true;
2559 ops = dev->rtnl_link_ops;
2560 if (!ops || !ops->dellink)
2561 return -EOPNOTSUPP;
2562 }
2563 }
2564
2565 if (!found)
2566 return -ENODEV;
2567
2568 for_each_netdev_safe(net, dev, aux) {
2569 if (dev->group == group) {
2570 const struct rtnl_link_ops *ops;
2571
2572 ops = dev->rtnl_link_ops;
2573 ops->dellink(dev, &list_kill);
2574 }
2575 }
2576 unregister_netdevice_many(&list_kill);
2577
2578 return 0;
2579 }
2580
2581 int rtnl_delete_link(struct net_device *dev)
2582 {
2583 const struct rtnl_link_ops *ops;
2584 LIST_HEAD(list_kill);
2585
2586 ops = dev->rtnl_link_ops;
2587 if (!ops || !ops->dellink)
2588 return -EOPNOTSUPP;
2589
2590 ops->dellink(dev, &list_kill);
2591 unregister_netdevice_many(&list_kill);
2592
2593 return 0;
2594 }
2595 EXPORT_SYMBOL_GPL(rtnl_delete_link);
2596
2597 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
2598 struct netlink_ext_ack *extack)
2599 {
2600 struct net *net = sock_net(skb->sk);
2601 struct net *tgt_net = net;
2602 struct net_device *dev = NULL;
2603 struct ifinfomsg *ifm;
2604 char ifname[IFNAMSIZ];
2605 struct nlattr *tb[IFLA_MAX+1];
2606 int err;
2607 int netnsid = -1;
2608
2609 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
2610 if (err < 0)
2611 return err;
2612
2613 err = rtnl_ensure_unique_netns(tb, extack, true);
2614 if (err < 0)
2615 return err;
2616
2617 if (tb[IFLA_IFNAME])
2618 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2619
2620 if (tb[IFLA_IF_NETNSID]) {
2621 netnsid = nla_get_s32(tb[IFLA_IF_NETNSID]);
2622 tgt_net = get_target_net(NETLINK_CB(skb).sk, netnsid);
2623 if (IS_ERR(tgt_net))
2624 return PTR_ERR(tgt_net);
2625 }
2626
2627 err = -EINVAL;
2628 ifm = nlmsg_data(nlh);
2629 if (ifm->ifi_index > 0)
2630 dev = __dev_get_by_index(tgt_net, ifm->ifi_index);
2631 else if (tb[IFLA_IFNAME])
2632 dev = __dev_get_by_name(tgt_net, ifname);
2633 else if (tb[IFLA_GROUP])
2634 err = rtnl_group_dellink(tgt_net, nla_get_u32(tb[IFLA_GROUP]));
2635 else
2636 goto out;
2637
2638 if (!dev) {
2639 if (tb[IFLA_IFNAME] || ifm->ifi_index > 0)
2640 err = -ENODEV;
2641
2642 goto out;
2643 }
2644
2645 err = rtnl_delete_link(dev);
2646
2647 out:
2648 if (netnsid >= 0)
2649 put_net(tgt_net);
2650
2651 return err;
2652 }
2653
2654 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
2655 {
2656 unsigned int old_flags;
2657 int err;
2658
2659 old_flags = dev->flags;
2660 if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
2661 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
2662 if (err < 0)
2663 return err;
2664 }
2665
2666 if (dev->rtnl_link_state == RTNL_LINK_INITIALIZED) {
2667 __dev_notify_flags(dev, old_flags, (old_flags ^ dev->flags));
2668 } else {
2669 dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
2670 __dev_notify_flags(dev, old_flags, ~0U);
2671 }
2672 return 0;
2673 }
2674 EXPORT_SYMBOL(rtnl_configure_link);
2675
2676 struct net_device *rtnl_create_link(struct net *net,
2677 const char *ifname, unsigned char name_assign_type,
2678 const struct rtnl_link_ops *ops, struct nlattr *tb[])
2679 {
2680 struct net_device *dev;
2681 unsigned int num_tx_queues = 1;
2682 unsigned int num_rx_queues = 1;
2683
2684 if (tb[IFLA_NUM_TX_QUEUES])
2685 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
2686 else if (ops->get_num_tx_queues)
2687 num_tx_queues = ops->get_num_tx_queues();
2688
2689 if (tb[IFLA_NUM_RX_QUEUES])
2690 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
2691 else if (ops->get_num_rx_queues)
2692 num_rx_queues = ops->get_num_rx_queues();
2693
2694 if (num_tx_queues < 1 || num_tx_queues > 4096)
2695 return ERR_PTR(-EINVAL);
2696
2697 if (num_rx_queues < 1 || num_rx_queues > 4096)
2698 return ERR_PTR(-EINVAL);
2699
2700 dev = alloc_netdev_mqs(ops->priv_size, ifname, name_assign_type,
2701 ops->setup, num_tx_queues, num_rx_queues);
2702 if (!dev)
2703 return ERR_PTR(-ENOMEM);
2704
2705 dev_net_set(dev, net);
2706 dev->rtnl_link_ops = ops;
2707 dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
2708
2709 if (tb[IFLA_MTU]) {
2710 u32 mtu = nla_get_u32(tb[IFLA_MTU]);
2711 int err;
2712
2713 err = dev_validate_mtu(dev, mtu);
2714 if (err) {
2715 free_netdev(dev);
2716 return ERR_PTR(err);
2717 }
2718 dev->mtu = mtu;
2719 }
2720 if (tb[IFLA_ADDRESS]) {
2721 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
2722 nla_len(tb[IFLA_ADDRESS]));
2723 dev->addr_assign_type = NET_ADDR_SET;
2724 }
2725 if (tb[IFLA_BROADCAST])
2726 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
2727 nla_len(tb[IFLA_BROADCAST]));
2728 if (tb[IFLA_TXQLEN])
2729 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
2730 if (tb[IFLA_OPERSTATE])
2731 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2732 if (tb[IFLA_LINKMODE])
2733 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
2734 if (tb[IFLA_GROUP])
2735 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2736
2737 return dev;
2738 }
2739 EXPORT_SYMBOL(rtnl_create_link);
2740
2741 static int rtnl_group_changelink(const struct sk_buff *skb,
2742 struct net *net, int group,
2743 struct ifinfomsg *ifm,
2744 struct netlink_ext_ack *extack,
2745 struct nlattr **tb)
2746 {
2747 struct net_device *dev, *aux;
2748 int err;
2749
2750 for_each_netdev_safe(net, dev, aux) {
2751 if (dev->group == group) {
2752 err = do_setlink(skb, dev, ifm, extack, tb, NULL, 0);
2753 if (err < 0)
2754 return err;
2755 }
2756 }
2757
2758 return 0;
2759 }
2760
2761 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
2762 struct netlink_ext_ack *extack)
2763 {
2764 struct net *net = sock_net(skb->sk);
2765 const struct rtnl_link_ops *ops;
2766 const struct rtnl_link_ops *m_ops = NULL;
2767 struct net_device *dev;
2768 struct net_device *master_dev = NULL;
2769 struct ifinfomsg *ifm;
2770 char kind[MODULE_NAME_LEN];
2771 char ifname[IFNAMSIZ];
2772 struct nlattr *tb[IFLA_MAX+1];
2773 struct nlattr *linkinfo[IFLA_INFO_MAX+1];
2774 unsigned char name_assign_type = NET_NAME_USER;
2775 int err;
2776
2777 #ifdef CONFIG_MODULES
2778 replay:
2779 #endif
2780 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
2781 if (err < 0)
2782 return err;
2783
2784 err = rtnl_ensure_unique_netns(tb, extack, false);
2785 if (err < 0)
2786 return err;
2787
2788 if (tb[IFLA_IFNAME])
2789 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2790 else
2791 ifname[0] = '\0';
2792
2793 ifm = nlmsg_data(nlh);
2794 if (ifm->ifi_index > 0)
2795 dev = __dev_get_by_index(net, ifm->ifi_index);
2796 else {
2797 if (ifname[0])
2798 dev = __dev_get_by_name(net, ifname);
2799 else
2800 dev = NULL;
2801 }
2802
2803 if (dev) {
2804 master_dev = netdev_master_upper_dev_get(dev);
2805 if (master_dev)
2806 m_ops = master_dev->rtnl_link_ops;
2807 }
2808
2809 err = validate_linkmsg(dev, tb);
2810 if (err < 0)
2811 return err;
2812
2813 if (tb[IFLA_LINKINFO]) {
2814 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
2815 tb[IFLA_LINKINFO], ifla_info_policy,
2816 NULL);
2817 if (err < 0)
2818 return err;
2819 } else
2820 memset(linkinfo, 0, sizeof(linkinfo));
2821
2822 if (linkinfo[IFLA_INFO_KIND]) {
2823 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
2824 ops = rtnl_link_ops_get(kind);
2825 } else {
2826 kind[0] = '\0';
2827 ops = NULL;
2828 }
2829
2830 if (1) {
2831 struct nlattr *attr[ops ? ops->maxtype + 1 : 1];
2832 struct nlattr *slave_attr[m_ops ? m_ops->slave_maxtype + 1 : 1];
2833 struct nlattr **data = NULL;
2834 struct nlattr **slave_data = NULL;
2835 struct net *dest_net, *link_net = NULL;
2836
2837 if (ops) {
2838 if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
2839 err = nla_parse_nested(attr, ops->maxtype,
2840 linkinfo[IFLA_INFO_DATA],
2841 ops->policy, NULL);
2842 if (err < 0)
2843 return err;
2844 data = attr;
2845 }
2846 if (ops->validate) {
2847 err = ops->validate(tb, data, extack);
2848 if (err < 0)
2849 return err;
2850 }
2851 }
2852
2853 if (m_ops) {
2854 if (m_ops->slave_maxtype &&
2855 linkinfo[IFLA_INFO_SLAVE_DATA]) {
2856 err = nla_parse_nested(slave_attr,
2857 m_ops->slave_maxtype,
2858 linkinfo[IFLA_INFO_SLAVE_DATA],
2859 m_ops->slave_policy,
2860 NULL);
2861 if (err < 0)
2862 return err;
2863 slave_data = slave_attr;
2864 }
2865 }
2866
2867 if (dev) {
2868 int status = 0;
2869
2870 if (nlh->nlmsg_flags & NLM_F_EXCL)
2871 return -EEXIST;
2872 if (nlh->nlmsg_flags & NLM_F_REPLACE)
2873 return -EOPNOTSUPP;
2874
2875 if (linkinfo[IFLA_INFO_DATA]) {
2876 if (!ops || ops != dev->rtnl_link_ops ||
2877 !ops->changelink)
2878 return -EOPNOTSUPP;
2879
2880 err = ops->changelink(dev, tb, data, extack);
2881 if (err < 0)
2882 return err;
2883 status |= DO_SETLINK_NOTIFY;
2884 }
2885
2886 if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
2887 if (!m_ops || !m_ops->slave_changelink)
2888 return -EOPNOTSUPP;
2889
2890 err = m_ops->slave_changelink(master_dev, dev,
2891 tb, slave_data,
2892 extack);
2893 if (err < 0)
2894 return err;
2895 status |= DO_SETLINK_NOTIFY;
2896 }
2897
2898 return do_setlink(skb, dev, ifm, extack, tb, ifname,
2899 status);
2900 }
2901
2902 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
2903 if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
2904 return rtnl_group_changelink(skb, net,
2905 nla_get_u32(tb[IFLA_GROUP]),
2906 ifm, extack, tb);
2907 return -ENODEV;
2908 }
2909
2910 if (tb[IFLA_MAP] || tb[IFLA_PROTINFO])
2911 return -EOPNOTSUPP;
2912
2913 if (!ops) {
2914 #ifdef CONFIG_MODULES
2915 if (kind[0]) {
2916 __rtnl_unlock();
2917 request_module("rtnl-link-%s", kind);
2918 rtnl_lock();
2919 ops = rtnl_link_ops_get(kind);
2920 if (ops)
2921 goto replay;
2922 }
2923 #endif
2924 return -EOPNOTSUPP;
2925 }
2926
2927 if (!ops->setup)
2928 return -EOPNOTSUPP;
2929
2930 if (!ifname[0]) {
2931 snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
2932 name_assign_type = NET_NAME_ENUM;
2933 }
2934
2935 dest_net = rtnl_link_get_net_capable(skb, net, tb, CAP_NET_ADMIN);
2936 if (IS_ERR(dest_net))
2937 return PTR_ERR(dest_net);
2938
2939 if (tb[IFLA_LINK_NETNSID]) {
2940 int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
2941
2942 link_net = get_net_ns_by_id(dest_net, id);
2943 if (!link_net) {
2944 err = -EINVAL;
2945 goto out;
2946 }
2947 err = -EPERM;
2948 if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
2949 goto out;
2950 }
2951
2952 dev = rtnl_create_link(link_net ? : dest_net, ifname,
2953 name_assign_type, ops, tb);
2954 if (IS_ERR(dev)) {
2955 err = PTR_ERR(dev);
2956 goto out;
2957 }
2958
2959 dev->ifindex = ifm->ifi_index;
2960
2961 if (ops->newlink) {
2962 err = ops->newlink(link_net ? : net, dev, tb, data,
2963 extack);
2964 /* Drivers should call free_netdev() in ->destructor
2965 * and unregister it on failure after registration
2966 * so that device could be finally freed in rtnl_unlock.
2967 */
2968 if (err < 0) {
2969 /* If device is not registered at all, free it now */
2970 if (dev->reg_state == NETREG_UNINITIALIZED)
2971 free_netdev(dev);
2972 goto out;
2973 }
2974 } else {
2975 err = register_netdevice(dev);
2976 if (err < 0) {
2977 free_netdev(dev);
2978 goto out;
2979 }
2980 }
2981 err = rtnl_configure_link(dev, ifm);
2982 if (err < 0)
2983 goto out_unregister;
2984 if (link_net) {
2985 err = dev_change_net_namespace(dev, dest_net, ifname);
2986 if (err < 0)
2987 goto out_unregister;
2988 }
2989 if (tb[IFLA_MASTER]) {
2990 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]),
2991 extack);
2992 if (err)
2993 goto out_unregister;
2994 }
2995 out:
2996 if (link_net)
2997 put_net(link_net);
2998 put_net(dest_net);
2999 return err;
3000 out_unregister:
3001 if (ops->newlink) {
3002 LIST_HEAD(list_kill);
3003
3004 ops->dellink(dev, &list_kill);
3005 unregister_netdevice_many(&list_kill);
3006 } else {
3007 unregister_netdevice(dev);
3008 }
3009 goto out;
3010 }
3011 }
3012
3013 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3014 struct netlink_ext_ack *extack)
3015 {
3016 struct net *net = sock_net(skb->sk);
3017 struct net *tgt_net = net;
3018 struct ifinfomsg *ifm;
3019 char ifname[IFNAMSIZ];
3020 struct nlattr *tb[IFLA_MAX+1];
3021 struct net_device *dev = NULL;
3022 struct sk_buff *nskb;
3023 int netnsid = -1;
3024 int err;
3025 u32 ext_filter_mask = 0;
3026
3027 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
3028 if (err < 0)
3029 return err;
3030
3031 err = rtnl_ensure_unique_netns(tb, extack, true);
3032 if (err < 0)
3033 return err;
3034
3035 if (tb[IFLA_IF_NETNSID]) {
3036 netnsid = nla_get_s32(tb[IFLA_IF_NETNSID]);
3037 tgt_net = get_target_net(NETLINK_CB(skb).sk, netnsid);
3038 if (IS_ERR(tgt_net))
3039 return PTR_ERR(tgt_net);
3040 }
3041
3042 if (tb[IFLA_IFNAME])
3043 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
3044
3045 if (tb[IFLA_EXT_MASK])
3046 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
3047
3048 err = -EINVAL;
3049 ifm = nlmsg_data(nlh);
3050 if (ifm->ifi_index > 0)
3051 dev = __dev_get_by_index(tgt_net, ifm->ifi_index);
3052 else if (tb[IFLA_IFNAME])
3053 dev = __dev_get_by_name(tgt_net, ifname);
3054 else
3055 goto out;
3056
3057 err = -ENODEV;
3058 if (dev == NULL)
3059 goto out;
3060
3061 err = -ENOBUFS;
3062 nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
3063 if (nskb == NULL)
3064 goto out;
3065
3066 err = rtnl_fill_ifinfo(nskb, dev, net,
3067 RTM_NEWLINK, NETLINK_CB(skb).portid,
3068 nlh->nlmsg_seq, 0, 0, ext_filter_mask,
3069 0, NULL, 0, netnsid, GFP_KERNEL);
3070 if (err < 0) {
3071 /* -EMSGSIZE implies BUG in if_nlmsg_size */
3072 WARN_ON(err == -EMSGSIZE);
3073 kfree_skb(nskb);
3074 } else
3075 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
3076 out:
3077 if (netnsid >= 0)
3078 put_net(tgt_net);
3079
3080 return err;
3081 }
3082
3083 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
3084 {
3085 struct net *net = sock_net(skb->sk);
3086 struct net_device *dev;
3087 struct nlattr *tb[IFLA_MAX+1];
3088 u32 ext_filter_mask = 0;
3089 u16 min_ifinfo_dump_size = 0;
3090 int hdrlen;
3091
3092 /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
3093 hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
3094 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
3095
3096 if (nlmsg_parse(nlh, hdrlen, tb, IFLA_MAX, ifla_policy, NULL) >= 0) {
3097 if (tb[IFLA_EXT_MASK])
3098 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
3099 }
3100
3101 if (!ext_filter_mask)
3102 return NLMSG_GOODSIZE;
3103 /*
3104 * traverse the list of net devices and compute the minimum
3105 * buffer size based upon the filter mask.
3106 */
3107 rcu_read_lock();
3108 for_each_netdev_rcu(net, dev) {
3109 min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
3110 if_nlmsg_size(dev,
3111 ext_filter_mask));
3112 }
3113 rcu_read_unlock();
3114
3115 return nlmsg_total_size(min_ifinfo_dump_size);
3116 }
3117
3118 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
3119 {
3120 int idx;
3121 int s_idx = cb->family;
3122 int type = cb->nlh->nlmsg_type - RTM_BASE;
3123
3124 if (s_idx == 0)
3125 s_idx = 1;
3126
3127 for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
3128 struct rtnl_link *handlers;
3129 rtnl_dumpit_func dumpit;
3130
3131 if (idx < s_idx || idx == PF_PACKET)
3132 continue;
3133
3134 handlers = rtnl_dereference(rtnl_msg_handlers[idx]);
3135 if (!handlers)
3136 continue;
3137
3138 dumpit = READ_ONCE(handlers[type].dumpit);
3139 if (!dumpit)
3140 continue;
3141
3142 if (idx > s_idx) {
3143 memset(&cb->args[0], 0, sizeof(cb->args));
3144 cb->prev_seq = 0;
3145 cb->seq = 0;
3146 }
3147 if (dumpit(skb, cb))
3148 break;
3149 }
3150 cb->family = idx;
3151
3152 return skb->len;
3153 }
3154
3155 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
3156 unsigned int change,
3157 u32 event, gfp_t flags, int *new_nsid,
3158 int new_ifindex)
3159 {
3160 struct net *net = dev_net(dev);
3161 struct sk_buff *skb;
3162 int err = -ENOBUFS;
3163 size_t if_info_size;
3164
3165 skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
3166 if (skb == NULL)
3167 goto errout;
3168
3169 err = rtnl_fill_ifinfo(skb, dev, dev_net(dev),
3170 type, 0, 0, change, 0, 0, event,
3171 new_nsid, new_ifindex, -1, flags);
3172 if (err < 0) {
3173 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
3174 WARN_ON(err == -EMSGSIZE);
3175 kfree_skb(skb);
3176 goto errout;
3177 }
3178 return skb;
3179 errout:
3180 if (err < 0)
3181 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
3182 return NULL;
3183 }
3184
3185 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags)
3186 {
3187 struct net *net = dev_net(dev);
3188
3189 rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
3190 }
3191
3192 static void rtmsg_ifinfo_event(int type, struct net_device *dev,
3193 unsigned int change, u32 event,
3194 gfp_t flags, int *new_nsid, int new_ifindex)
3195 {
3196 struct sk_buff *skb;
3197
3198 if (dev->reg_state != NETREG_REGISTERED)
3199 return;
3200
3201 skb = rtmsg_ifinfo_build_skb(type, dev, change, event, flags, new_nsid,
3202 new_ifindex);
3203 if (skb)
3204 rtmsg_ifinfo_send(skb, dev, flags);
3205 }
3206
3207 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
3208 gfp_t flags)
3209 {
3210 rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
3211 NULL, 0);
3212 }
3213
3214 void rtmsg_ifinfo_newnet(int type, struct net_device *dev, unsigned int change,
3215 gfp_t flags, int *new_nsid, int new_ifindex)
3216 {
3217 rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
3218 new_nsid, new_ifindex);
3219 }
3220
3221 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
3222 struct net_device *dev,
3223 u8 *addr, u16 vid, u32 pid, u32 seq,
3224 int type, unsigned int flags,
3225 int nlflags, u16 ndm_state)
3226 {
3227 struct nlmsghdr *nlh;
3228 struct ndmsg *ndm;
3229
3230 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
3231 if (!nlh)
3232 return -EMSGSIZE;
3233
3234 ndm = nlmsg_data(nlh);
3235 ndm->ndm_family = AF_BRIDGE;
3236 ndm->ndm_pad1 = 0;
3237 ndm->ndm_pad2 = 0;
3238 ndm->ndm_flags = flags;
3239 ndm->ndm_type = 0;
3240 ndm->ndm_ifindex = dev->ifindex;
3241 ndm->ndm_state = ndm_state;
3242
3243 if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
3244 goto nla_put_failure;
3245 if (vid)
3246 if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
3247 goto nla_put_failure;
3248
3249 nlmsg_end(skb, nlh);
3250 return 0;
3251
3252 nla_put_failure:
3253 nlmsg_cancel(skb, nlh);
3254 return -EMSGSIZE;
3255 }
3256
3257 static inline size_t rtnl_fdb_nlmsg_size(void)
3258 {
3259 return NLMSG_ALIGN(sizeof(struct ndmsg)) +
3260 nla_total_size(ETH_ALEN) + /* NDA_LLADDR */
3261 nla_total_size(sizeof(u16)) + /* NDA_VLAN */
3262 0;
3263 }
3264
3265 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type,
3266 u16 ndm_state)
3267 {
3268 struct net *net = dev_net(dev);
3269 struct sk_buff *skb;
3270 int err = -ENOBUFS;
3271
3272 skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
3273 if (!skb)
3274 goto errout;
3275
3276 err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
3277 0, 0, type, NTF_SELF, 0, ndm_state);
3278 if (err < 0) {
3279 kfree_skb(skb);
3280 goto errout;
3281 }
3282
3283 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
3284 return;
3285 errout:
3286 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
3287 }
3288
3289 /**
3290 * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
3291 */
3292 int ndo_dflt_fdb_add(struct ndmsg *ndm,
3293 struct nlattr *tb[],
3294 struct net_device *dev,
3295 const unsigned char *addr, u16 vid,
3296 u16 flags)
3297 {
3298 int err = -EINVAL;
3299
3300 /* If aging addresses are supported device will need to
3301 * implement its own handler for this.
3302 */
3303 if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
3304 pr_info("%s: FDB only supports static addresses\n", dev->name);
3305 return err;
3306 }
3307
3308 if (vid) {
3309 pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
3310 return err;
3311 }
3312
3313 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
3314 err = dev_uc_add_excl(dev, addr);
3315 else if (is_multicast_ether_addr(addr))
3316 err = dev_mc_add_excl(dev, addr);
3317
3318 /* Only return duplicate errors if NLM_F_EXCL is set */
3319 if (err == -EEXIST && !(flags & NLM_F_EXCL))
3320 err = 0;
3321
3322 return err;
3323 }
3324 EXPORT_SYMBOL(ndo_dflt_fdb_add);
3325
3326 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid,
3327 struct netlink_ext_ack *extack)
3328 {
3329 u16 vid = 0;
3330
3331 if (vlan_attr) {
3332 if (nla_len(vlan_attr) != sizeof(u16)) {
3333 NL_SET_ERR_MSG(extack, "invalid vlan attribute size");
3334 return -EINVAL;
3335 }
3336
3337 vid = nla_get_u16(vlan_attr);
3338
3339 if (!vid || vid >= VLAN_VID_MASK) {
3340 NL_SET_ERR_MSG(extack, "invalid vlan id");
3341 return -EINVAL;
3342 }
3343 }
3344 *p_vid = vid;
3345 return 0;
3346 }
3347
3348 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh,
3349 struct netlink_ext_ack *extack)
3350 {
3351 struct net *net = sock_net(skb->sk);
3352 struct ndmsg *ndm;
3353 struct nlattr *tb[NDA_MAX+1];
3354 struct net_device *dev;
3355 u8 *addr;
3356 u16 vid;
3357 int err;
3358
3359 err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL, extack);
3360 if (err < 0)
3361 return err;
3362
3363 ndm = nlmsg_data(nlh);
3364 if (ndm->ndm_ifindex == 0) {
3365 NL_SET_ERR_MSG(extack, "invalid ifindex");
3366 return -EINVAL;
3367 }
3368
3369 dev = __dev_get_by_index(net, ndm->ndm_ifindex);
3370 if (dev == NULL) {
3371 NL_SET_ERR_MSG(extack, "unknown ifindex");
3372 return -ENODEV;
3373 }
3374
3375 if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
3376 NL_SET_ERR_MSG(extack, "invalid address");
3377 return -EINVAL;
3378 }
3379
3380 if (dev->type != ARPHRD_ETHER) {
3381 NL_SET_ERR_MSG(extack, "FDB add only supported for Ethernet devices");
3382 return -EINVAL;
3383 }
3384
3385 addr = nla_data(tb[NDA_LLADDR]);
3386
3387 err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
3388 if (err)
3389 return err;
3390
3391 err = -EOPNOTSUPP;
3392
3393 /* Support fdb on master device the net/bridge default case */
3394 if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
3395 (dev->priv_flags & IFF_BRIDGE_PORT)) {
3396 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3397 const struct net_device_ops *ops = br_dev->netdev_ops;
3398
3399 err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
3400 nlh->nlmsg_flags);
3401 if (err)
3402 goto out;
3403 else
3404 ndm->ndm_flags &= ~NTF_MASTER;
3405 }
3406
3407 /* Embedded bridge, macvlan, and any other device support */
3408 if ((ndm->ndm_flags & NTF_SELF)) {
3409 if (dev->netdev_ops->ndo_fdb_add)
3410 err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
3411 vid,
3412 nlh->nlmsg_flags);
3413 else
3414 err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
3415 nlh->nlmsg_flags);
3416
3417 if (!err) {
3418 rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH,
3419 ndm->ndm_state);
3420 ndm->ndm_flags &= ~NTF_SELF;
3421 }
3422 }
3423 out:
3424 return err;
3425 }
3426
3427 /**
3428 * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
3429 */
3430 int ndo_dflt_fdb_del(struct ndmsg *ndm,
3431 struct nlattr *tb[],
3432 struct net_device *dev,
3433 const unsigned char *addr, u16 vid)
3434 {
3435 int err = -EINVAL;
3436
3437 /* If aging addresses are supported device will need to
3438 * implement its own handler for this.
3439 */
3440 if (!(ndm->ndm_state & NUD_PERMANENT)) {
3441 pr_info("%s: FDB only supports static addresses\n", dev->name);
3442 return err;
3443 }
3444
3445 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
3446 err = dev_uc_del(dev, addr);
3447 else if (is_multicast_ether_addr(addr))
3448 err = dev_mc_del(dev, addr);
3449
3450 return err;
3451 }
3452 EXPORT_SYMBOL(ndo_dflt_fdb_del);
3453
3454 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh,
3455 struct netlink_ext_ack *extack)
3456 {
3457 struct net *net = sock_net(skb->sk);
3458 struct ndmsg *ndm;
3459 struct nlattr *tb[NDA_MAX+1];
3460 struct net_device *dev;
3461 int err = -EINVAL;
3462 __u8 *addr;
3463 u16 vid;
3464
3465 if (!netlink_capable(skb, CAP_NET_ADMIN))
3466 return -EPERM;
3467
3468 err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL, extack);
3469 if (err < 0)
3470 return err;
3471
3472 ndm = nlmsg_data(nlh);
3473 if (ndm->ndm_ifindex == 0) {
3474 NL_SET_ERR_MSG(extack, "invalid ifindex");
3475 return -EINVAL;
3476 }
3477
3478 dev = __dev_get_by_index(net, ndm->ndm_ifindex);
3479 if (dev == NULL) {
3480 NL_SET_ERR_MSG(extack, "unknown ifindex");
3481 return -ENODEV;
3482 }
3483
3484 if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
3485 NL_SET_ERR_MSG(extack, "invalid address");
3486 return -EINVAL;
3487 }
3488
3489 if (dev->type != ARPHRD_ETHER) {
3490 NL_SET_ERR_MSG(extack, "FDB delete only supported for Ethernet devices");
3491 return -EINVAL;
3492 }
3493
3494 addr = nla_data(tb[NDA_LLADDR]);
3495
3496 err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
3497 if (err)
3498 return err;
3499
3500 err = -EOPNOTSUPP;
3501
3502 /* Support fdb on master device the net/bridge default case */
3503 if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
3504 (dev->priv_flags & IFF_BRIDGE_PORT)) {
3505 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3506 const struct net_device_ops *ops = br_dev->netdev_ops;
3507
3508 if (ops->ndo_fdb_del)
3509 err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid);
3510
3511 if (err)
3512 goto out;
3513 else
3514 ndm->ndm_flags &= ~NTF_MASTER;
3515 }
3516
3517 /* Embedded bridge, macvlan, and any other device support */
3518 if (ndm->ndm_flags & NTF_SELF) {
3519 if (dev->netdev_ops->ndo_fdb_del)
3520 err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr,
3521 vid);
3522 else
3523 err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
3524
3525 if (!err) {
3526 rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH,
3527 ndm->ndm_state);
3528 ndm->ndm_flags &= ~NTF_SELF;
3529 }
3530 }
3531 out:
3532 return err;
3533 }
3534
3535 static int nlmsg_populate_fdb(struct sk_buff *skb,
3536 struct netlink_callback *cb,
3537 struct net_device *dev,
3538 int *idx,
3539 struct netdev_hw_addr_list *list)
3540 {
3541 struct netdev_hw_addr *ha;
3542 int err;
3543 u32 portid, seq;
3544
3545 portid = NETLINK_CB(cb->skb).portid;
3546 seq = cb->nlh->nlmsg_seq;
3547
3548 list_for_each_entry(ha, &list->list, list) {
3549 if (*idx < cb->args[2])
3550 goto skip;
3551
3552 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
3553 portid, seq,
3554 RTM_NEWNEIGH, NTF_SELF,
3555 NLM_F_MULTI, NUD_PERMANENT);
3556 if (err < 0)
3557 return err;
3558 skip:
3559 *idx += 1;
3560 }
3561 return 0;
3562 }
3563
3564 /**
3565 * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
3566 * @nlh: netlink message header
3567 * @dev: netdevice
3568 *
3569 * Default netdevice operation to dump the existing unicast address list.
3570 * Returns number of addresses from list put in skb.
3571 */
3572 int ndo_dflt_fdb_dump(struct sk_buff *skb,
3573 struct netlink_callback *cb,
3574 struct net_device *dev,
3575 struct net_device *filter_dev,
3576 int *idx)
3577 {
3578 int err;
3579
3580 if (dev->type != ARPHRD_ETHER)
3581 return -EINVAL;
3582
3583 netif_addr_lock_bh(dev);
3584 err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->uc);
3585 if (err)
3586 goto out;
3587 err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->mc);
3588 out:
3589 netif_addr_unlock_bh(dev);
3590 return err;
3591 }
3592 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
3593
3594 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
3595 {
3596 struct net_device *dev;
3597 struct nlattr *tb[IFLA_MAX+1];
3598 struct net_device *br_dev = NULL;
3599 const struct net_device_ops *ops = NULL;
3600 const struct net_device_ops *cops = NULL;
3601 struct ifinfomsg *ifm = nlmsg_data(cb->nlh);
3602 struct net *net = sock_net(skb->sk);
3603 struct hlist_head *head;
3604 int brport_idx = 0;
3605 int br_idx = 0;
3606 int h, s_h;
3607 int idx = 0, s_idx;
3608 int err = 0;
3609 int fidx = 0;
3610
3611 /* A hack to preserve kernel<->userspace interface.
3612 * Before Linux v4.12 this code accepted ndmsg since iproute2 v3.3.0.
3613 * However, ndmsg is shorter than ifinfomsg thus nlmsg_parse() bails.
3614 * So, check for ndmsg with an optional u32 attribute (not used here).
3615 * Fortunately these sizes don't conflict with the size of ifinfomsg
3616 * with an optional attribute.
3617 */
3618 if (nlmsg_len(cb->nlh) != sizeof(struct ndmsg) &&
3619 (nlmsg_len(cb->nlh) != sizeof(struct ndmsg) +
3620 nla_attr_size(sizeof(u32)))) {
3621 err = nlmsg_parse(cb->nlh, sizeof(struct ifinfomsg), tb,
3622 IFLA_MAX, ifla_policy, NULL);
3623 if (err < 0) {
3624 return -EINVAL;
3625 } else if (err == 0) {
3626 if (tb[IFLA_MASTER])
3627 br_idx = nla_get_u32(tb[IFLA_MASTER]);
3628 }
3629
3630 brport_idx = ifm->ifi_index;
3631 }
3632
3633 if (br_idx) {
3634 br_dev = __dev_get_by_index(net, br_idx);
3635 if (!br_dev)
3636 return -ENODEV;
3637
3638 ops = br_dev->netdev_ops;
3639 }
3640
3641 s_h = cb->args[0];
3642 s_idx = cb->args[1];
3643
3644 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3645 idx = 0;
3646 head = &net->dev_index_head[h];
3647 hlist_for_each_entry(dev, head, index_hlist) {
3648
3649 if (brport_idx && (dev->ifindex != brport_idx))
3650 continue;
3651
3652 if (!br_idx) { /* user did not specify a specific bridge */
3653 if (dev->priv_flags & IFF_BRIDGE_PORT) {
3654 br_dev = netdev_master_upper_dev_get(dev);
3655 cops = br_dev->netdev_ops;
3656 }
3657 } else {
3658 if (dev != br_dev &&
3659 !(dev->priv_flags & IFF_BRIDGE_PORT))
3660 continue;
3661
3662 if (br_dev != netdev_master_upper_dev_get(dev) &&
3663 !(dev->priv_flags & IFF_EBRIDGE))
3664 continue;
3665 cops = ops;
3666 }
3667
3668 if (idx < s_idx)
3669 goto cont;
3670
3671 if (dev->priv_flags & IFF_BRIDGE_PORT) {
3672 if (cops && cops->ndo_fdb_dump) {
3673 err = cops->ndo_fdb_dump(skb, cb,
3674 br_dev, dev,
3675 &fidx);
3676 if (err == -EMSGSIZE)
3677 goto out;
3678 }
3679 }
3680
3681 if (dev->netdev_ops->ndo_fdb_dump)
3682 err = dev->netdev_ops->ndo_fdb_dump(skb, cb,
3683 dev, NULL,
3684 &fidx);
3685 else
3686 err = ndo_dflt_fdb_dump(skb, cb, dev, NULL,
3687 &fidx);
3688 if (err == -EMSGSIZE)
3689 goto out;
3690
3691 cops = NULL;
3692
3693 /* reset fdb offset to 0 for rest of the interfaces */
3694 cb->args[2] = 0;
3695 fidx = 0;
3696 cont:
3697 idx++;
3698 }
3699 }
3700
3701 out:
3702 cb->args[0] = h;
3703 cb->args[1] = idx;
3704 cb->args[2] = fidx;
3705
3706 return skb->len;
3707 }
3708
3709 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
3710 unsigned int attrnum, unsigned int flag)
3711 {
3712 if (mask & flag)
3713 return nla_put_u8(skb, attrnum, !!(flags & flag));
3714 return 0;
3715 }
3716
3717 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
3718 struct net_device *dev, u16 mode,
3719 u32 flags, u32 mask, int nlflags,
3720 u32 filter_mask,
3721 int (*vlan_fill)(struct sk_buff *skb,
3722 struct net_device *dev,
3723 u32 filter_mask))
3724 {
3725 struct nlmsghdr *nlh;
3726 struct ifinfomsg *ifm;
3727 struct nlattr *br_afspec;
3728 struct nlattr *protinfo;
3729 u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
3730 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3731 int err = 0;
3732
3733 nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
3734 if (nlh == NULL)
3735 return -EMSGSIZE;
3736
3737 ifm = nlmsg_data(nlh);
3738 ifm->ifi_family = AF_BRIDGE;
3739 ifm->__ifi_pad = 0;
3740 ifm->ifi_type = dev->type;
3741 ifm->ifi_index = dev->ifindex;
3742 ifm->ifi_flags = dev_get_flags(dev);
3743 ifm->ifi_change = 0;
3744
3745
3746 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
3747 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
3748 nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
3749 (br_dev &&
3750 nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
3751 (dev->addr_len &&
3752 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
3753 (dev->ifindex != dev_get_iflink(dev) &&
3754 nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
3755 goto nla_put_failure;
3756
3757 br_afspec = nla_nest_start(skb, IFLA_AF_SPEC);
3758 if (!br_afspec)
3759 goto nla_put_failure;
3760
3761 if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
3762 nla_nest_cancel(skb, br_afspec);
3763 goto nla_put_failure;
3764 }
3765
3766 if (mode != BRIDGE_MODE_UNDEF) {
3767 if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
3768 nla_nest_cancel(skb, br_afspec);
3769 goto nla_put_failure;
3770 }
3771 }
3772 if (vlan_fill) {
3773 err = vlan_fill(skb, dev, filter_mask);
3774 if (err) {
3775 nla_nest_cancel(skb, br_afspec);
3776 goto nla_put_failure;
3777 }
3778 }
3779 nla_nest_end(skb, br_afspec);
3780
3781 protinfo = nla_nest_start(skb, IFLA_PROTINFO | NLA_F_NESTED);
3782 if (!protinfo)
3783 goto nla_put_failure;
3784
3785 if (brport_nla_put_flag(skb, flags, mask,
3786 IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
3787 brport_nla_put_flag(skb, flags, mask,
3788 IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
3789 brport_nla_put_flag(skb, flags, mask,
3790 IFLA_BRPORT_FAST_LEAVE,
3791 BR_MULTICAST_FAST_LEAVE) ||
3792 brport_nla_put_flag(skb, flags, mask,
3793 IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
3794 brport_nla_put_flag(skb, flags, mask,
3795 IFLA_BRPORT_LEARNING, BR_LEARNING) ||
3796 brport_nla_put_flag(skb, flags, mask,
3797 IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
3798 brport_nla_put_flag(skb, flags, mask,
3799 IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
3800 brport_nla_put_flag(skb, flags, mask,
3801 IFLA_BRPORT_PROXYARP, BR_PROXYARP)) {
3802 nla_nest_cancel(skb, protinfo);
3803 goto nla_put_failure;
3804 }
3805
3806 nla_nest_end(skb, protinfo);
3807
3808 nlmsg_end(skb, nlh);
3809 return 0;
3810 nla_put_failure:
3811 nlmsg_cancel(skb, nlh);
3812 return err ? err : -EMSGSIZE;
3813 }
3814 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink);
3815
3816 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
3817 {
3818 struct net *net = sock_net(skb->sk);
3819 struct net_device *dev;
3820 int idx = 0;
3821 u32 portid = NETLINK_CB(cb->skb).portid;
3822 u32 seq = cb->nlh->nlmsg_seq;
3823 u32 filter_mask = 0;
3824 int err;
3825
3826 if (nlmsg_len(cb->nlh) > sizeof(struct ifinfomsg)) {
3827 struct nlattr *extfilt;
3828
3829 extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct ifinfomsg),
3830 IFLA_EXT_MASK);
3831 if (extfilt) {
3832 if (nla_len(extfilt) < sizeof(filter_mask))
3833 return -EINVAL;
3834
3835 filter_mask = nla_get_u32(extfilt);
3836 }
3837 }
3838
3839 rcu_read_lock();
3840 for_each_netdev_rcu(net, dev) {
3841 const struct net_device_ops *ops = dev->netdev_ops;
3842 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3843
3844 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
3845 if (idx >= cb->args[0]) {
3846 err = br_dev->netdev_ops->ndo_bridge_getlink(
3847 skb, portid, seq, dev,
3848 filter_mask, NLM_F_MULTI);
3849 if (err < 0 && err != -EOPNOTSUPP) {
3850 if (likely(skb->len))
3851 break;
3852
3853 goto out_err;
3854 }
3855 }
3856 idx++;
3857 }
3858
3859 if (ops->ndo_bridge_getlink) {
3860 if (idx >= cb->args[0]) {
3861 err = ops->ndo_bridge_getlink(skb, portid,
3862 seq, dev,
3863 filter_mask,
3864 NLM_F_MULTI);
3865 if (err < 0 && err != -EOPNOTSUPP) {
3866 if (likely(skb->len))
3867 break;
3868
3869 goto out_err;
3870 }
3871 }
3872 idx++;
3873 }
3874 }
3875 err = skb->len;
3876 out_err:
3877 rcu_read_unlock();
3878 cb->args[0] = idx;
3879
3880 return err;
3881 }
3882
3883 static inline size_t bridge_nlmsg_size(void)
3884 {
3885 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3886 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3887 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3888 + nla_total_size(sizeof(u32)) /* IFLA_MASTER */
3889 + nla_total_size(sizeof(u32)) /* IFLA_MTU */
3890 + nla_total_size(sizeof(u32)) /* IFLA_LINK */
3891 + nla_total_size(sizeof(u32)) /* IFLA_OPERSTATE */
3892 + nla_total_size(sizeof(u8)) /* IFLA_PROTINFO */
3893 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
3894 + nla_total_size(sizeof(u16)) /* IFLA_BRIDGE_FLAGS */
3895 + nla_total_size(sizeof(u16)); /* IFLA_BRIDGE_MODE */
3896 }
3897
3898 static int rtnl_bridge_notify(struct net_device *dev)
3899 {
3900 struct net *net = dev_net(dev);
3901 struct sk_buff *skb;
3902 int err = -EOPNOTSUPP;
3903
3904 if (!dev->netdev_ops->ndo_bridge_getlink)
3905 return 0;
3906
3907 skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
3908 if (!skb) {
3909 err = -ENOMEM;
3910 goto errout;
3911 }
3912
3913 err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
3914 if (err < 0)
3915 goto errout;
3916
3917 if (!skb->len)
3918 goto errout;
3919
3920 rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
3921 return 0;
3922 errout:
3923 WARN_ON(err == -EMSGSIZE);
3924 kfree_skb(skb);
3925 if (err)
3926 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
3927 return err;
3928 }
3929
3930 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3931 struct netlink_ext_ack *extack)
3932 {
3933 struct net *net = sock_net(skb->sk);
3934 struct ifinfomsg *ifm;
3935 struct net_device *dev;
3936 struct nlattr *br_spec, *attr = NULL;
3937 int rem, err = -EOPNOTSUPP;
3938 u16 flags = 0;
3939 bool have_flags = false;
3940
3941 if (nlmsg_len(nlh) < sizeof(*ifm))
3942 return -EINVAL;
3943
3944 ifm = nlmsg_data(nlh);
3945 if (ifm->ifi_family != AF_BRIDGE)
3946 return -EPFNOSUPPORT;
3947
3948 dev = __dev_get_by_index(net, ifm->ifi_index);
3949 if (!dev) {
3950 NL_SET_ERR_MSG(extack, "unknown ifindex");
3951 return -ENODEV;
3952 }
3953
3954 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3955 if (br_spec) {
3956 nla_for_each_nested(attr, br_spec, rem) {
3957 if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3958 if (nla_len(attr) < sizeof(flags))
3959 return -EINVAL;
3960
3961 have_flags = true;
3962 flags = nla_get_u16(attr);
3963 break;
3964 }
3965 }
3966 }
3967
3968 if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3969 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3970
3971 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
3972 err = -EOPNOTSUPP;
3973 goto out;
3974 }
3975
3976 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags);
3977 if (err)
3978 goto out;
3979
3980 flags &= ~BRIDGE_FLAGS_MASTER;
3981 }
3982
3983 if ((flags & BRIDGE_FLAGS_SELF)) {
3984 if (!dev->netdev_ops->ndo_bridge_setlink)
3985 err = -EOPNOTSUPP;
3986 else
3987 err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
3988 flags);
3989 if (!err) {
3990 flags &= ~BRIDGE_FLAGS_SELF;
3991
3992 /* Generate event to notify upper layer of bridge
3993 * change
3994 */
3995 err = rtnl_bridge_notify(dev);
3996 }
3997 }
3998
3999 if (have_flags)
4000 memcpy(nla_data(attr), &flags, sizeof(flags));
4001 out:
4002 return err;
4003 }
4004
4005 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
4006 struct netlink_ext_ack *extack)
4007 {
4008 struct net *net = sock_net(skb->sk);
4009 struct ifinfomsg *ifm;
4010 struct net_device *dev;
4011 struct nlattr *br_spec, *attr = NULL;
4012 int rem, err = -EOPNOTSUPP;
4013 u16 flags = 0;
4014 bool have_flags = false;
4015
4016 if (nlmsg_len(nlh) < sizeof(*ifm))
4017 return -EINVAL;
4018
4019 ifm = nlmsg_data(nlh);
4020 if (ifm->ifi_family != AF_BRIDGE)
4021 return -EPFNOSUPPORT;
4022
4023 dev = __dev_get_by_index(net, ifm->ifi_index);
4024 if (!dev) {
4025 NL_SET_ERR_MSG(extack, "unknown ifindex");
4026 return -ENODEV;
4027 }
4028
4029 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
4030 if (br_spec) {
4031 nla_for_each_nested(attr, br_spec, rem) {
4032 if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
4033 if (nla_len(attr) < sizeof(flags))
4034 return -EINVAL;
4035
4036 have_flags = true;
4037 flags = nla_get_u16(attr);
4038 break;
4039 }
4040 }
4041 }
4042
4043 if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
4044 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4045
4046 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
4047 err = -EOPNOTSUPP;
4048 goto out;
4049 }
4050
4051 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
4052 if (err)
4053 goto out;
4054
4055 flags &= ~BRIDGE_FLAGS_MASTER;
4056 }
4057
4058 if ((flags & BRIDGE_FLAGS_SELF)) {
4059 if (!dev->netdev_ops->ndo_bridge_dellink)
4060 err = -EOPNOTSUPP;
4061 else
4062 err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
4063 flags);
4064
4065 if (!err) {
4066 flags &= ~BRIDGE_FLAGS_SELF;
4067
4068 /* Generate event to notify upper layer of bridge
4069 * change
4070 */
4071 err = rtnl_bridge_notify(dev);
4072 }
4073 }
4074
4075 if (have_flags)
4076 memcpy(nla_data(attr), &flags, sizeof(flags));
4077 out:
4078 return err;
4079 }
4080
4081 static bool stats_attr_valid(unsigned int mask, int attrid, int idxattr)
4082 {
4083 return (mask & IFLA_STATS_FILTER_BIT(attrid)) &&
4084 (!idxattr || idxattr == attrid);
4085 }
4086
4087 #define IFLA_OFFLOAD_XSTATS_FIRST (IFLA_OFFLOAD_XSTATS_UNSPEC + 1)
4088 static int rtnl_get_offload_stats_attr_size(int attr_id)
4089 {
4090 switch (attr_id) {
4091 case IFLA_OFFLOAD_XSTATS_CPU_HIT:
4092 return sizeof(struct rtnl_link_stats64);
4093 }
4094
4095 return 0;
4096 }
4097
4098 static int rtnl_get_offload_stats(struct sk_buff *skb, struct net_device *dev,
4099 int *prividx)
4100 {
4101 struct nlattr *attr = NULL;
4102 int attr_id, size;
4103 void *attr_data;
4104 int err;
4105
4106 if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
4107 dev->netdev_ops->ndo_get_offload_stats))
4108 return -ENODATA;
4109
4110 for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
4111 attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
4112 if (attr_id < *prividx)
4113 continue;
4114
4115 size = rtnl_get_offload_stats_attr_size(attr_id);
4116 if (!size)
4117 continue;
4118
4119 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
4120 continue;
4121
4122 attr = nla_reserve_64bit(skb, attr_id, size,
4123 IFLA_OFFLOAD_XSTATS_UNSPEC);
4124 if (!attr)
4125 goto nla_put_failure;
4126
4127 attr_data = nla_data(attr);
4128 memset(attr_data, 0, size);
4129 err = dev->netdev_ops->ndo_get_offload_stats(attr_id, dev,
4130 attr_data);
4131 if (err)
4132 goto get_offload_stats_failure;
4133 }
4134
4135 if (!attr)
4136 return -ENODATA;
4137
4138 *prividx = 0;
4139 return 0;
4140
4141 nla_put_failure:
4142 err = -EMSGSIZE;
4143 get_offload_stats_failure:
4144 *prividx = attr_id;
4145 return err;
4146 }
4147
4148 static int rtnl_get_offload_stats_size(const struct net_device *dev)
4149 {
4150 int nla_size = 0;
4151 int attr_id;
4152 int size;
4153
4154 if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
4155 dev->netdev_ops->ndo_get_offload_stats))
4156 return 0;
4157
4158 for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
4159 attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
4160 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
4161 continue;
4162 size = rtnl_get_offload_stats_attr_size(attr_id);
4163 nla_size += nla_total_size_64bit(size);
4164 }
4165
4166 if (nla_size != 0)
4167 nla_size += nla_total_size(0);
4168
4169 return nla_size;
4170 }
4171
4172 static int rtnl_fill_statsinfo(struct sk_buff *skb, struct net_device *dev,
4173 int type, u32 pid, u32 seq, u32 change,
4174 unsigned int flags, unsigned int filter_mask,
4175 int *idxattr, int *prividx)
4176 {
4177 struct if_stats_msg *ifsm;
4178 struct nlmsghdr *nlh;
4179 struct nlattr *attr;
4180 int s_prividx = *prividx;
4181 int err;
4182
4183 ASSERT_RTNL();
4184
4185 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifsm), flags);
4186 if (!nlh)
4187 return -EMSGSIZE;
4188
4189 ifsm = nlmsg_data(nlh);
4190 ifsm->family = PF_UNSPEC;
4191 ifsm->pad1 = 0;
4192 ifsm->pad2 = 0;
4193 ifsm->ifindex = dev->ifindex;
4194 ifsm->filter_mask = filter_mask;
4195
4196 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, *idxattr)) {
4197 struct rtnl_link_stats64 *sp;
4198
4199 attr = nla_reserve_64bit(skb, IFLA_STATS_LINK_64,
4200 sizeof(struct rtnl_link_stats64),
4201 IFLA_STATS_UNSPEC);
4202 if (!attr)
4203 goto nla_put_failure;
4204
4205 sp = nla_data(attr);
4206 dev_get_stats(dev, sp);
4207 }
4208
4209 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, *idxattr)) {
4210 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
4211
4212 if (ops && ops->fill_linkxstats) {
4213 *idxattr = IFLA_STATS_LINK_XSTATS;
4214 attr = nla_nest_start(skb,
4215 IFLA_STATS_LINK_XSTATS);
4216 if (!attr)
4217 goto nla_put_failure;
4218
4219 err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
4220 nla_nest_end(skb, attr);
4221 if (err)
4222 goto nla_put_failure;
4223 *idxattr = 0;
4224 }
4225 }
4226
4227 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE,
4228 *idxattr)) {
4229 const struct rtnl_link_ops *ops = NULL;
4230 const struct net_device *master;
4231
4232 master = netdev_master_upper_dev_get(dev);
4233 if (master)
4234 ops = master->rtnl_link_ops;
4235 if (ops && ops->fill_linkxstats) {
4236 *idxattr = IFLA_STATS_LINK_XSTATS_SLAVE;
4237 attr = nla_nest_start(skb,
4238 IFLA_STATS_LINK_XSTATS_SLAVE);
4239 if (!attr)
4240 goto nla_put_failure;
4241
4242 err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
4243 nla_nest_end(skb, attr);
4244 if (err)
4245 goto nla_put_failure;
4246 *idxattr = 0;
4247 }
4248 }
4249
4250 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS,
4251 *idxattr)) {
4252 *idxattr = IFLA_STATS_LINK_OFFLOAD_XSTATS;
4253 attr = nla_nest_start(skb, IFLA_STATS_LINK_OFFLOAD_XSTATS);
4254 if (!attr)
4255 goto nla_put_failure;
4256
4257 err = rtnl_get_offload_stats(skb, dev, prividx);
4258 if (err == -ENODATA)
4259 nla_nest_cancel(skb, attr);
4260 else
4261 nla_nest_end(skb, attr);
4262
4263 if (err && err != -ENODATA)
4264 goto nla_put_failure;
4265 *idxattr = 0;
4266 }
4267
4268 if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, *idxattr)) {
4269 struct rtnl_af_ops *af_ops;
4270
4271 *idxattr = IFLA_STATS_AF_SPEC;
4272 attr = nla_nest_start(skb, IFLA_STATS_AF_SPEC);
4273 if (!attr)
4274 goto nla_put_failure;
4275
4276 rcu_read_lock();
4277 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
4278 if (af_ops->fill_stats_af) {
4279 struct nlattr *af;
4280 int err;
4281
4282 af = nla_nest_start(skb, af_ops->family);
4283 if (!af) {
4284 rcu_read_unlock();
4285 goto nla_put_failure;
4286 }
4287 err = af_ops->fill_stats_af(skb, dev);
4288
4289 if (err == -ENODATA) {
4290 nla_nest_cancel(skb, af);
4291 } else if (err < 0) {
4292 rcu_read_unlock();
4293 goto nla_put_failure;
4294 }
4295
4296 nla_nest_end(skb, af);
4297 }
4298 }
4299 rcu_read_unlock();
4300
4301 nla_nest_end(skb, attr);
4302
4303 *idxattr = 0;
4304 }
4305
4306 nlmsg_end(skb, nlh);
4307
4308 return 0;
4309
4310 nla_put_failure:
4311 /* not a multi message or no progress mean a real error */
4312 if (!(flags & NLM_F_MULTI) || s_prividx == *prividx)
4313 nlmsg_cancel(skb, nlh);
4314 else
4315 nlmsg_end(skb, nlh);
4316
4317 return -EMSGSIZE;
4318 }
4319
4320 static size_t if_nlmsg_stats_size(const struct net_device *dev,
4321 u32 filter_mask)
4322 {
4323 size_t size = 0;
4324
4325 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, 0))
4326 size += nla_total_size_64bit(sizeof(struct rtnl_link_stats64));
4327
4328 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, 0)) {
4329 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
4330 int attr = IFLA_STATS_LINK_XSTATS;
4331
4332 if (ops && ops->get_linkxstats_size) {
4333 size += nla_total_size(ops->get_linkxstats_size(dev,
4334 attr));
4335 /* for IFLA_STATS_LINK_XSTATS */
4336 size += nla_total_size(0);
4337 }
4338 }
4339
4340 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE, 0)) {
4341 struct net_device *_dev = (struct net_device *)dev;
4342 const struct rtnl_link_ops *ops = NULL;
4343 const struct net_device *master;
4344
4345 /* netdev_master_upper_dev_get can't take const */
4346 master = netdev_master_upper_dev_get(_dev);
4347 if (master)
4348 ops = master->rtnl_link_ops;
4349 if (ops && ops->get_linkxstats_size) {
4350 int attr = IFLA_STATS_LINK_XSTATS_SLAVE;
4351
4352 size += nla_total_size(ops->get_linkxstats_size(dev,
4353 attr));
4354 /* for IFLA_STATS_LINK_XSTATS_SLAVE */
4355 size += nla_total_size(0);
4356 }
4357 }
4358
4359 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS, 0))
4360 size += rtnl_get_offload_stats_size(dev);
4361
4362 if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, 0)) {
4363 struct rtnl_af_ops *af_ops;
4364
4365 /* for IFLA_STATS_AF_SPEC */
4366 size += nla_total_size(0);
4367
4368 rcu_read_lock();
4369 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
4370 if (af_ops->get_stats_af_size) {
4371 size += nla_total_size(
4372 af_ops->get_stats_af_size(dev));
4373
4374 /* for AF_* */
4375 size += nla_total_size(0);
4376 }
4377 }
4378 rcu_read_unlock();
4379 }
4380
4381 return size;
4382 }
4383
4384 static int rtnl_stats_get(struct sk_buff *skb, struct nlmsghdr *nlh,
4385 struct netlink_ext_ack *extack)
4386 {
4387 struct net *net = sock_net(skb->sk);
4388 struct net_device *dev = NULL;
4389 int idxattr = 0, prividx = 0;
4390 struct if_stats_msg *ifsm;
4391 struct sk_buff *nskb;
4392 u32 filter_mask;
4393 int err;
4394
4395 if (nlmsg_len(nlh) < sizeof(*ifsm))
4396 return -EINVAL;
4397
4398 ifsm = nlmsg_data(nlh);
4399 if (ifsm->ifindex > 0)
4400 dev = __dev_get_by_index(net, ifsm->ifindex);
4401 else
4402 return -EINVAL;
4403
4404 if (!dev)
4405 return -ENODEV;
4406
4407 filter_mask = ifsm->filter_mask;
4408 if (!filter_mask)
4409 return -EINVAL;
4410
4411 nskb = nlmsg_new(if_nlmsg_stats_size(dev, filter_mask), GFP_KERNEL);
4412 if (!nskb)
4413 return -ENOBUFS;
4414
4415 err = rtnl_fill_statsinfo(nskb, dev, RTM_NEWSTATS,
4416 NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
4417 0, filter_mask, &idxattr, &prividx);
4418 if (err < 0) {
4419 /* -EMSGSIZE implies BUG in if_nlmsg_stats_size */
4420 WARN_ON(err == -EMSGSIZE);
4421 kfree_skb(nskb);
4422 } else {
4423 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
4424 }
4425
4426 return err;
4427 }
4428
4429 static int rtnl_stats_dump(struct sk_buff *skb, struct netlink_callback *cb)
4430 {
4431 int h, s_h, err, s_idx, s_idxattr, s_prividx;
4432 struct net *net = sock_net(skb->sk);
4433 unsigned int flags = NLM_F_MULTI;
4434 struct if_stats_msg *ifsm;
4435 struct hlist_head *head;
4436 struct net_device *dev;
4437 u32 filter_mask = 0;
4438 int idx = 0;
4439
4440 s_h = cb->args[0];
4441 s_idx = cb->args[1];
4442 s_idxattr = cb->args[2];
4443 s_prividx = cb->args[3];
4444
4445 cb->seq = net->dev_base_seq;
4446
4447 if (nlmsg_len(cb->nlh) < sizeof(*ifsm))
4448 return -EINVAL;
4449
4450 ifsm = nlmsg_data(cb->nlh);
4451 filter_mask = ifsm->filter_mask;
4452 if (!filter_mask)
4453 return -EINVAL;
4454
4455 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4456 idx = 0;
4457 head = &net->dev_index_head[h];
4458 hlist_for_each_entry(dev, head, index_hlist) {
4459 if (idx < s_idx)
4460 goto cont;
4461 err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS,
4462 NETLINK_CB(cb->skb).portid,
4463 cb->nlh->nlmsg_seq, 0,
4464 flags, filter_mask,
4465 &s_idxattr, &s_prividx);
4466 /* If we ran out of room on the first message,
4467 * we're in trouble
4468 */
4469 WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
4470
4471 if (err < 0)
4472 goto out;
4473 s_prividx = 0;
4474 s_idxattr = 0;
4475 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4476 cont:
4477 idx++;
4478 }
4479 }
4480 out:
4481 cb->args[3] = s_prividx;
4482 cb->args[2] = s_idxattr;
4483 cb->args[1] = idx;
4484 cb->args[0] = h;
4485
4486 return skb->len;
4487 }
4488
4489 /* Process one rtnetlink message. */
4490
4491 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh,
4492 struct netlink_ext_ack *extack)
4493 {
4494 struct net *net = sock_net(skb->sk);
4495 struct rtnl_link *handlers;
4496 int err = -EOPNOTSUPP;
4497 rtnl_doit_func doit;
4498 unsigned int flags;
4499 int kind;
4500 int family;
4501 int type;
4502
4503 type = nlh->nlmsg_type;
4504 if (type > RTM_MAX)
4505 return -EOPNOTSUPP;
4506
4507 type -= RTM_BASE;
4508
4509 /* All the messages must have at least 1 byte length */
4510 if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
4511 return 0;
4512
4513 family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
4514 kind = type&3;
4515
4516 if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
4517 return -EPERM;
4518
4519 if (family >= ARRAY_SIZE(rtnl_msg_handlers))
4520 family = PF_UNSPEC;
4521
4522 rcu_read_lock();
4523 handlers = rcu_dereference(rtnl_msg_handlers[family]);
4524 if (!handlers) {
4525 family = PF_UNSPEC;
4526 handlers = rcu_dereference(rtnl_msg_handlers[family]);
4527 }
4528
4529 if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
4530 struct sock *rtnl;
4531 rtnl_dumpit_func dumpit;
4532 u16 min_dump_alloc = 0;
4533
4534 dumpit = READ_ONCE(handlers[type].dumpit);
4535 if (!dumpit) {
4536 family = PF_UNSPEC;
4537 handlers = rcu_dereference(rtnl_msg_handlers[PF_UNSPEC]);
4538 if (!handlers)
4539 goto err_unlock;
4540
4541 dumpit = READ_ONCE(handlers[type].dumpit);
4542 if (!dumpit)
4543 goto err_unlock;
4544 }
4545
4546 refcount_inc(&rtnl_msg_handlers_ref[family]);
4547
4548 if (type == RTM_GETLINK - RTM_BASE)
4549 min_dump_alloc = rtnl_calcit(skb, nlh);
4550
4551 rcu_read_unlock();
4552
4553 rtnl = net->rtnl;
4554 {
4555 struct netlink_dump_control c = {
4556 .dump = dumpit,
4557 .min_dump_alloc = min_dump_alloc,
4558 };
4559 err = netlink_dump_start(rtnl, skb, nlh, &c);
4560 }
4561 refcount_dec(&rtnl_msg_handlers_ref[family]);
4562 return err;
4563 }
4564
4565 doit = READ_ONCE(handlers[type].doit);
4566 if (!doit) {
4567 family = PF_UNSPEC;
4568 handlers = rcu_dereference(rtnl_msg_handlers[family]);
4569 }
4570
4571 flags = READ_ONCE(handlers[type].flags);
4572 if (flags & RTNL_FLAG_DOIT_UNLOCKED) {
4573 refcount_inc(&rtnl_msg_handlers_ref[family]);
4574 doit = READ_ONCE(handlers[type].doit);
4575 rcu_read_unlock();
4576 if (doit)
4577 err = doit(skb, nlh, extack);
4578 refcount_dec(&rtnl_msg_handlers_ref[family]);
4579 return err;
4580 }
4581
4582 rcu_read_unlock();
4583
4584 rtnl_lock();
4585 handlers = rtnl_dereference(rtnl_msg_handlers[family]);
4586 if (handlers) {
4587 doit = READ_ONCE(handlers[type].doit);
4588 if (doit)
4589 err = doit(skb, nlh, extack);
4590 }
4591 rtnl_unlock();
4592 return err;
4593
4594 err_unlock:
4595 rcu_read_unlock();
4596 return -EOPNOTSUPP;
4597 }
4598
4599 static void rtnetlink_rcv(struct sk_buff *skb)
4600 {
4601 netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
4602 }
4603
4604 static int rtnetlink_bind(struct net *net, int group)
4605 {
4606 switch (group) {
4607 case RTNLGRP_IPV4_MROUTE_R:
4608 case RTNLGRP_IPV6_MROUTE_R:
4609 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
4610 return -EPERM;
4611 break;
4612 }
4613 return 0;
4614 }
4615
4616 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
4617 {
4618 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
4619
4620 switch (event) {
4621 case NETDEV_REBOOT:
4622 case NETDEV_CHANGEMTU:
4623 case NETDEV_CHANGEADDR:
4624 case NETDEV_CHANGENAME:
4625 case NETDEV_FEAT_CHANGE:
4626 case NETDEV_BONDING_FAILOVER:
4627 case NETDEV_POST_TYPE_CHANGE:
4628 case NETDEV_NOTIFY_PEERS:
4629 case NETDEV_CHANGEUPPER:
4630 case NETDEV_RESEND_IGMP:
4631 case NETDEV_CHANGEINFODATA:
4632 case NETDEV_CHANGELOWERSTATE:
4633 case NETDEV_CHANGE_TX_QUEUE_LEN:
4634 rtmsg_ifinfo_event(RTM_NEWLINK, dev, 0, rtnl_get_event(event),
4635 GFP_KERNEL, NULL, 0);
4636 break;
4637 default:
4638 break;
4639 }
4640 return NOTIFY_DONE;
4641 }
4642
4643 static struct notifier_block rtnetlink_dev_notifier = {
4644 .notifier_call = rtnetlink_event,
4645 };
4646
4647
4648 static int __net_init rtnetlink_net_init(struct net *net)
4649 {
4650 struct sock *sk;
4651 struct netlink_kernel_cfg cfg = {
4652 .groups = RTNLGRP_MAX,
4653 .input = rtnetlink_rcv,
4654 .cb_mutex = &rtnl_mutex,
4655 .flags = NL_CFG_F_NONROOT_RECV,
4656 .bind = rtnetlink_bind,
4657 };
4658
4659 sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
4660 if (!sk)
4661 return -ENOMEM;
4662 net->rtnl = sk;
4663 return 0;
4664 }
4665
4666 static void __net_exit rtnetlink_net_exit(struct net *net)
4667 {
4668 netlink_kernel_release(net->rtnl);
4669 net->rtnl = NULL;
4670 }
4671
4672 static struct pernet_operations rtnetlink_net_ops = {
4673 .init = rtnetlink_net_init,
4674 .exit = rtnetlink_net_exit,
4675 };
4676
4677 void __init rtnetlink_init(void)
4678 {
4679 int i;
4680
4681 for (i = 0; i < ARRAY_SIZE(rtnl_msg_handlers_ref); i++)
4682 refcount_set(&rtnl_msg_handlers_ref[i], 1);
4683
4684 if (register_pernet_subsys(&rtnetlink_net_ops))
4685 panic("rtnetlink_init: cannot initialize rtnetlink\n");
4686
4687 register_netdevice_notifier(&rtnetlink_dev_notifier);
4688
4689 rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
4690 rtnl_dump_ifinfo, 0);
4691 rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, 0);
4692 rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, 0);
4693 rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, 0);
4694
4695 rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, 0);
4696 rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, 0);
4697 rtnl_register(PF_UNSPEC, RTM_GETNETCONF, NULL, rtnl_dump_all, 0);
4698
4699 rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, 0);
4700 rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, 0);
4701 rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, 0);
4702
4703 rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, 0);
4704 rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, 0);
4705 rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, 0);
4706
4707 rtnl_register(PF_UNSPEC, RTM_GETSTATS, rtnl_stats_get, rtnl_stats_dump,
4708 0);
4709 }
ubuntu-bionic-kernel mirror