?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
x86/speculation/mds: Fix documentation typo
[linux/fpc-iii.git] / net / core / rtnetlink.c
blob925af6b430179ab01da503c5b5ba97046809bdcb
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(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(&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 * The caller must hold the rtnl_mutex.
483 */
484 void __rtnl_af_unregister(struct rtnl_af_ops *ops)
485 {
486 list_del(&ops->list);
487 }
488 EXPORT_SYMBOL_GPL(__rtnl_af_unregister);
489
490 /**
491 * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
492 * @ops: struct rtnl_af_ops * to unregister
493 */
494 void rtnl_af_unregister(struct rtnl_af_ops *ops)
495 {
496 rtnl_lock();
497 __rtnl_af_unregister(ops);
498 rtnl_unlock();
499 }
500 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
501
502 static size_t rtnl_link_get_af_size(const struct net_device *dev,
503 u32 ext_filter_mask)
504 {
505 struct rtnl_af_ops *af_ops;
506 size_t size;
507
508 /* IFLA_AF_SPEC */
509 size = nla_total_size(sizeof(struct nlattr));
510
511 list_for_each_entry(af_ops, &rtnl_af_ops, list) {
512 if (af_ops->get_link_af_size) {
513 /* AF_* + nested data */
514 size += nla_total_size(sizeof(struct nlattr)) +
515 af_ops->get_link_af_size(dev, ext_filter_mask);
516 }
517 }
518
519 return size;
520 }
521
522 static bool rtnl_have_link_slave_info(const struct net_device *dev)
523 {
524 struct net_device *master_dev;
525
526 master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
527 if (master_dev && master_dev->rtnl_link_ops)
528 return true;
529 return false;
530 }
531
532 static int rtnl_link_slave_info_fill(struct sk_buff *skb,
533 const struct net_device *dev)
534 {
535 struct net_device *master_dev;
536 const struct rtnl_link_ops *ops;
537 struct nlattr *slave_data;
538 int err;
539
540 master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
541 if (!master_dev)
542 return 0;
543 ops = master_dev->rtnl_link_ops;
544 if (!ops)
545 return 0;
546 if (nla_put_string(skb, IFLA_INFO_SLAVE_KIND, ops->kind) < 0)
547 return -EMSGSIZE;
548 if (ops->fill_slave_info) {
549 slave_data = nla_nest_start(skb, IFLA_INFO_SLAVE_DATA);
550 if (!slave_data)
551 return -EMSGSIZE;
552 err = ops->fill_slave_info(skb, master_dev, dev);
553 if (err < 0)
554 goto err_cancel_slave_data;
555 nla_nest_end(skb, slave_data);
556 }
557 return 0;
558
559 err_cancel_slave_data:
560 nla_nest_cancel(skb, slave_data);
561 return err;
562 }
563
564 static int rtnl_link_info_fill(struct sk_buff *skb,
565 const struct net_device *dev)
566 {
567 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
568 struct nlattr *data;
569 int err;
570
571 if (!ops)
572 return 0;
573 if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
574 return -EMSGSIZE;
575 if (ops->fill_xstats) {
576 err = ops->fill_xstats(skb, dev);
577 if (err < 0)
578 return err;
579 }
580 if (ops->fill_info) {
581 data = nla_nest_start(skb, IFLA_INFO_DATA);
582 if (data == NULL)
583 return -EMSGSIZE;
584 err = ops->fill_info(skb, dev);
585 if (err < 0)
586 goto err_cancel_data;
587 nla_nest_end(skb, data);
588 }
589 return 0;
590
591 err_cancel_data:
592 nla_nest_cancel(skb, data);
593 return err;
594 }
595
596 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
597 {
598 struct nlattr *linkinfo;
599 int err = -EMSGSIZE;
600
601 linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
602 if (linkinfo == NULL)
603 goto out;
604
605 err = rtnl_link_info_fill(skb, dev);
606 if (err < 0)
607 goto err_cancel_link;
608
609 err = rtnl_link_slave_info_fill(skb, dev);
610 if (err < 0)
611 goto err_cancel_link;
612
613 nla_nest_end(skb, linkinfo);
614 return 0;
615
616 err_cancel_link:
617 nla_nest_cancel(skb, linkinfo);
618 out:
619 return err;
620 }
621
622 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
623 {
624 struct sock *rtnl = net->rtnl;
625 int err = 0;
626
627 NETLINK_CB(skb).dst_group = group;
628 if (echo)
629 refcount_inc(&skb->users);
630 netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
631 if (echo)
632 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
633 return err;
634 }
635
636 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
637 {
638 struct sock *rtnl = net->rtnl;
639
640 return nlmsg_unicast(rtnl, skb, pid);
641 }
642 EXPORT_SYMBOL(rtnl_unicast);
643
644 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
645 struct nlmsghdr *nlh, gfp_t flags)
646 {
647 struct sock *rtnl = net->rtnl;
648 int report = 0;
649
650 if (nlh)
651 report = nlmsg_report(nlh);
652
653 nlmsg_notify(rtnl, skb, pid, group, report, flags);
654 }
655 EXPORT_SYMBOL(rtnl_notify);
656
657 void rtnl_set_sk_err(struct net *net, u32 group, int error)
658 {
659 struct sock *rtnl = net->rtnl;
660
661 netlink_set_err(rtnl, 0, group, error);
662 }
663 EXPORT_SYMBOL(rtnl_set_sk_err);
664
665 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
666 {
667 struct nlattr *mx;
668 int i, valid = 0;
669
670 mx = nla_nest_start(skb, RTA_METRICS);
671 if (mx == NULL)
672 return -ENOBUFS;
673
674 for (i = 0; i < RTAX_MAX; i++) {
675 if (metrics[i]) {
676 if (i == RTAX_CC_ALGO - 1) {
677 char tmp[TCP_CA_NAME_MAX], *name;
678
679 name = tcp_ca_get_name_by_key(metrics[i], tmp);
680 if (!name)
681 continue;
682 if (nla_put_string(skb, i + 1, name))
683 goto nla_put_failure;
684 } else if (i == RTAX_FEATURES - 1) {
685 u32 user_features = metrics[i] & RTAX_FEATURE_MASK;
686
687 if (!user_features)
688 continue;
689 BUILD_BUG_ON(RTAX_FEATURE_MASK & DST_FEATURE_MASK);
690 if (nla_put_u32(skb, i + 1, user_features))
691 goto nla_put_failure;
692 } else {
693 if (nla_put_u32(skb, i + 1, metrics[i]))
694 goto nla_put_failure;
695 }
696 valid++;
697 }
698 }
699
700 if (!valid) {
701 nla_nest_cancel(skb, mx);
702 return 0;
703 }
704
705 return nla_nest_end(skb, mx);
706
707 nla_put_failure:
708 nla_nest_cancel(skb, mx);
709 return -EMSGSIZE;
710 }
711 EXPORT_SYMBOL(rtnetlink_put_metrics);
712
713 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
714 long expires, u32 error)
715 {
716 struct rta_cacheinfo ci = {
717 .rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse),
718 .rta_used = dst->__use,
719 .rta_clntref = atomic_read(&(dst->__refcnt)),
720 .rta_error = error,
721 .rta_id = id,
722 };
723
724 if (expires) {
725 unsigned long clock;
726
727 clock = jiffies_to_clock_t(abs(expires));
728 clock = min_t(unsigned long, clock, INT_MAX);
729 ci.rta_expires = (expires > 0) ? clock : -clock;
730 }
731 return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
732 }
733 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
734
735 static void set_operstate(struct net_device *dev, unsigned char transition)
736 {
737 unsigned char operstate = dev->operstate;
738
739 switch (transition) {
740 case IF_OPER_UP:
741 if ((operstate == IF_OPER_DORMANT ||
742 operstate == IF_OPER_UNKNOWN) &&
743 !netif_dormant(dev))
744 operstate = IF_OPER_UP;
745 break;
746
747 case IF_OPER_DORMANT:
748 if (operstate == IF_OPER_UP ||
749 operstate == IF_OPER_UNKNOWN)
750 operstate = IF_OPER_DORMANT;
751 break;
752 }
753
754 if (dev->operstate != operstate) {
755 write_lock_bh(&dev_base_lock);
756 dev->operstate = operstate;
757 write_unlock_bh(&dev_base_lock);
758 netdev_state_change(dev);
759 }
760 }
761
762 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
763 {
764 return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
765 (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
766 }
767
768 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
769 const struct ifinfomsg *ifm)
770 {
771 unsigned int flags = ifm->ifi_flags;
772
773 /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
774 if (ifm->ifi_change)
775 flags = (flags & ifm->ifi_change) |
776 (rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
777
778 return flags;
779 }
780
781 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
782 const struct rtnl_link_stats64 *b)
783 {
784 a->rx_packets = b->rx_packets;
785 a->tx_packets = b->tx_packets;
786 a->rx_bytes = b->rx_bytes;
787 a->tx_bytes = b->tx_bytes;
788 a->rx_errors = b->rx_errors;
789 a->tx_errors = b->tx_errors;
790 a->rx_dropped = b->rx_dropped;
791 a->tx_dropped = b->tx_dropped;
792
793 a->multicast = b->multicast;
794 a->collisions = b->collisions;
795
796 a->rx_length_errors = b->rx_length_errors;
797 a->rx_over_errors = b->rx_over_errors;
798 a->rx_crc_errors = b->rx_crc_errors;
799 a->rx_frame_errors = b->rx_frame_errors;
800 a->rx_fifo_errors = b->rx_fifo_errors;
801 a->rx_missed_errors = b->rx_missed_errors;
802
803 a->tx_aborted_errors = b->tx_aborted_errors;
804 a->tx_carrier_errors = b->tx_carrier_errors;
805 a->tx_fifo_errors = b->tx_fifo_errors;
806 a->tx_heartbeat_errors = b->tx_heartbeat_errors;
807 a->tx_window_errors = b->tx_window_errors;
808
809 a->rx_compressed = b->rx_compressed;
810 a->tx_compressed = b->tx_compressed;
811
812 a->rx_nohandler = b->rx_nohandler;
813 }
814
815 /* All VF info */
816 static inline int rtnl_vfinfo_size(const struct net_device *dev,
817 u32 ext_filter_mask)
818 {
819 if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF)) {
820 int num_vfs = dev_num_vf(dev->dev.parent);
821 size_t size = nla_total_size(0);
822 size += num_vfs *
823 (nla_total_size(0) +
824 nla_total_size(sizeof(struct ifla_vf_mac)) +
825 nla_total_size(sizeof(struct ifla_vf_vlan)) +
826 nla_total_size(0) + /* nest IFLA_VF_VLAN_LIST */
827 nla_total_size(MAX_VLAN_LIST_LEN *
828 sizeof(struct ifla_vf_vlan_info)) +
829 nla_total_size(sizeof(struct ifla_vf_spoofchk)) +
830 nla_total_size(sizeof(struct ifla_vf_tx_rate)) +
831 nla_total_size(sizeof(struct ifla_vf_rate)) +
832 nla_total_size(sizeof(struct ifla_vf_link_state)) +
833 nla_total_size(sizeof(struct ifla_vf_rss_query_en)) +
834 nla_total_size(0) + /* nest IFLA_VF_STATS */
835 /* IFLA_VF_STATS_RX_PACKETS */
836 nla_total_size_64bit(sizeof(__u64)) +
837 /* IFLA_VF_STATS_TX_PACKETS */
838 nla_total_size_64bit(sizeof(__u64)) +
839 /* IFLA_VF_STATS_RX_BYTES */
840 nla_total_size_64bit(sizeof(__u64)) +
841 /* IFLA_VF_STATS_TX_BYTES */
842 nla_total_size_64bit(sizeof(__u64)) +
843 /* IFLA_VF_STATS_BROADCAST */
844 nla_total_size_64bit(sizeof(__u64)) +
845 /* IFLA_VF_STATS_MULTICAST */
846 nla_total_size_64bit(sizeof(__u64)) +
847 nla_total_size(sizeof(struct ifla_vf_trust)));
848 return size;
849 } else
850 return 0;
851 }
852
853 static size_t rtnl_port_size(const struct net_device *dev,
854 u32 ext_filter_mask)
855 {
856 size_t port_size = nla_total_size(4) /* PORT_VF */
857 + nla_total_size(PORT_PROFILE_MAX) /* PORT_PROFILE */
858 + nla_total_size(PORT_UUID_MAX) /* PORT_INSTANCE_UUID */
859 + nla_total_size(PORT_UUID_MAX) /* PORT_HOST_UUID */
860 + nla_total_size(1) /* PROT_VDP_REQUEST */
861 + nla_total_size(2); /* PORT_VDP_RESPONSE */
862 size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
863 size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
864 + port_size;
865 size_t port_self_size = nla_total_size(sizeof(struct nlattr))
866 + port_size;
867
868 if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
869 !(ext_filter_mask & RTEXT_FILTER_VF))
870 return 0;
871 if (dev_num_vf(dev->dev.parent))
872 return port_self_size + vf_ports_size +
873 vf_port_size * dev_num_vf(dev->dev.parent);
874 else
875 return port_self_size;
876 }
877
878 static size_t rtnl_xdp_size(void)
879 {
880 size_t xdp_size = nla_total_size(0) + /* nest IFLA_XDP */
881 nla_total_size(1) + /* XDP_ATTACHED */
882 nla_total_size(4); /* XDP_PROG_ID */
883
884 return xdp_size;
885 }
886
887 static noinline size_t if_nlmsg_size(const struct net_device *dev,
888 u32 ext_filter_mask)
889 {
890 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
891 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
892 + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
893 + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
894 + nla_total_size_64bit(sizeof(struct rtnl_link_ifmap))
895 + nla_total_size(sizeof(struct rtnl_link_stats))
896 + nla_total_size_64bit(sizeof(struct rtnl_link_stats64))
897 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
898 + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
899 + nla_total_size(4) /* IFLA_TXQLEN */
900 + nla_total_size(4) /* IFLA_WEIGHT */
901 + nla_total_size(4) /* IFLA_MTU */
902 + nla_total_size(4) /* IFLA_LINK */
903 + nla_total_size(4) /* IFLA_MASTER */
904 + nla_total_size(1) /* IFLA_CARRIER */
905 + nla_total_size(4) /* IFLA_PROMISCUITY */
906 + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
907 + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
908 + nla_total_size(4) /* IFLA_GSO_MAX_SEGS */
909 + nla_total_size(4) /* IFLA_GSO_MAX_SIZE */
910 + nla_total_size(1) /* IFLA_OPERSTATE */
911 + nla_total_size(1) /* IFLA_LINKMODE */
912 + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
913 + nla_total_size(4) /* IFLA_LINK_NETNSID */
914 + nla_total_size(4) /* IFLA_GROUP */
915 + nla_total_size(ext_filter_mask
916 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
917 + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
918 + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
919 + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
920 + rtnl_link_get_af_size(dev, ext_filter_mask) /* IFLA_AF_SPEC */
921 + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */
922 + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_SWITCH_ID */
923 + nla_total_size(IFNAMSIZ) /* IFLA_PHYS_PORT_NAME */
924 + rtnl_xdp_size() /* IFLA_XDP */
925 + nla_total_size(4) /* IFLA_EVENT */
926 + nla_total_size(1); /* IFLA_PROTO_DOWN */
927
928 }
929
930 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
931 {
932 struct nlattr *vf_ports;
933 struct nlattr *vf_port;
934 int vf;
935 int err;
936
937 vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
938 if (!vf_ports)
939 return -EMSGSIZE;
940
941 for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
942 vf_port = nla_nest_start(skb, IFLA_VF_PORT);
943 if (!vf_port)
944 goto nla_put_failure;
945 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
946 goto nla_put_failure;
947 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
948 if (err == -EMSGSIZE)
949 goto nla_put_failure;
950 if (err) {
951 nla_nest_cancel(skb, vf_port);
952 continue;
953 }
954 nla_nest_end(skb, vf_port);
955 }
956
957 nla_nest_end(skb, vf_ports);
958
959 return 0;
960
961 nla_put_failure:
962 nla_nest_cancel(skb, vf_ports);
963 return -EMSGSIZE;
964 }
965
966 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
967 {
968 struct nlattr *port_self;
969 int err;
970
971 port_self = nla_nest_start(skb, IFLA_PORT_SELF);
972 if (!port_self)
973 return -EMSGSIZE;
974
975 err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
976 if (err) {
977 nla_nest_cancel(skb, port_self);
978 return (err == -EMSGSIZE) ? err : 0;
979 }
980
981 nla_nest_end(skb, port_self);
982
983 return 0;
984 }
985
986 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
987 u32 ext_filter_mask)
988 {
989 int err;
990
991 if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
992 !(ext_filter_mask & RTEXT_FILTER_VF))
993 return 0;
994
995 err = rtnl_port_self_fill(skb, dev);
996 if (err)
997 return err;
998
999 if (dev_num_vf(dev->dev.parent)) {
1000 err = rtnl_vf_ports_fill(skb, dev);
1001 if (err)
1002 return err;
1003 }
1004
1005 return 0;
1006 }
1007
1008 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
1009 {
1010 int err;
1011 struct netdev_phys_item_id ppid;
1012
1013 err = dev_get_phys_port_id(dev, &ppid);
1014 if (err) {
1015 if (err == -EOPNOTSUPP)
1016 return 0;
1017 return err;
1018 }
1019
1020 if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
1021 return -EMSGSIZE;
1022
1023 return 0;
1024 }
1025
1026 static int rtnl_phys_port_name_fill(struct sk_buff *skb, struct net_device *dev)
1027 {
1028 char name[IFNAMSIZ];
1029 int err;
1030
1031 err = dev_get_phys_port_name(dev, name, sizeof(name));
1032 if (err) {
1033 if (err == -EOPNOTSUPP)
1034 return 0;
1035 return err;
1036 }
1037
1038 if (nla_put_string(skb, IFLA_PHYS_PORT_NAME, name))
1039 return -EMSGSIZE;
1040
1041 return 0;
1042 }
1043
1044 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev)
1045 {
1046 int err;
1047 struct switchdev_attr attr = {
1048 .orig_dev = dev,
1049 .id = SWITCHDEV_ATTR_ID_PORT_PARENT_ID,
1050 .flags = SWITCHDEV_F_NO_RECURSE,
1051 };
1052
1053 err = switchdev_port_attr_get(dev, &attr);
1054 if (err) {
1055 if (err == -EOPNOTSUPP)
1056 return 0;
1057 return err;
1058 }
1059
1060 if (nla_put(skb, IFLA_PHYS_SWITCH_ID, attr.u.ppid.id_len,
1061 attr.u.ppid.id))
1062 return -EMSGSIZE;
1063
1064 return 0;
1065 }
1066
1067 static noinline_for_stack int rtnl_fill_stats(struct sk_buff *skb,
1068 struct net_device *dev)
1069 {
1070 struct rtnl_link_stats64 *sp;
1071 struct nlattr *attr;
1072
1073 attr = nla_reserve_64bit(skb, IFLA_STATS64,
1074 sizeof(struct rtnl_link_stats64), IFLA_PAD);
1075 if (!attr)
1076 return -EMSGSIZE;
1077
1078 sp = nla_data(attr);
1079 dev_get_stats(dev, sp);
1080
1081 attr = nla_reserve(skb, IFLA_STATS,
1082 sizeof(struct rtnl_link_stats));
1083 if (!attr)
1084 return -EMSGSIZE;
1085
1086 copy_rtnl_link_stats(nla_data(attr), sp);
1087
1088 return 0;
1089 }
1090
1091 static noinline_for_stack int rtnl_fill_vfinfo(struct sk_buff *skb,
1092 struct net_device *dev,
1093 int vfs_num,
1094 struct nlattr *vfinfo)
1095 {
1096 struct ifla_vf_rss_query_en vf_rss_query_en;
1097 struct nlattr *vf, *vfstats, *vfvlanlist;
1098 struct ifla_vf_link_state vf_linkstate;
1099 struct ifla_vf_vlan_info vf_vlan_info;
1100 struct ifla_vf_spoofchk vf_spoofchk;
1101 struct ifla_vf_tx_rate vf_tx_rate;
1102 struct ifla_vf_stats vf_stats;
1103 struct ifla_vf_trust vf_trust;
1104 struct ifla_vf_vlan vf_vlan;
1105 struct ifla_vf_rate vf_rate;
1106 struct ifla_vf_mac vf_mac;
1107 struct ifla_vf_info ivi;
1108
1109 memset(&ivi, 0, sizeof(ivi));
1110
1111 /* Not all SR-IOV capable drivers support the
1112 * spoofcheck and "RSS query enable" query. Preset to
1113 * -1 so the user space tool can detect that the driver
1114 * didn't report anything.
1115 */
1116 ivi.spoofchk = -1;
1117 ivi.rss_query_en = -1;
1118 ivi.trusted = -1;
1119 /* The default value for VF link state is "auto"
1120 * IFLA_VF_LINK_STATE_AUTO which equals zero
1121 */
1122 ivi.linkstate = 0;
1123 /* VLAN Protocol by default is 802.1Q */
1124 ivi.vlan_proto = htons(ETH_P_8021Q);
1125 if (dev->netdev_ops->ndo_get_vf_config(dev, vfs_num, &ivi))
1126 return 0;
1127
1128 memset(&vf_vlan_info, 0, sizeof(vf_vlan_info));
1129
1130 vf_mac.vf =
1131 vf_vlan.vf =
1132 vf_vlan_info.vf =
1133 vf_rate.vf =
1134 vf_tx_rate.vf =
1135 vf_spoofchk.vf =
1136 vf_linkstate.vf =
1137 vf_rss_query_en.vf =
1138 vf_trust.vf = ivi.vf;
1139
1140 memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1141 vf_vlan.vlan = ivi.vlan;
1142 vf_vlan.qos = ivi.qos;
1143 vf_vlan_info.vlan = ivi.vlan;
1144 vf_vlan_info.qos = ivi.qos;
1145 vf_vlan_info.vlan_proto = ivi.vlan_proto;
1146 vf_tx_rate.rate = ivi.max_tx_rate;
1147 vf_rate.min_tx_rate = ivi.min_tx_rate;
1148 vf_rate.max_tx_rate = ivi.max_tx_rate;
1149 vf_spoofchk.setting = ivi.spoofchk;
1150 vf_linkstate.link_state = ivi.linkstate;
1151 vf_rss_query_en.setting = ivi.rss_query_en;
1152 vf_trust.setting = ivi.trusted;
1153 vf = nla_nest_start(skb, IFLA_VF_INFO);
1154 if (!vf)
1155 goto nla_put_vfinfo_failure;
1156 if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1157 nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1158 nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1159 &vf_rate) ||
1160 nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1161 &vf_tx_rate) ||
1162 nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1163 &vf_spoofchk) ||
1164 nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1165 &vf_linkstate) ||
1166 nla_put(skb, IFLA_VF_RSS_QUERY_EN,
1167 sizeof(vf_rss_query_en),
1168 &vf_rss_query_en) ||
1169 nla_put(skb, IFLA_VF_TRUST,
1170 sizeof(vf_trust), &vf_trust))
1171 goto nla_put_vf_failure;
1172 vfvlanlist = nla_nest_start(skb, IFLA_VF_VLAN_LIST);
1173 if (!vfvlanlist)
1174 goto nla_put_vf_failure;
1175 if (nla_put(skb, IFLA_VF_VLAN_INFO, sizeof(vf_vlan_info),
1176 &vf_vlan_info)) {
1177 nla_nest_cancel(skb, vfvlanlist);
1178 goto nla_put_vf_failure;
1179 }
1180 nla_nest_end(skb, vfvlanlist);
1181 memset(&vf_stats, 0, sizeof(vf_stats));
1182 if (dev->netdev_ops->ndo_get_vf_stats)
1183 dev->netdev_ops->ndo_get_vf_stats(dev, vfs_num,
1184 &vf_stats);
1185 vfstats = nla_nest_start(skb, IFLA_VF_STATS);
1186 if (!vfstats)
1187 goto nla_put_vf_failure;
1188 if (nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_PACKETS,
1189 vf_stats.rx_packets, IFLA_VF_STATS_PAD) ||
1190 nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_PACKETS,
1191 vf_stats.tx_packets, IFLA_VF_STATS_PAD) ||
1192 nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_BYTES,
1193 vf_stats.rx_bytes, IFLA_VF_STATS_PAD) ||
1194 nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_BYTES,
1195 vf_stats.tx_bytes, IFLA_VF_STATS_PAD) ||
1196 nla_put_u64_64bit(skb, IFLA_VF_STATS_BROADCAST,
1197 vf_stats.broadcast, IFLA_VF_STATS_PAD) ||
1198 nla_put_u64_64bit(skb, IFLA_VF_STATS_MULTICAST,
1199 vf_stats.multicast, IFLA_VF_STATS_PAD)) {
1200 nla_nest_cancel(skb, vfstats);
1201 goto nla_put_vf_failure;
1202 }
1203 nla_nest_end(skb, vfstats);
1204 nla_nest_end(skb, vf);
1205 return 0;
1206
1207 nla_put_vf_failure:
1208 nla_nest_cancel(skb, vf);
1209 nla_put_vfinfo_failure:
1210 nla_nest_cancel(skb, vfinfo);
1211 return -EMSGSIZE;
1212 }
1213
1214 static int rtnl_fill_link_ifmap(struct sk_buff *skb, struct net_device *dev)
1215 {
1216 struct rtnl_link_ifmap map;
1217
1218 memset(&map, 0, sizeof(map));
1219 map.mem_start = dev->mem_start;
1220 map.mem_end = dev->mem_end;
1221 map.base_addr = dev->base_addr;
1222 map.irq = dev->irq;
1223 map.dma = dev->dma;
1224 map.port = dev->if_port;
1225
1226 if (nla_put_64bit(skb, IFLA_MAP, sizeof(map), &map, IFLA_PAD))
1227 return -EMSGSIZE;
1228
1229 return 0;
1230 }
1231
1232 static u8 rtnl_xdp_attached_mode(struct net_device *dev, u32 *prog_id)
1233 {
1234 const struct net_device_ops *ops = dev->netdev_ops;
1235 const struct bpf_prog *generic_xdp_prog;
1236
1237 ASSERT_RTNL();
1238
1239 *prog_id = 0;
1240 generic_xdp_prog = rtnl_dereference(dev->xdp_prog);
1241 if (generic_xdp_prog) {
1242 *prog_id = generic_xdp_prog->aux->id;
1243 return XDP_ATTACHED_SKB;
1244 }
1245 if (!ops->ndo_xdp)
1246 return XDP_ATTACHED_NONE;
1247
1248 return __dev_xdp_attached(dev, ops->ndo_xdp, prog_id);
1249 }
1250
1251 static int rtnl_xdp_fill(struct sk_buff *skb, struct net_device *dev)
1252 {
1253 struct nlattr *xdp;
1254 u32 prog_id;
1255 int err;
1256
1257 xdp = nla_nest_start(skb, IFLA_XDP);
1258 if (!xdp)
1259 return -EMSGSIZE;
1260
1261 err = nla_put_u8(skb, IFLA_XDP_ATTACHED,
1262 rtnl_xdp_attached_mode(dev, &prog_id));
1263 if (err)
1264 goto err_cancel;
1265
1266 if (prog_id) {
1267 err = nla_put_u32(skb, IFLA_XDP_PROG_ID, prog_id);
1268 if (err)
1269 goto err_cancel;
1270 }
1271
1272 nla_nest_end(skb, xdp);
1273 return 0;
1274
1275 err_cancel:
1276 nla_nest_cancel(skb, xdp);
1277 return err;
1278 }
1279
1280 static u32 rtnl_get_event(unsigned long event)
1281 {
1282 u32 rtnl_event_type = IFLA_EVENT_NONE;
1283
1284 switch (event) {
1285 case NETDEV_REBOOT:
1286 rtnl_event_type = IFLA_EVENT_REBOOT;
1287 break;
1288 case NETDEV_FEAT_CHANGE:
1289 rtnl_event_type = IFLA_EVENT_FEATURES;
1290 break;
1291 case NETDEV_BONDING_FAILOVER:
1292 rtnl_event_type = IFLA_EVENT_BONDING_FAILOVER;
1293 break;
1294 case NETDEV_NOTIFY_PEERS:
1295 rtnl_event_type = IFLA_EVENT_NOTIFY_PEERS;
1296 break;
1297 case NETDEV_RESEND_IGMP:
1298 rtnl_event_type = IFLA_EVENT_IGMP_RESEND;
1299 break;
1300 case NETDEV_CHANGEINFODATA:
1301 rtnl_event_type = IFLA_EVENT_BONDING_OPTIONS;
1302 break;
1303 default:
1304 break;
1305 }
1306
1307 return rtnl_event_type;
1308 }
1309
1310 static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
1311 int type, u32 pid, u32 seq, u32 change,
1312 unsigned int flags, u32 ext_filter_mask,
1313 u32 event)
1314 {
1315 struct ifinfomsg *ifm;
1316 struct nlmsghdr *nlh;
1317 struct nlattr *af_spec;
1318 struct rtnl_af_ops *af_ops;
1319 struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1320
1321 ASSERT_RTNL();
1322 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
1323 if (nlh == NULL)
1324 return -EMSGSIZE;
1325
1326 ifm = nlmsg_data(nlh);
1327 ifm->ifi_family = AF_UNSPEC;
1328 ifm->__ifi_pad = 0;
1329 ifm->ifi_type = dev->type;
1330 ifm->ifi_index = dev->ifindex;
1331 ifm->ifi_flags = dev_get_flags(dev);
1332 ifm->ifi_change = change;
1333
1334 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
1335 nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
1336 nla_put_u8(skb, IFLA_OPERSTATE,
1337 netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
1338 nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
1339 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
1340 nla_put_u32(skb, IFLA_GROUP, dev->group) ||
1341 nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
1342 nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
1343 nla_put_u32(skb, IFLA_GSO_MAX_SEGS, dev->gso_max_segs) ||
1344 nla_put_u32(skb, IFLA_GSO_MAX_SIZE, dev->gso_max_size) ||
1345 #ifdef CONFIG_RPS
1346 nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1347 #endif
1348 (dev->ifindex != dev_get_iflink(dev) &&
1349 nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
1350 (upper_dev &&
1351 nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex)) ||
1352 nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1353 (dev->qdisc &&
1354 nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
1355 (dev->ifalias &&
1356 nla_put_string(skb, IFLA_IFALIAS, dev->ifalias)) ||
1357 nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1358 atomic_read(&dev->carrier_changes)) ||
1359 nla_put_u8(skb, IFLA_PROTO_DOWN, dev->proto_down))
1360 goto nla_put_failure;
1361
1362 if (event != IFLA_EVENT_NONE) {
1363 if (nla_put_u32(skb, IFLA_EVENT, event))
1364 goto nla_put_failure;
1365 }
1366
1367 if (rtnl_fill_link_ifmap(skb, dev))
1368 goto nla_put_failure;
1369
1370 if (dev->addr_len) {
1371 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1372 nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1373 goto nla_put_failure;
1374 }
1375
1376 if (rtnl_phys_port_id_fill(skb, dev))
1377 goto nla_put_failure;
1378
1379 if (rtnl_phys_port_name_fill(skb, dev))
1380 goto nla_put_failure;
1381
1382 if (rtnl_phys_switch_id_fill(skb, dev))
1383 goto nla_put_failure;
1384
1385 if (rtnl_fill_stats(skb, dev))
1386 goto nla_put_failure;
1387
1388 if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF) &&
1389 nla_put_u32(skb, IFLA_NUM_VF, dev_num_vf(dev->dev.parent)))
1390 goto nla_put_failure;
1391
1392 if (dev->netdev_ops->ndo_get_vf_config && dev->dev.parent &&
1393 ext_filter_mask & RTEXT_FILTER_VF) {
1394 int i;
1395 struct nlattr *vfinfo;
1396 int num_vfs = dev_num_vf(dev->dev.parent);
1397
1398 vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
1399 if (!vfinfo)
1400 goto nla_put_failure;
1401 for (i = 0; i < num_vfs; i++) {
1402 if (rtnl_fill_vfinfo(skb, dev, i, vfinfo))
1403 goto nla_put_failure;
1404 }
1405
1406 nla_nest_end(skb, vfinfo);
1407 }
1408
1409 if (rtnl_port_fill(skb, dev, ext_filter_mask))
1410 goto nla_put_failure;
1411
1412 if (rtnl_xdp_fill(skb, dev))
1413 goto nla_put_failure;
1414
1415 if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1416 if (rtnl_link_fill(skb, dev) < 0)
1417 goto nla_put_failure;
1418 }
1419
1420 if (dev->rtnl_link_ops &&
1421 dev->rtnl_link_ops->get_link_net) {
1422 struct net *link_net = dev->rtnl_link_ops->get_link_net(dev);
1423
1424 if (!net_eq(dev_net(dev), link_net)) {
1425 int id = peernet2id_alloc(dev_net(dev), link_net);
1426
1427 if (nla_put_s32(skb, IFLA_LINK_NETNSID, id))
1428 goto nla_put_failure;
1429 }
1430 }
1431
1432 if (!(af_spec = nla_nest_start(skb, IFLA_AF_SPEC)))
1433 goto nla_put_failure;
1434
1435 list_for_each_entry(af_ops, &rtnl_af_ops, list) {
1436 if (af_ops->fill_link_af) {
1437 struct nlattr *af;
1438 int err;
1439
1440 if (!(af = nla_nest_start(skb, af_ops->family)))
1441 goto nla_put_failure;
1442
1443 err = af_ops->fill_link_af(skb, dev, ext_filter_mask);
1444
1445 /*
1446 * Caller may return ENODATA to indicate that there
1447 * was no data to be dumped. This is not an error, it
1448 * means we should trim the attribute header and
1449 * continue.
1450 */
1451 if (err == -ENODATA)
1452 nla_nest_cancel(skb, af);
1453 else if (err < 0)
1454 goto nla_put_failure;
1455
1456 nla_nest_end(skb, af);
1457 }
1458 }
1459
1460 nla_nest_end(skb, af_spec);
1461
1462 nlmsg_end(skb, nlh);
1463 return 0;
1464
1465 nla_put_failure:
1466 nlmsg_cancel(skb, nlh);
1467 return -EMSGSIZE;
1468 }
1469
1470 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1471 [IFLA_IFNAME] = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1472 [IFLA_ADDRESS] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1473 [IFLA_BROADCAST] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1474 [IFLA_MAP] = { .len = sizeof(struct rtnl_link_ifmap) },
1475 [IFLA_MTU] = { .type = NLA_U32 },
1476 [IFLA_LINK] = { .type = NLA_U32 },
1477 [IFLA_MASTER] = { .type = NLA_U32 },
1478 [IFLA_CARRIER] = { .type = NLA_U8 },
1479 [IFLA_TXQLEN] = { .type = NLA_U32 },
1480 [IFLA_WEIGHT] = { .type = NLA_U32 },
1481 [IFLA_OPERSTATE] = { .type = NLA_U8 },
1482 [IFLA_LINKMODE] = { .type = NLA_U8 },
1483 [IFLA_LINKINFO] = { .type = NLA_NESTED },
1484 [IFLA_NET_NS_PID] = { .type = NLA_U32 },
1485 [IFLA_NET_NS_FD] = { .type = NLA_U32 },
1486 /* IFLA_IFALIAS is a string, but policy is set to NLA_BINARY to
1487 * allow 0-length string (needed to remove an alias).
1488 */
1489 [IFLA_IFALIAS] = { .type = NLA_BINARY, .len = IFALIASZ - 1 },
1490 [IFLA_VFINFO_LIST] = {. type = NLA_NESTED },
1491 [IFLA_VF_PORTS] = { .type = NLA_NESTED },
1492 [IFLA_PORT_SELF] = { .type = NLA_NESTED },
1493 [IFLA_AF_SPEC] = { .type = NLA_NESTED },
1494 [IFLA_EXT_MASK] = { .type = NLA_U32 },
1495 [IFLA_PROMISCUITY] = { .type = NLA_U32 },
1496 [IFLA_NUM_TX_QUEUES] = { .type = NLA_U32 },
1497 [IFLA_NUM_RX_QUEUES] = { .type = NLA_U32 },
1498 [IFLA_PHYS_PORT_ID] = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1499 [IFLA_CARRIER_CHANGES] = { .type = NLA_U32 }, /* ignored */
1500 [IFLA_PHYS_SWITCH_ID] = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1501 [IFLA_LINK_NETNSID] = { .type = NLA_S32 },
1502 [IFLA_PROTO_DOWN] = { .type = NLA_U8 },
1503 [IFLA_XDP] = { .type = NLA_NESTED },
1504 [IFLA_EVENT] = { .type = NLA_U32 },
1505 [IFLA_GROUP] = { .type = NLA_U32 },
1506 };
1507
1508 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1509 [IFLA_INFO_KIND] = { .type = NLA_STRING },
1510 [IFLA_INFO_DATA] = { .type = NLA_NESTED },
1511 [IFLA_INFO_SLAVE_KIND] = { .type = NLA_STRING },
1512 [IFLA_INFO_SLAVE_DATA] = { .type = NLA_NESTED },
1513 };
1514
1515 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1516 [IFLA_VF_MAC] = { .len = sizeof(struct ifla_vf_mac) },
1517 [IFLA_VF_VLAN] = { .len = sizeof(struct ifla_vf_vlan) },
1518 [IFLA_VF_VLAN_LIST] = { .type = NLA_NESTED },
1519 [IFLA_VF_TX_RATE] = { .len = sizeof(struct ifla_vf_tx_rate) },
1520 [IFLA_VF_SPOOFCHK] = { .len = sizeof(struct ifla_vf_spoofchk) },
1521 [IFLA_VF_RATE] = { .len = sizeof(struct ifla_vf_rate) },
1522 [IFLA_VF_LINK_STATE] = { .len = sizeof(struct ifla_vf_link_state) },
1523 [IFLA_VF_RSS_QUERY_EN] = { .len = sizeof(struct ifla_vf_rss_query_en) },
1524 [IFLA_VF_STATS] = { .type = NLA_NESTED },
1525 [IFLA_VF_TRUST] = { .len = sizeof(struct ifla_vf_trust) },
1526 [IFLA_VF_IB_NODE_GUID] = { .len = sizeof(struct ifla_vf_guid) },
1527 [IFLA_VF_IB_PORT_GUID] = { .len = sizeof(struct ifla_vf_guid) },
1528 };
1529
1530 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1531 [IFLA_PORT_VF] = { .type = NLA_U32 },
1532 [IFLA_PORT_PROFILE] = { .type = NLA_STRING,
1533 .len = PORT_PROFILE_MAX },
1534 [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1535 .len = PORT_UUID_MAX },
1536 [IFLA_PORT_HOST_UUID] = { .type = NLA_STRING,
1537 .len = PORT_UUID_MAX },
1538 [IFLA_PORT_REQUEST] = { .type = NLA_U8, },
1539 [IFLA_PORT_RESPONSE] = { .type = NLA_U16, },
1540
1541 /* Unused, but we need to keep it here since user space could
1542 * fill it. It's also broken with regard to NLA_BINARY use in
1543 * combination with structs.
1544 */
1545 [IFLA_PORT_VSI_TYPE] = { .type = NLA_BINARY,
1546 .len = sizeof(struct ifla_port_vsi) },
1547 };
1548
1549 static const struct nla_policy ifla_xdp_policy[IFLA_XDP_MAX + 1] = {
1550 [IFLA_XDP_FD] = { .type = NLA_S32 },
1551 [IFLA_XDP_ATTACHED] = { .type = NLA_U8 },
1552 [IFLA_XDP_FLAGS] = { .type = NLA_U32 },
1553 [IFLA_XDP_PROG_ID] = { .type = NLA_U32 },
1554 };
1555
1556 static const struct rtnl_link_ops *linkinfo_to_kind_ops(const struct nlattr *nla)
1557 {
1558 const struct rtnl_link_ops *ops = NULL;
1559 struct nlattr *linfo[IFLA_INFO_MAX + 1];
1560
1561 if (nla_parse_nested(linfo, IFLA_INFO_MAX, nla,
1562 ifla_info_policy, NULL) < 0)
1563 return NULL;
1564
1565 if (linfo[IFLA_INFO_KIND]) {
1566 char kind[MODULE_NAME_LEN];
1567
1568 nla_strlcpy(kind, linfo[IFLA_INFO_KIND], sizeof(kind));
1569 ops = rtnl_link_ops_get(kind);
1570 }
1571
1572 return ops;
1573 }
1574
1575 static bool link_master_filtered(struct net_device *dev, int master_idx)
1576 {
1577 struct net_device *master;
1578
1579 if (!master_idx)
1580 return false;
1581
1582 master = netdev_master_upper_dev_get(dev);
1583 if (!master || master->ifindex != master_idx)
1584 return true;
1585
1586 return false;
1587 }
1588
1589 static bool link_kind_filtered(const struct net_device *dev,
1590 const struct rtnl_link_ops *kind_ops)
1591 {
1592 if (kind_ops && dev->rtnl_link_ops != kind_ops)
1593 return true;
1594
1595 return false;
1596 }
1597
1598 static bool link_dump_filtered(struct net_device *dev,
1599 int master_idx,
1600 const struct rtnl_link_ops *kind_ops)
1601 {
1602 if (link_master_filtered(dev, master_idx) ||
1603 link_kind_filtered(dev, kind_ops))
1604 return true;
1605
1606 return false;
1607 }
1608
1609 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1610 {
1611 struct net *net = sock_net(skb->sk);
1612 int h, s_h;
1613 int idx = 0, s_idx;
1614 struct net_device *dev;
1615 struct hlist_head *head;
1616 struct nlattr *tb[IFLA_MAX+1];
1617 u32 ext_filter_mask = 0;
1618 const struct rtnl_link_ops *kind_ops = NULL;
1619 unsigned int flags = NLM_F_MULTI;
1620 int master_idx = 0;
1621 int err;
1622 int hdrlen;
1623
1624 s_h = cb->args[0];
1625 s_idx = cb->args[1];
1626
1627 /* A hack to preserve kernel<->userspace interface.
1628 * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
1629 * However, before Linux v3.9 the code here assumed rtgenmsg and that's
1630 * what iproute2 < v3.9.0 used.
1631 * We can detect the old iproute2. Even including the IFLA_EXT_MASK
1632 * attribute, its netlink message is shorter than struct ifinfomsg.
1633 */
1634 hdrlen = nlmsg_len(cb->nlh) < sizeof(struct ifinfomsg) ?
1635 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
1636
1637 if (nlmsg_parse(cb->nlh, hdrlen, tb, IFLA_MAX,
1638 ifla_policy, NULL) >= 0) {
1639 if (tb[IFLA_EXT_MASK])
1640 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1641
1642 if (tb[IFLA_MASTER])
1643 master_idx = nla_get_u32(tb[IFLA_MASTER]);
1644
1645 if (tb[IFLA_LINKINFO])
1646 kind_ops = linkinfo_to_kind_ops(tb[IFLA_LINKINFO]);
1647
1648 if (master_idx || kind_ops)
1649 flags |= NLM_F_DUMP_FILTERED;
1650 }
1651
1652 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1653 idx = 0;
1654 head = &net->dev_index_head[h];
1655 hlist_for_each_entry(dev, head, index_hlist) {
1656 if (link_dump_filtered(dev, master_idx, kind_ops))
1657 goto cont;
1658 if (idx < s_idx)
1659 goto cont;
1660 err = rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
1661 NETLINK_CB(cb->skb).portid,
1662 cb->nlh->nlmsg_seq, 0,
1663 flags,
1664 ext_filter_mask, 0);
1665
1666 if (err < 0) {
1667 if (likely(skb->len))
1668 goto out;
1669
1670 goto out_err;
1671 }
1672 cont:
1673 idx++;
1674 }
1675 }
1676 out:
1677 err = skb->len;
1678 out_err:
1679 cb->args[1] = idx;
1680 cb->args[0] = h;
1681 cb->seq = net->dev_base_seq;
1682 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1683
1684 return err;
1685 }
1686
1687 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len,
1688 struct netlink_ext_ack *exterr)
1689 {
1690 return nla_parse(tb, IFLA_MAX, head, len, ifla_policy, exterr);
1691 }
1692 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
1693
1694 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1695 {
1696 struct net *net;
1697 /* Examine the link attributes and figure out which
1698 * network namespace we are talking about.
1699 */
1700 if (tb[IFLA_NET_NS_PID])
1701 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1702 else if (tb[IFLA_NET_NS_FD])
1703 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1704 else
1705 net = get_net(src_net);
1706 return net;
1707 }
1708 EXPORT_SYMBOL(rtnl_link_get_net);
1709
1710 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1711 {
1712 if (dev) {
1713 if (tb[IFLA_ADDRESS] &&
1714 nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1715 return -EINVAL;
1716
1717 if (tb[IFLA_BROADCAST] &&
1718 nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1719 return -EINVAL;
1720 }
1721
1722 if (tb[IFLA_AF_SPEC]) {
1723 struct nlattr *af;
1724 int rem, err;
1725
1726 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1727 const struct rtnl_af_ops *af_ops;
1728
1729 if (!(af_ops = rtnl_af_lookup(nla_type(af))))
1730 return -EAFNOSUPPORT;
1731
1732 if (!af_ops->set_link_af)
1733 return -EOPNOTSUPP;
1734
1735 if (af_ops->validate_link_af) {
1736 err = af_ops->validate_link_af(dev, af);
1737 if (err < 0)
1738 return err;
1739 }
1740 }
1741 }
1742
1743 return 0;
1744 }
1745
1746 static int handle_infiniband_guid(struct net_device *dev, struct ifla_vf_guid *ivt,
1747 int guid_type)
1748 {
1749 const struct net_device_ops *ops = dev->netdev_ops;
1750
1751 return ops->ndo_set_vf_guid(dev, ivt->vf, ivt->guid, guid_type);
1752 }
1753
1754 static int handle_vf_guid(struct net_device *dev, struct ifla_vf_guid *ivt, int guid_type)
1755 {
1756 if (dev->type != ARPHRD_INFINIBAND)
1757 return -EOPNOTSUPP;
1758
1759 return handle_infiniband_guid(dev, ivt, guid_type);
1760 }
1761
1762 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb)
1763 {
1764 const struct net_device_ops *ops = dev->netdev_ops;
1765 int err = -EINVAL;
1766
1767 if (tb[IFLA_VF_MAC]) {
1768 struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]);
1769
1770 err = -EOPNOTSUPP;
1771 if (ops->ndo_set_vf_mac)
1772 err = ops->ndo_set_vf_mac(dev, ivm->vf,
1773 ivm->mac);
1774 if (err < 0)
1775 return err;
1776 }
1777
1778 if (tb[IFLA_VF_VLAN]) {
1779 struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]);
1780
1781 err = -EOPNOTSUPP;
1782 if (ops->ndo_set_vf_vlan)
1783 err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan,
1784 ivv->qos,
1785 htons(ETH_P_8021Q));
1786 if (err < 0)
1787 return err;
1788 }
1789
1790 if (tb[IFLA_VF_VLAN_LIST]) {
1791 struct ifla_vf_vlan_info *ivvl[MAX_VLAN_LIST_LEN];
1792 struct nlattr *attr;
1793 int rem, len = 0;
1794
1795 err = -EOPNOTSUPP;
1796 if (!ops->ndo_set_vf_vlan)
1797 return err;
1798
1799 nla_for_each_nested(attr, tb[IFLA_VF_VLAN_LIST], rem) {
1800 if (nla_type(attr) != IFLA_VF_VLAN_INFO ||
1801 nla_len(attr) < NLA_HDRLEN) {
1802 return -EINVAL;
1803 }
1804 if (len >= MAX_VLAN_LIST_LEN)
1805 return -EOPNOTSUPP;
1806 ivvl[len] = nla_data(attr);
1807
1808 len++;
1809 }
1810 if (len == 0)
1811 return -EINVAL;
1812
1813 err = ops->ndo_set_vf_vlan(dev, ivvl[0]->vf, ivvl[0]->vlan,
1814 ivvl[0]->qos, ivvl[0]->vlan_proto);
1815 if (err < 0)
1816 return err;
1817 }
1818
1819 if (tb[IFLA_VF_TX_RATE]) {
1820 struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]);
1821 struct ifla_vf_info ivf;
1822
1823 err = -EOPNOTSUPP;
1824 if (ops->ndo_get_vf_config)
1825 err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf);
1826 if (err < 0)
1827 return err;
1828
1829 err = -EOPNOTSUPP;
1830 if (ops->ndo_set_vf_rate)
1831 err = ops->ndo_set_vf_rate(dev, ivt->vf,
1832 ivf.min_tx_rate,
1833 ivt->rate);
1834 if (err < 0)
1835 return err;
1836 }
1837
1838 if (tb[IFLA_VF_RATE]) {
1839 struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]);
1840
1841 err = -EOPNOTSUPP;
1842 if (ops->ndo_set_vf_rate)
1843 err = ops->ndo_set_vf_rate(dev, ivt->vf,
1844 ivt->min_tx_rate,
1845 ivt->max_tx_rate);
1846 if (err < 0)
1847 return err;
1848 }
1849
1850 if (tb[IFLA_VF_SPOOFCHK]) {
1851 struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]);
1852
1853 err = -EOPNOTSUPP;
1854 if (ops->ndo_set_vf_spoofchk)
1855 err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
1856 ivs->setting);
1857 if (err < 0)
1858 return err;
1859 }
1860
1861 if (tb[IFLA_VF_LINK_STATE]) {
1862 struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]);
1863
1864 err = -EOPNOTSUPP;
1865 if (ops->ndo_set_vf_link_state)
1866 err = ops->ndo_set_vf_link_state(dev, ivl->vf,
1867 ivl->link_state);
1868 if (err < 0)
1869 return err;
1870 }
1871
1872 if (tb[IFLA_VF_RSS_QUERY_EN]) {
1873 struct ifla_vf_rss_query_en *ivrssq_en;
1874
1875 err = -EOPNOTSUPP;
1876 ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]);
1877 if (ops->ndo_set_vf_rss_query_en)
1878 err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf,
1879 ivrssq_en->setting);
1880 if (err < 0)
1881 return err;
1882 }
1883
1884 if (tb[IFLA_VF_TRUST]) {
1885 struct ifla_vf_trust *ivt = nla_data(tb[IFLA_VF_TRUST]);
1886
1887 err = -EOPNOTSUPP;
1888 if (ops->ndo_set_vf_trust)
1889 err = ops->ndo_set_vf_trust(dev, ivt->vf, ivt->setting);
1890 if (err < 0)
1891 return err;
1892 }
1893
1894 if (tb[IFLA_VF_IB_NODE_GUID]) {
1895 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_NODE_GUID]);
1896
1897 if (!ops->ndo_set_vf_guid)
1898 return -EOPNOTSUPP;
1899
1900 return handle_vf_guid(dev, ivt, IFLA_VF_IB_NODE_GUID);
1901 }
1902
1903 if (tb[IFLA_VF_IB_PORT_GUID]) {
1904 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_PORT_GUID]);
1905
1906 if (!ops->ndo_set_vf_guid)
1907 return -EOPNOTSUPP;
1908
1909 return handle_vf_guid(dev, ivt, IFLA_VF_IB_PORT_GUID);
1910 }
1911
1912 return err;
1913 }
1914
1915 static int do_set_master(struct net_device *dev, int ifindex)
1916 {
1917 struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
1918 const struct net_device_ops *ops;
1919 int err;
1920
1921 if (upper_dev) {
1922 if (upper_dev->ifindex == ifindex)
1923 return 0;
1924 ops = upper_dev->netdev_ops;
1925 if (ops->ndo_del_slave) {
1926 err = ops->ndo_del_slave(upper_dev, dev);
1927 if (err)
1928 return err;
1929 } else {
1930 return -EOPNOTSUPP;
1931 }
1932 }
1933
1934 if (ifindex) {
1935 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
1936 if (!upper_dev)
1937 return -EINVAL;
1938 ops = upper_dev->netdev_ops;
1939 if (ops->ndo_add_slave) {
1940 err = ops->ndo_add_slave(upper_dev, dev);
1941 if (err)
1942 return err;
1943 } else {
1944 return -EOPNOTSUPP;
1945 }
1946 }
1947 return 0;
1948 }
1949
1950 #define DO_SETLINK_MODIFIED 0x01
1951 /* notify flag means notify + modified. */
1952 #define DO_SETLINK_NOTIFY 0x03
1953 static int do_setlink(const struct sk_buff *skb,
1954 struct net_device *dev, struct ifinfomsg *ifm,
1955 struct netlink_ext_ack *extack,
1956 struct nlattr **tb, char *ifname, int status)
1957 {
1958 const struct net_device_ops *ops = dev->netdev_ops;
1959 int err;
1960
1961 err = validate_linkmsg(dev, tb);
1962 if (err < 0)
1963 return err;
1964
1965 if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]) {
1966 struct net *net = rtnl_link_get_net(dev_net(dev), tb);
1967 if (IS_ERR(net)) {
1968 err = PTR_ERR(net);
1969 goto errout;
1970 }
1971 if (!netlink_ns_capable(skb, net->user_ns, CAP_NET_ADMIN)) {
1972 put_net(net);
1973 err = -EPERM;
1974 goto errout;
1975 }
1976 err = dev_change_net_namespace(dev, net, ifname);
1977 put_net(net);
1978 if (err)
1979 goto errout;
1980 status |= DO_SETLINK_MODIFIED;
1981 }
1982
1983 if (tb[IFLA_MAP]) {
1984 struct rtnl_link_ifmap *u_map;
1985 struct ifmap k_map;
1986
1987 if (!ops->ndo_set_config) {
1988 err = -EOPNOTSUPP;
1989 goto errout;
1990 }
1991
1992 if (!netif_device_present(dev)) {
1993 err = -ENODEV;
1994 goto errout;
1995 }
1996
1997 u_map = nla_data(tb[IFLA_MAP]);
1998 k_map.mem_start = (unsigned long) u_map->mem_start;
1999 k_map.mem_end = (unsigned long) u_map->mem_end;
2000 k_map.base_addr = (unsigned short) u_map->base_addr;
2001 k_map.irq = (unsigned char) u_map->irq;
2002 k_map.dma = (unsigned char) u_map->dma;
2003 k_map.port = (unsigned char) u_map->port;
2004
2005 err = ops->ndo_set_config(dev, &k_map);
2006 if (err < 0)
2007 goto errout;
2008
2009 status |= DO_SETLINK_NOTIFY;
2010 }
2011
2012 if (tb[IFLA_ADDRESS]) {
2013 struct sockaddr *sa;
2014 int len;
2015
2016 len = sizeof(sa_family_t) + max_t(size_t, dev->addr_len,
2017 sizeof(*sa));
2018 sa = kmalloc(len, GFP_KERNEL);
2019 if (!sa) {
2020 err = -ENOMEM;
2021 goto errout;
2022 }
2023 sa->sa_family = dev->type;
2024 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
2025 dev->addr_len);
2026 err = dev_set_mac_address(dev, sa);
2027 kfree(sa);
2028 if (err)
2029 goto errout;
2030 status |= DO_SETLINK_MODIFIED;
2031 }
2032
2033 if (tb[IFLA_MTU]) {
2034 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
2035 if (err < 0)
2036 goto errout;
2037 status |= DO_SETLINK_MODIFIED;
2038 }
2039
2040 if (tb[IFLA_GROUP]) {
2041 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2042 status |= DO_SETLINK_NOTIFY;
2043 }
2044
2045 /*
2046 * Interface selected by interface index but interface
2047 * name provided implies that a name change has been
2048 * requested.
2049 */
2050 if (ifm->ifi_index > 0 && ifname[0]) {
2051 err = dev_change_name(dev, ifname);
2052 if (err < 0)
2053 goto errout;
2054 status |= DO_SETLINK_MODIFIED;
2055 }
2056
2057 if (tb[IFLA_IFALIAS]) {
2058 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
2059 nla_len(tb[IFLA_IFALIAS]));
2060 if (err < 0)
2061 goto errout;
2062 status |= DO_SETLINK_NOTIFY;
2063 }
2064
2065 if (tb[IFLA_BROADCAST]) {
2066 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
2067 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
2068 }
2069
2070 if (ifm->ifi_flags || ifm->ifi_change) {
2071 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
2072 if (err < 0)
2073 goto errout;
2074 }
2075
2076 if (tb[IFLA_MASTER]) {
2077 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
2078 if (err)
2079 goto errout;
2080 status |= DO_SETLINK_MODIFIED;
2081 }
2082
2083 if (tb[IFLA_CARRIER]) {
2084 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
2085 if (err)
2086 goto errout;
2087 status |= DO_SETLINK_MODIFIED;
2088 }
2089
2090 if (tb[IFLA_TXQLEN]) {
2091 unsigned int value = nla_get_u32(tb[IFLA_TXQLEN]);
2092 unsigned int orig_len = dev->tx_queue_len;
2093
2094 if (dev->tx_queue_len ^ value) {
2095 dev->tx_queue_len = value;
2096 err = call_netdevice_notifiers(
2097 NETDEV_CHANGE_TX_QUEUE_LEN, dev);
2098 err = notifier_to_errno(err);
2099 if (err) {
2100 dev->tx_queue_len = orig_len;
2101 goto errout;
2102 }
2103 status |= DO_SETLINK_MODIFIED;
2104 }
2105 }
2106
2107 if (tb[IFLA_OPERSTATE])
2108 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2109
2110 if (tb[IFLA_LINKMODE]) {
2111 unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
2112
2113 write_lock_bh(&dev_base_lock);
2114 if (dev->link_mode ^ value)
2115 status |= DO_SETLINK_NOTIFY;
2116 dev->link_mode = value;
2117 write_unlock_bh(&dev_base_lock);
2118 }
2119
2120 if (tb[IFLA_VFINFO_LIST]) {
2121 struct nlattr *vfinfo[IFLA_VF_MAX + 1];
2122 struct nlattr *attr;
2123 int rem;
2124
2125 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
2126 if (nla_type(attr) != IFLA_VF_INFO ||
2127 nla_len(attr) < NLA_HDRLEN) {
2128 err = -EINVAL;
2129 goto errout;
2130 }
2131 err = nla_parse_nested(vfinfo, IFLA_VF_MAX, attr,
2132 ifla_vf_policy, NULL);
2133 if (err < 0)
2134 goto errout;
2135 err = do_setvfinfo(dev, vfinfo);
2136 if (err < 0)
2137 goto errout;
2138 status |= DO_SETLINK_NOTIFY;
2139 }
2140 }
2141 err = 0;
2142
2143 if (tb[IFLA_VF_PORTS]) {
2144 struct nlattr *port[IFLA_PORT_MAX+1];
2145 struct nlattr *attr;
2146 int vf;
2147 int rem;
2148
2149 err = -EOPNOTSUPP;
2150 if (!ops->ndo_set_vf_port)
2151 goto errout;
2152
2153 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
2154 if (nla_type(attr) != IFLA_VF_PORT ||
2155 nla_len(attr) < NLA_HDRLEN) {
2156 err = -EINVAL;
2157 goto errout;
2158 }
2159 err = nla_parse_nested(port, IFLA_PORT_MAX, attr,
2160 ifla_port_policy, NULL);
2161 if (err < 0)
2162 goto errout;
2163 if (!port[IFLA_PORT_VF]) {
2164 err = -EOPNOTSUPP;
2165 goto errout;
2166 }
2167 vf = nla_get_u32(port[IFLA_PORT_VF]);
2168 err = ops->ndo_set_vf_port(dev, vf, port);
2169 if (err < 0)
2170 goto errout;
2171 status |= DO_SETLINK_NOTIFY;
2172 }
2173 }
2174 err = 0;
2175
2176 if (tb[IFLA_PORT_SELF]) {
2177 struct nlattr *port[IFLA_PORT_MAX+1];
2178
2179 err = nla_parse_nested(port, IFLA_PORT_MAX,
2180 tb[IFLA_PORT_SELF], ifla_port_policy,
2181 NULL);
2182 if (err < 0)
2183 goto errout;
2184
2185 err = -EOPNOTSUPP;
2186 if (ops->ndo_set_vf_port)
2187 err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
2188 if (err < 0)
2189 goto errout;
2190 status |= DO_SETLINK_NOTIFY;
2191 }
2192
2193 if (tb[IFLA_AF_SPEC]) {
2194 struct nlattr *af;
2195 int rem;
2196
2197 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2198 const struct rtnl_af_ops *af_ops;
2199
2200 if (!(af_ops = rtnl_af_lookup(nla_type(af))))
2201 BUG();
2202
2203 err = af_ops->set_link_af(dev, af);
2204 if (err < 0)
2205 goto errout;
2206
2207 status |= DO_SETLINK_NOTIFY;
2208 }
2209 }
2210 err = 0;
2211
2212 if (tb[IFLA_PROTO_DOWN]) {
2213 err = dev_change_proto_down(dev,
2214 nla_get_u8(tb[IFLA_PROTO_DOWN]));
2215 if (err)
2216 goto errout;
2217 status |= DO_SETLINK_NOTIFY;
2218 }
2219
2220 if (tb[IFLA_XDP]) {
2221 struct nlattr *xdp[IFLA_XDP_MAX + 1];
2222 u32 xdp_flags = 0;
2223
2224 err = nla_parse_nested(xdp, IFLA_XDP_MAX, tb[IFLA_XDP],
2225 ifla_xdp_policy, NULL);
2226 if (err < 0)
2227 goto errout;
2228
2229 if (xdp[IFLA_XDP_ATTACHED] || xdp[IFLA_XDP_PROG_ID]) {
2230 err = -EINVAL;
2231 goto errout;
2232 }
2233
2234 if (xdp[IFLA_XDP_FLAGS]) {
2235 xdp_flags = nla_get_u32(xdp[IFLA_XDP_FLAGS]);
2236 if (xdp_flags & ~XDP_FLAGS_MASK) {
2237 err = -EINVAL;
2238 goto errout;
2239 }
2240 if (hweight32(xdp_flags & XDP_FLAGS_MODES) > 1) {
2241 err = -EINVAL;
2242 goto errout;
2243 }
2244 }
2245
2246 if (xdp[IFLA_XDP_FD]) {
2247 err = dev_change_xdp_fd(dev, extack,
2248 nla_get_s32(xdp[IFLA_XDP_FD]),
2249 xdp_flags);
2250 if (err)
2251 goto errout;
2252 status |= DO_SETLINK_NOTIFY;
2253 }
2254 }
2255
2256 errout:
2257 if (status & DO_SETLINK_MODIFIED) {
2258 if ((status & DO_SETLINK_NOTIFY) == DO_SETLINK_NOTIFY)
2259 netdev_state_change(dev);
2260
2261 if (err < 0)
2262 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",
2263 dev->name);
2264 }
2265
2266 return err;
2267 }
2268
2269 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
2270 struct netlink_ext_ack *extack)
2271 {
2272 struct net *net = sock_net(skb->sk);
2273 struct ifinfomsg *ifm;
2274 struct net_device *dev;
2275 int err;
2276 struct nlattr *tb[IFLA_MAX+1];
2277 char ifname[IFNAMSIZ];
2278
2279 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy,
2280 extack);
2281 if (err < 0)
2282 goto errout;
2283
2284 if (tb[IFLA_IFNAME])
2285 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2286 else
2287 ifname[0] = '\0';
2288
2289 err = -EINVAL;
2290 ifm = nlmsg_data(nlh);
2291 if (ifm->ifi_index > 0)
2292 dev = __dev_get_by_index(net, ifm->ifi_index);
2293 else if (tb[IFLA_IFNAME])
2294 dev = __dev_get_by_name(net, ifname);
2295 else
2296 goto errout;
2297
2298 if (dev == NULL) {
2299 err = -ENODEV;
2300 goto errout;
2301 }
2302
2303 err = do_setlink(skb, dev, ifm, extack, tb, ifname, 0);
2304 errout:
2305 return err;
2306 }
2307
2308 static int rtnl_group_dellink(const struct net *net, int group)
2309 {
2310 struct net_device *dev, *aux;
2311 LIST_HEAD(list_kill);
2312 bool found = false;
2313
2314 if (!group)
2315 return -EPERM;
2316
2317 for_each_netdev(net, dev) {
2318 if (dev->group == group) {
2319 const struct rtnl_link_ops *ops;
2320
2321 found = true;
2322 ops = dev->rtnl_link_ops;
2323 if (!ops || !ops->dellink)
2324 return -EOPNOTSUPP;
2325 }
2326 }
2327
2328 if (!found)
2329 return -ENODEV;
2330
2331 for_each_netdev_safe(net, dev, aux) {
2332 if (dev->group == group) {
2333 const struct rtnl_link_ops *ops;
2334
2335 ops = dev->rtnl_link_ops;
2336 ops->dellink(dev, &list_kill);
2337 }
2338 }
2339 unregister_netdevice_many(&list_kill);
2340
2341 return 0;
2342 }
2343
2344 int rtnl_delete_link(struct net_device *dev)
2345 {
2346 const struct rtnl_link_ops *ops;
2347 LIST_HEAD(list_kill);
2348
2349 ops = dev->rtnl_link_ops;
2350 if (!ops || !ops->dellink)
2351 return -EOPNOTSUPP;
2352
2353 ops->dellink(dev, &list_kill);
2354 unregister_netdevice_many(&list_kill);
2355
2356 return 0;
2357 }
2358 EXPORT_SYMBOL_GPL(rtnl_delete_link);
2359
2360 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
2361 struct netlink_ext_ack *extack)
2362 {
2363 struct net *net = sock_net(skb->sk);
2364 struct net_device *dev;
2365 struct ifinfomsg *ifm;
2366 char ifname[IFNAMSIZ];
2367 struct nlattr *tb[IFLA_MAX+1];
2368 int err;
2369
2370 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
2371 if (err < 0)
2372 return err;
2373
2374 if (tb[IFLA_IFNAME])
2375 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2376
2377 ifm = nlmsg_data(nlh);
2378 if (ifm->ifi_index > 0)
2379 dev = __dev_get_by_index(net, ifm->ifi_index);
2380 else if (tb[IFLA_IFNAME])
2381 dev = __dev_get_by_name(net, ifname);
2382 else if (tb[IFLA_GROUP])
2383 return rtnl_group_dellink(net, nla_get_u32(tb[IFLA_GROUP]));
2384 else
2385 return -EINVAL;
2386
2387 if (!dev)
2388 return -ENODEV;
2389
2390 return rtnl_delete_link(dev);
2391 }
2392
2393 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
2394 {
2395 unsigned int old_flags;
2396 int err;
2397
2398 old_flags = dev->flags;
2399 if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
2400 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
2401 if (err < 0)
2402 return err;
2403 }
2404
2405 if (dev->rtnl_link_state == RTNL_LINK_INITIALIZED) {
2406 __dev_notify_flags(dev, old_flags, 0U);
2407 } else {
2408 dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
2409 __dev_notify_flags(dev, old_flags, ~0U);
2410 }
2411 return 0;
2412 }
2413 EXPORT_SYMBOL(rtnl_configure_link);
2414
2415 struct net_device *rtnl_create_link(struct net *net,
2416 const char *ifname, unsigned char name_assign_type,
2417 const struct rtnl_link_ops *ops, struct nlattr *tb[])
2418 {
2419 struct net_device *dev;
2420 unsigned int num_tx_queues = 1;
2421 unsigned int num_rx_queues = 1;
2422
2423 if (tb[IFLA_NUM_TX_QUEUES])
2424 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
2425 else if (ops->get_num_tx_queues)
2426 num_tx_queues = ops->get_num_tx_queues();
2427
2428 if (tb[IFLA_NUM_RX_QUEUES])
2429 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
2430 else if (ops->get_num_rx_queues)
2431 num_rx_queues = ops->get_num_rx_queues();
2432
2433 if (num_tx_queues < 1 || num_tx_queues > 4096)
2434 return ERR_PTR(-EINVAL);
2435
2436 if (num_rx_queues < 1 || num_rx_queues > 4096)
2437 return ERR_PTR(-EINVAL);
2438
2439 dev = alloc_netdev_mqs(ops->priv_size, ifname, name_assign_type,
2440 ops->setup, num_tx_queues, num_rx_queues);
2441 if (!dev)
2442 return ERR_PTR(-ENOMEM);
2443
2444 dev_net_set(dev, net);
2445 dev->rtnl_link_ops = ops;
2446 dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
2447
2448 if (tb[IFLA_MTU])
2449 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
2450 if (tb[IFLA_ADDRESS]) {
2451 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
2452 nla_len(tb[IFLA_ADDRESS]));
2453 dev->addr_assign_type = NET_ADDR_SET;
2454 }
2455 if (tb[IFLA_BROADCAST])
2456 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
2457 nla_len(tb[IFLA_BROADCAST]));
2458 if (tb[IFLA_TXQLEN])
2459 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
2460 if (tb[IFLA_OPERSTATE])
2461 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2462 if (tb[IFLA_LINKMODE])
2463 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
2464 if (tb[IFLA_GROUP])
2465 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2466
2467 return dev;
2468 }
2469 EXPORT_SYMBOL(rtnl_create_link);
2470
2471 static int rtnl_group_changelink(const struct sk_buff *skb,
2472 struct net *net, int group,
2473 struct ifinfomsg *ifm,
2474 struct netlink_ext_ack *extack,
2475 struct nlattr **tb)
2476 {
2477 struct net_device *dev, *aux;
2478 int err;
2479
2480 for_each_netdev_safe(net, dev, aux) {
2481 if (dev->group == group) {
2482 err = do_setlink(skb, dev, ifm, extack, tb, NULL, 0);
2483 if (err < 0)
2484 return err;
2485 }
2486 }
2487
2488 return 0;
2489 }
2490
2491 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
2492 struct netlink_ext_ack *extack)
2493 {
2494 struct net *net = sock_net(skb->sk);
2495 const struct rtnl_link_ops *ops;
2496 const struct rtnl_link_ops *m_ops = NULL;
2497 struct net_device *dev;
2498 struct net_device *master_dev = NULL;
2499 struct ifinfomsg *ifm;
2500 char kind[MODULE_NAME_LEN];
2501 char ifname[IFNAMSIZ];
2502 struct nlattr *tb[IFLA_MAX+1];
2503 struct nlattr *linkinfo[IFLA_INFO_MAX+1];
2504 unsigned char name_assign_type = NET_NAME_USER;
2505 int err;
2506
2507 #ifdef CONFIG_MODULES
2508 replay:
2509 #endif
2510 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
2511 if (err < 0)
2512 return err;
2513
2514 if (tb[IFLA_IFNAME])
2515 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2516 else
2517 ifname[0] = '\0';
2518
2519 ifm = nlmsg_data(nlh);
2520 if (ifm->ifi_index > 0)
2521 dev = __dev_get_by_index(net, ifm->ifi_index);
2522 else {
2523 if (ifname[0])
2524 dev = __dev_get_by_name(net, ifname);
2525 else
2526 dev = NULL;
2527 }
2528
2529 if (dev) {
2530 master_dev = netdev_master_upper_dev_get(dev);
2531 if (master_dev)
2532 m_ops = master_dev->rtnl_link_ops;
2533 }
2534
2535 err = validate_linkmsg(dev, tb);
2536 if (err < 0)
2537 return err;
2538
2539 if (tb[IFLA_LINKINFO]) {
2540 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
2541 tb[IFLA_LINKINFO], ifla_info_policy,
2542 NULL);
2543 if (err < 0)
2544 return err;
2545 } else
2546 memset(linkinfo, 0, sizeof(linkinfo));
2547
2548 if (linkinfo[IFLA_INFO_KIND]) {
2549 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
2550 ops = rtnl_link_ops_get(kind);
2551 } else {
2552 kind[0] = '\0';
2553 ops = NULL;
2554 }
2555
2556 if (1) {
2557 struct nlattr *attr[ops ? ops->maxtype + 1 : 1];
2558 struct nlattr *slave_attr[m_ops ? m_ops->slave_maxtype + 1 : 1];
2559 struct nlattr **data = NULL;
2560 struct nlattr **slave_data = NULL;
2561 struct net *dest_net, *link_net = NULL;
2562
2563 if (ops) {
2564 if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
2565 err = nla_parse_nested(attr, ops->maxtype,
2566 linkinfo[IFLA_INFO_DATA],
2567 ops->policy, NULL);
2568 if (err < 0)
2569 return err;
2570 data = attr;
2571 }
2572 if (ops->validate) {
2573 err = ops->validate(tb, data, extack);
2574 if (err < 0)
2575 return err;
2576 }
2577 }
2578
2579 if (m_ops) {
2580 if (m_ops->slave_maxtype &&
2581 linkinfo[IFLA_INFO_SLAVE_DATA]) {
2582 err = nla_parse_nested(slave_attr,
2583 m_ops->slave_maxtype,
2584 linkinfo[IFLA_INFO_SLAVE_DATA],
2585 m_ops->slave_policy,
2586 NULL);
2587 if (err < 0)
2588 return err;
2589 slave_data = slave_attr;
2590 }
2591 if (m_ops->slave_validate) {
2592 err = m_ops->slave_validate(tb, slave_data,
2593 extack);
2594 if (err < 0)
2595 return err;
2596 }
2597 }
2598
2599 if (dev) {
2600 int status = 0;
2601
2602 if (nlh->nlmsg_flags & NLM_F_EXCL)
2603 return -EEXIST;
2604 if (nlh->nlmsg_flags & NLM_F_REPLACE)
2605 return -EOPNOTSUPP;
2606
2607 if (linkinfo[IFLA_INFO_DATA]) {
2608 if (!ops || ops != dev->rtnl_link_ops ||
2609 !ops->changelink)
2610 return -EOPNOTSUPP;
2611
2612 err = ops->changelink(dev, tb, data, extack);
2613 if (err < 0)
2614 return err;
2615 status |= DO_SETLINK_NOTIFY;
2616 }
2617
2618 if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
2619 if (!m_ops || !m_ops->slave_changelink)
2620 return -EOPNOTSUPP;
2621
2622 err = m_ops->slave_changelink(master_dev, dev,
2623 tb, slave_data,
2624 extack);
2625 if (err < 0)
2626 return err;
2627 status |= DO_SETLINK_NOTIFY;
2628 }
2629
2630 return do_setlink(skb, dev, ifm, extack, tb, ifname,
2631 status);
2632 }
2633
2634 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
2635 if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
2636 return rtnl_group_changelink(skb, net,
2637 nla_get_u32(tb[IFLA_GROUP]),
2638 ifm, extack, tb);
2639 return -ENODEV;
2640 }
2641
2642 if (tb[IFLA_MAP] || tb[IFLA_PROTINFO])
2643 return -EOPNOTSUPP;
2644
2645 if (!ops) {
2646 #ifdef CONFIG_MODULES
2647 if (kind[0]) {
2648 __rtnl_unlock();
2649 request_module("rtnl-link-%s", kind);
2650 rtnl_lock();
2651 ops = rtnl_link_ops_get(kind);
2652 if (ops)
2653 goto replay;
2654 }
2655 #endif
2656 return -EOPNOTSUPP;
2657 }
2658
2659 if (!ops->setup)
2660 return -EOPNOTSUPP;
2661
2662 if (!ifname[0]) {
2663 snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
2664 name_assign_type = NET_NAME_ENUM;
2665 }
2666
2667 dest_net = rtnl_link_get_net(net, tb);
2668 if (IS_ERR(dest_net))
2669 return PTR_ERR(dest_net);
2670
2671 err = -EPERM;
2672 if (!netlink_ns_capable(skb, dest_net->user_ns, CAP_NET_ADMIN))
2673 goto out;
2674
2675 if (tb[IFLA_LINK_NETNSID]) {
2676 int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
2677
2678 link_net = get_net_ns_by_id(dest_net, id);
2679 if (!link_net) {
2680 err = -EINVAL;
2681 goto out;
2682 }
2683 err = -EPERM;
2684 if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
2685 goto out;
2686 }
2687
2688 dev = rtnl_create_link(link_net ? : dest_net, ifname,
2689 name_assign_type, ops, tb);
2690 if (IS_ERR(dev)) {
2691 err = PTR_ERR(dev);
2692 goto out;
2693 }
2694
2695 dev->ifindex = ifm->ifi_index;
2696
2697 if (ops->newlink) {
2698 err = ops->newlink(link_net ? : net, dev, tb, data,
2699 extack);
2700 /* Drivers should call free_netdev() in ->destructor
2701 * and unregister it on failure after registration
2702 * so that device could be finally freed in rtnl_unlock.
2703 */
2704 if (err < 0) {
2705 /* If device is not registered at all, free it now */
2706 if (dev->reg_state == NETREG_UNINITIALIZED)
2707 free_netdev(dev);
2708 goto out;
2709 }
2710 } else {
2711 err = register_netdevice(dev);
2712 if (err < 0) {
2713 free_netdev(dev);
2714 goto out;
2715 }
2716 }
2717 err = rtnl_configure_link(dev, ifm);
2718 if (err < 0)
2719 goto out_unregister;
2720 if (link_net) {
2721 err = dev_change_net_namespace(dev, dest_net, ifname);
2722 if (err < 0)
2723 goto out_unregister;
2724 }
2725 if (tb[IFLA_MASTER]) {
2726 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]));
2727 if (err)
2728 goto out_unregister;
2729 }
2730 out:
2731 if (link_net)
2732 put_net(link_net);
2733 put_net(dest_net);
2734 return err;
2735 out_unregister:
2736 if (ops->newlink) {
2737 LIST_HEAD(list_kill);
2738
2739 ops->dellink(dev, &list_kill);
2740 unregister_netdevice_many(&list_kill);
2741 } else {
2742 unregister_netdevice(dev);
2743 }
2744 goto out;
2745 }
2746 }
2747
2748 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr *nlh,
2749 struct netlink_ext_ack *extack)
2750 {
2751 struct net *net = sock_net(skb->sk);
2752 struct ifinfomsg *ifm;
2753 char ifname[IFNAMSIZ];
2754 struct nlattr *tb[IFLA_MAX+1];
2755 struct net_device *dev = NULL;
2756 struct sk_buff *nskb;
2757 int err;
2758 u32 ext_filter_mask = 0;
2759
2760 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
2761 if (err < 0)
2762 return err;
2763
2764 if (tb[IFLA_IFNAME])
2765 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2766
2767 if (tb[IFLA_EXT_MASK])
2768 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2769
2770 ifm = nlmsg_data(nlh);
2771 if (ifm->ifi_index > 0)
2772 dev = __dev_get_by_index(net, ifm->ifi_index);
2773 else if (tb[IFLA_IFNAME])
2774 dev = __dev_get_by_name(net, ifname);
2775 else
2776 return -EINVAL;
2777
2778 if (dev == NULL)
2779 return -ENODEV;
2780
2781 nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
2782 if (nskb == NULL)
2783 return -ENOBUFS;
2784
2785 err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).portid,
2786 nlh->nlmsg_seq, 0, 0, ext_filter_mask, 0);
2787 if (err < 0) {
2788 /* -EMSGSIZE implies BUG in if_nlmsg_size */
2789 WARN_ON(err == -EMSGSIZE);
2790 kfree_skb(nskb);
2791 } else
2792 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
2793
2794 return err;
2795 }
2796
2797 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
2798 {
2799 struct net *net = sock_net(skb->sk);
2800 struct net_device *dev;
2801 struct nlattr *tb[IFLA_MAX+1];
2802 u32 ext_filter_mask = 0;
2803 u16 min_ifinfo_dump_size = 0;
2804 int hdrlen;
2805
2806 /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
2807 hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
2808 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
2809
2810 if (nlmsg_parse(nlh, hdrlen, tb, IFLA_MAX, ifla_policy, NULL) >= 0) {
2811 if (tb[IFLA_EXT_MASK])
2812 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
2813 }
2814
2815 if (!ext_filter_mask)
2816 return NLMSG_GOODSIZE;
2817 /*
2818 * traverse the list of net devices and compute the minimum
2819 * buffer size based upon the filter mask.
2820 */
2821 rcu_read_lock();
2822 for_each_netdev_rcu(net, dev) {
2823 min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
2824 if_nlmsg_size(dev,
2825 ext_filter_mask));
2826 }
2827 rcu_read_unlock();
2828
2829 return nlmsg_total_size(min_ifinfo_dump_size);
2830 }
2831
2832 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
2833 {
2834 int idx;
2835 int s_idx = cb->family;
2836
2837 if (s_idx == 0)
2838 s_idx = 1;
2839
2840 for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
2841 int type = cb->nlh->nlmsg_type-RTM_BASE;
2842 struct rtnl_link *handlers;
2843 rtnl_dumpit_func dumpit;
2844
2845 if (idx < s_idx || idx == PF_PACKET)
2846 continue;
2847
2848 handlers = rtnl_dereference(rtnl_msg_handlers[idx]);
2849 if (!handlers)
2850 continue;
2851
2852 dumpit = READ_ONCE(handlers[type].dumpit);
2853 if (!dumpit)
2854 continue;
2855
2856 if (idx > s_idx) {
2857 memset(&cb->args[0], 0, sizeof(cb->args));
2858 cb->prev_seq = 0;
2859 cb->seq = 0;
2860 }
2861 if (dumpit(skb, cb))
2862 break;
2863 }
2864 cb->family = idx;
2865
2866 return skb->len;
2867 }
2868
2869 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
2870 unsigned int change,
2871 u32 event, gfp_t flags)
2872 {
2873 struct net *net = dev_net(dev);
2874 struct sk_buff *skb;
2875 int err = -ENOBUFS;
2876 size_t if_info_size;
2877
2878 skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
2879 if (skb == NULL)
2880 goto errout;
2881
2882 err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0, 0, event);
2883 if (err < 0) {
2884 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
2885 WARN_ON(err == -EMSGSIZE);
2886 kfree_skb(skb);
2887 goto errout;
2888 }
2889 return skb;
2890 errout:
2891 if (err < 0)
2892 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
2893 return NULL;
2894 }
2895
2896 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags)
2897 {
2898 struct net *net = dev_net(dev);
2899
2900 rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
2901 }
2902
2903 static void rtmsg_ifinfo_event(int type, struct net_device *dev,
2904 unsigned int change, u32 event,
2905 gfp_t flags)
2906 {
2907 struct sk_buff *skb;
2908
2909 if (dev->reg_state != NETREG_REGISTERED)
2910 return;
2911
2912 skb = rtmsg_ifinfo_build_skb(type, dev, change, event, flags);
2913 if (skb)
2914 rtmsg_ifinfo_send(skb, dev, flags);
2915 }
2916
2917 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
2918 gfp_t flags)
2919 {
2920 rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags);
2921 }
2922 EXPORT_SYMBOL(rtmsg_ifinfo);
2923
2924 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
2925 struct net_device *dev,
2926 u8 *addr, u16 vid, u32 pid, u32 seq,
2927 int type, unsigned int flags,
2928 int nlflags, u16 ndm_state)
2929 {
2930 struct nlmsghdr *nlh;
2931 struct ndmsg *ndm;
2932
2933 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
2934 if (!nlh)
2935 return -EMSGSIZE;
2936
2937 ndm = nlmsg_data(nlh);
2938 ndm->ndm_family = AF_BRIDGE;
2939 ndm->ndm_pad1 = 0;
2940 ndm->ndm_pad2 = 0;
2941 ndm->ndm_flags = flags;
2942 ndm->ndm_type = 0;
2943 ndm->ndm_ifindex = dev->ifindex;
2944 ndm->ndm_state = ndm_state;
2945
2946 if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
2947 goto nla_put_failure;
2948 if (vid)
2949 if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
2950 goto nla_put_failure;
2951
2952 nlmsg_end(skb, nlh);
2953 return 0;
2954
2955 nla_put_failure:
2956 nlmsg_cancel(skb, nlh);
2957 return -EMSGSIZE;
2958 }
2959
2960 static inline size_t rtnl_fdb_nlmsg_size(void)
2961 {
2962 return NLMSG_ALIGN(sizeof(struct ndmsg)) +
2963 nla_total_size(ETH_ALEN) + /* NDA_LLADDR */
2964 nla_total_size(sizeof(u16)) + /* NDA_VLAN */
2965 0;
2966 }
2967
2968 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type,
2969 u16 ndm_state)
2970 {
2971 struct net *net = dev_net(dev);
2972 struct sk_buff *skb;
2973 int err = -ENOBUFS;
2974
2975 skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
2976 if (!skb)
2977 goto errout;
2978
2979 err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
2980 0, 0, type, NTF_SELF, 0, ndm_state);
2981 if (err < 0) {
2982 kfree_skb(skb);
2983 goto errout;
2984 }
2985
2986 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
2987 return;
2988 errout:
2989 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
2990 }
2991
2992 /**
2993 * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
2994 */
2995 int ndo_dflt_fdb_add(struct ndmsg *ndm,
2996 struct nlattr *tb[],
2997 struct net_device *dev,
2998 const unsigned char *addr, u16 vid,
2999 u16 flags)
3000 {
3001 int err = -EINVAL;
3002
3003 /* If aging addresses are supported device will need to
3004 * implement its own handler for this.
3005 */
3006 if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
3007 pr_info("%s: FDB only supports static addresses\n", dev->name);
3008 return err;
3009 }
3010
3011 if (vid) {
3012 pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
3013 return err;
3014 }
3015
3016 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
3017 err = dev_uc_add_excl(dev, addr);
3018 else if (is_multicast_ether_addr(addr))
3019 err = dev_mc_add_excl(dev, addr);
3020
3021 /* Only return duplicate errors if NLM_F_EXCL is set */
3022 if (err == -EEXIST && !(flags & NLM_F_EXCL))
3023 err = 0;
3024
3025 return err;
3026 }
3027 EXPORT_SYMBOL(ndo_dflt_fdb_add);
3028
3029 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid)
3030 {
3031 u16 vid = 0;
3032
3033 if (vlan_attr) {
3034 if (nla_len(vlan_attr) != sizeof(u16)) {
3035 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan\n");
3036 return -EINVAL;
3037 }
3038
3039 vid = nla_get_u16(vlan_attr);
3040
3041 if (!vid || vid >= VLAN_VID_MASK) {
3042 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid vlan id %d\n",
3043 vid);
3044 return -EINVAL;
3045 }
3046 }
3047 *p_vid = vid;
3048 return 0;
3049 }
3050
3051 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh,
3052 struct netlink_ext_ack *extack)
3053 {
3054 struct net *net = sock_net(skb->sk);
3055 struct ndmsg *ndm;
3056 struct nlattr *tb[NDA_MAX+1];
3057 struct net_device *dev;
3058 u8 *addr;
3059 u16 vid;
3060 int err;
3061
3062 err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL, extack);
3063 if (err < 0)
3064 return err;
3065
3066 ndm = nlmsg_data(nlh);
3067 if (ndm->ndm_ifindex == 0) {
3068 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid ifindex\n");
3069 return -EINVAL;
3070 }
3071
3072 dev = __dev_get_by_index(net, ndm->ndm_ifindex);
3073 if (dev == NULL) {
3074 pr_info("PF_BRIDGE: RTM_NEWNEIGH with unknown ifindex\n");
3075 return -ENODEV;
3076 }
3077
3078 if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
3079 pr_info("PF_BRIDGE: RTM_NEWNEIGH with invalid address\n");
3080 return -EINVAL;
3081 }
3082
3083 if (dev->type != ARPHRD_ETHER) {
3084 NL_SET_ERR_MSG(extack, "FDB add only supported for Ethernet devices");
3085 return -EINVAL;
3086 }
3087
3088 addr = nla_data(tb[NDA_LLADDR]);
3089
3090 err = fdb_vid_parse(tb[NDA_VLAN], &vid);
3091 if (err)
3092 return err;
3093
3094 err = -EOPNOTSUPP;
3095
3096 /* Support fdb on master device the net/bridge default case */
3097 if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
3098 (dev->priv_flags & IFF_BRIDGE_PORT)) {
3099 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3100 const struct net_device_ops *ops = br_dev->netdev_ops;
3101
3102 err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
3103 nlh->nlmsg_flags);
3104 if (err)
3105 goto out;
3106 else
3107 ndm->ndm_flags &= ~NTF_MASTER;
3108 }
3109
3110 /* Embedded bridge, macvlan, and any other device support */
3111 if ((ndm->ndm_flags & NTF_SELF)) {
3112 if (dev->netdev_ops->ndo_fdb_add)
3113 err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
3114 vid,
3115 nlh->nlmsg_flags);
3116 else
3117 err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
3118 nlh->nlmsg_flags);
3119
3120 if (!err) {
3121 rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH,
3122 ndm->ndm_state);
3123 ndm->ndm_flags &= ~NTF_SELF;
3124 }
3125 }
3126 out:
3127 return err;
3128 }
3129
3130 /**
3131 * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
3132 */
3133 int ndo_dflt_fdb_del(struct ndmsg *ndm,
3134 struct nlattr *tb[],
3135 struct net_device *dev,
3136 const unsigned char *addr, u16 vid)
3137 {
3138 int err = -EINVAL;
3139
3140 /* If aging addresses are supported device will need to
3141 * implement its own handler for this.
3142 */
3143 if (!(ndm->ndm_state & NUD_PERMANENT)) {
3144 pr_info("%s: FDB only supports static addresses\n", dev->name);
3145 return err;
3146 }
3147
3148 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
3149 err = dev_uc_del(dev, addr);
3150 else if (is_multicast_ether_addr(addr))
3151 err = dev_mc_del(dev, addr);
3152
3153 return err;
3154 }
3155 EXPORT_SYMBOL(ndo_dflt_fdb_del);
3156
3157 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh,
3158 struct netlink_ext_ack *extack)
3159 {
3160 struct net *net = sock_net(skb->sk);
3161 struct ndmsg *ndm;
3162 struct nlattr *tb[NDA_MAX+1];
3163 struct net_device *dev;
3164 int err = -EINVAL;
3165 __u8 *addr;
3166 u16 vid;
3167
3168 if (!netlink_capable(skb, CAP_NET_ADMIN))
3169 return -EPERM;
3170
3171 err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL, extack);
3172 if (err < 0)
3173 return err;
3174
3175 ndm = nlmsg_data(nlh);
3176 if (ndm->ndm_ifindex == 0) {
3177 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid ifindex\n");
3178 return -EINVAL;
3179 }
3180
3181 dev = __dev_get_by_index(net, ndm->ndm_ifindex);
3182 if (dev == NULL) {
3183 pr_info("PF_BRIDGE: RTM_DELNEIGH with unknown ifindex\n");
3184 return -ENODEV;
3185 }
3186
3187 if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
3188 pr_info("PF_BRIDGE: RTM_DELNEIGH with invalid address\n");
3189 return -EINVAL;
3190 }
3191
3192 if (dev->type != ARPHRD_ETHER) {
3193 NL_SET_ERR_MSG(extack, "FDB delete only supported for Ethernet devices");
3194 return -EINVAL;
3195 }
3196
3197 addr = nla_data(tb[NDA_LLADDR]);
3198
3199 err = fdb_vid_parse(tb[NDA_VLAN], &vid);
3200 if (err)
3201 return err;
3202
3203 err = -EOPNOTSUPP;
3204
3205 /* Support fdb on master device the net/bridge default case */
3206 if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
3207 (dev->priv_flags & IFF_BRIDGE_PORT)) {
3208 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3209 const struct net_device_ops *ops = br_dev->netdev_ops;
3210
3211 if (ops->ndo_fdb_del)
3212 err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid);
3213
3214 if (err)
3215 goto out;
3216 else
3217 ndm->ndm_flags &= ~NTF_MASTER;
3218 }
3219
3220 /* Embedded bridge, macvlan, and any other device support */
3221 if (ndm->ndm_flags & NTF_SELF) {
3222 if (dev->netdev_ops->ndo_fdb_del)
3223 err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr,
3224 vid);
3225 else
3226 err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
3227
3228 if (!err) {
3229 rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH,
3230 ndm->ndm_state);
3231 ndm->ndm_flags &= ~NTF_SELF;
3232 }
3233 }
3234 out:
3235 return err;
3236 }
3237
3238 static int nlmsg_populate_fdb(struct sk_buff *skb,
3239 struct netlink_callback *cb,
3240 struct net_device *dev,
3241 int *idx,
3242 struct netdev_hw_addr_list *list)
3243 {
3244 struct netdev_hw_addr *ha;
3245 int err;
3246 u32 portid, seq;
3247
3248 portid = NETLINK_CB(cb->skb).portid;
3249 seq = cb->nlh->nlmsg_seq;
3250
3251 list_for_each_entry(ha, &list->list, list) {
3252 if (*idx < cb->args[2])
3253 goto skip;
3254
3255 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
3256 portid, seq,
3257 RTM_NEWNEIGH, NTF_SELF,
3258 NLM_F_MULTI, NUD_PERMANENT);
3259 if (err < 0)
3260 return err;
3261 skip:
3262 *idx += 1;
3263 }
3264 return 0;
3265 }
3266
3267 /**
3268 * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
3269 * @nlh: netlink message header
3270 * @dev: netdevice
3271 *
3272 * Default netdevice operation to dump the existing unicast address list.
3273 * Returns number of addresses from list put in skb.
3274 */
3275 int ndo_dflt_fdb_dump(struct sk_buff *skb,
3276 struct netlink_callback *cb,
3277 struct net_device *dev,
3278 struct net_device *filter_dev,
3279 int *idx)
3280 {
3281 int err;
3282
3283 if (dev->type != ARPHRD_ETHER)
3284 return -EINVAL;
3285
3286 netif_addr_lock_bh(dev);
3287 err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->uc);
3288 if (err)
3289 goto out;
3290 err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->mc);
3291 out:
3292 netif_addr_unlock_bh(dev);
3293 return err;
3294 }
3295 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
3296
3297 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
3298 {
3299 struct net_device *dev;
3300 struct nlattr *tb[IFLA_MAX+1];
3301 struct net_device *br_dev = NULL;
3302 const struct net_device_ops *ops = NULL;
3303 const struct net_device_ops *cops = NULL;
3304 struct ifinfomsg *ifm = nlmsg_data(cb->nlh);
3305 struct net *net = sock_net(skb->sk);
3306 struct hlist_head *head;
3307 int brport_idx = 0;
3308 int br_idx = 0;
3309 int h, s_h;
3310 int idx = 0, s_idx;
3311 int err = 0;
3312 int fidx = 0;
3313
3314 /* A hack to preserve kernel<->userspace interface.
3315 * Before Linux v4.12 this code accepted ndmsg since iproute2 v3.3.0.
3316 * However, ndmsg is shorter than ifinfomsg thus nlmsg_parse() bails.
3317 * So, check for ndmsg with an optional u32 attribute (not used here).
3318 * Fortunately these sizes don't conflict with the size of ifinfomsg
3319 * with an optional attribute.
3320 */
3321 if (nlmsg_len(cb->nlh) != sizeof(struct ndmsg) &&
3322 (nlmsg_len(cb->nlh) != sizeof(struct ndmsg) +
3323 nla_attr_size(sizeof(u32)))) {
3324 err = nlmsg_parse(cb->nlh, sizeof(struct ifinfomsg), tb,
3325 IFLA_MAX, ifla_policy, NULL);
3326 if (err < 0) {
3327 return -EINVAL;
3328 } else if (err == 0) {
3329 if (tb[IFLA_MASTER])
3330 br_idx = nla_get_u32(tb[IFLA_MASTER]);
3331 }
3332
3333 brport_idx = ifm->ifi_index;
3334 }
3335
3336 if (br_idx) {
3337 br_dev = __dev_get_by_index(net, br_idx);
3338 if (!br_dev)
3339 return -ENODEV;
3340
3341 ops = br_dev->netdev_ops;
3342 }
3343
3344 s_h = cb->args[0];
3345 s_idx = cb->args[1];
3346
3347 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3348 idx = 0;
3349 head = &net->dev_index_head[h];
3350 hlist_for_each_entry(dev, head, index_hlist) {
3351
3352 if (brport_idx && (dev->ifindex != brport_idx))
3353 continue;
3354
3355 if (!br_idx) { /* user did not specify a specific bridge */
3356 if (dev->priv_flags & IFF_BRIDGE_PORT) {
3357 br_dev = netdev_master_upper_dev_get(dev);
3358 cops = br_dev->netdev_ops;
3359 }
3360 } else {
3361 if (dev != br_dev &&
3362 !(dev->priv_flags & IFF_BRIDGE_PORT))
3363 continue;
3364
3365 if (br_dev != netdev_master_upper_dev_get(dev) &&
3366 !(dev->priv_flags & IFF_EBRIDGE))
3367 continue;
3368 cops = ops;
3369 }
3370
3371 if (idx < s_idx)
3372 goto cont;
3373
3374 if (dev->priv_flags & IFF_BRIDGE_PORT) {
3375 if (cops && cops->ndo_fdb_dump) {
3376 err = cops->ndo_fdb_dump(skb, cb,
3377 br_dev, dev,
3378 &fidx);
3379 if (err == -EMSGSIZE)
3380 goto out;
3381 }
3382 }
3383
3384 if (dev->netdev_ops->ndo_fdb_dump)
3385 err = dev->netdev_ops->ndo_fdb_dump(skb, cb,
3386 dev, NULL,
3387 &fidx);
3388 else
3389 err = ndo_dflt_fdb_dump(skb, cb, dev, NULL,
3390 &fidx);
3391 if (err == -EMSGSIZE)
3392 goto out;
3393
3394 cops = NULL;
3395
3396 /* reset fdb offset to 0 for rest of the interfaces */
3397 cb->args[2] = 0;
3398 fidx = 0;
3399 cont:
3400 idx++;
3401 }
3402 }
3403
3404 out:
3405 cb->args[0] = h;
3406 cb->args[1] = idx;
3407 cb->args[2] = fidx;
3408
3409 return skb->len;
3410 }
3411
3412 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
3413 unsigned int attrnum, unsigned int flag)
3414 {
3415 if (mask & flag)
3416 return nla_put_u8(skb, attrnum, !!(flags & flag));
3417 return 0;
3418 }
3419
3420 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
3421 struct net_device *dev, u16 mode,
3422 u32 flags, u32 mask, int nlflags,
3423 u32 filter_mask,
3424 int (*vlan_fill)(struct sk_buff *skb,
3425 struct net_device *dev,
3426 u32 filter_mask))
3427 {
3428 struct nlmsghdr *nlh;
3429 struct ifinfomsg *ifm;
3430 struct nlattr *br_afspec;
3431 struct nlattr *protinfo;
3432 u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
3433 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3434 int err = 0;
3435
3436 nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
3437 if (nlh == NULL)
3438 return -EMSGSIZE;
3439
3440 ifm = nlmsg_data(nlh);
3441 ifm->ifi_family = AF_BRIDGE;
3442 ifm->__ifi_pad = 0;
3443 ifm->ifi_type = dev->type;
3444 ifm->ifi_index = dev->ifindex;
3445 ifm->ifi_flags = dev_get_flags(dev);
3446 ifm->ifi_change = 0;
3447
3448
3449 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
3450 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
3451 nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
3452 (br_dev &&
3453 nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
3454 (dev->addr_len &&
3455 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
3456 (dev->ifindex != dev_get_iflink(dev) &&
3457 nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
3458 goto nla_put_failure;
3459
3460 br_afspec = nla_nest_start(skb, IFLA_AF_SPEC);
3461 if (!br_afspec)
3462 goto nla_put_failure;
3463
3464 if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
3465 nla_nest_cancel(skb, br_afspec);
3466 goto nla_put_failure;
3467 }
3468
3469 if (mode != BRIDGE_MODE_UNDEF) {
3470 if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
3471 nla_nest_cancel(skb, br_afspec);
3472 goto nla_put_failure;
3473 }
3474 }
3475 if (vlan_fill) {
3476 err = vlan_fill(skb, dev, filter_mask);
3477 if (err) {
3478 nla_nest_cancel(skb, br_afspec);
3479 goto nla_put_failure;
3480 }
3481 }
3482 nla_nest_end(skb, br_afspec);
3483
3484 protinfo = nla_nest_start(skb, IFLA_PROTINFO | NLA_F_NESTED);
3485 if (!protinfo)
3486 goto nla_put_failure;
3487
3488 if (brport_nla_put_flag(skb, flags, mask,
3489 IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
3490 brport_nla_put_flag(skb, flags, mask,
3491 IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
3492 brport_nla_put_flag(skb, flags, mask,
3493 IFLA_BRPORT_FAST_LEAVE,
3494 BR_MULTICAST_FAST_LEAVE) ||
3495 brport_nla_put_flag(skb, flags, mask,
3496 IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
3497 brport_nla_put_flag(skb, flags, mask,
3498 IFLA_BRPORT_LEARNING, BR_LEARNING) ||
3499 brport_nla_put_flag(skb, flags, mask,
3500 IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
3501 brport_nla_put_flag(skb, flags, mask,
3502 IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
3503 brport_nla_put_flag(skb, flags, mask,
3504 IFLA_BRPORT_PROXYARP, BR_PROXYARP)) {
3505 nla_nest_cancel(skb, protinfo);
3506 goto nla_put_failure;
3507 }
3508
3509 nla_nest_end(skb, protinfo);
3510
3511 nlmsg_end(skb, nlh);
3512 return 0;
3513 nla_put_failure:
3514 nlmsg_cancel(skb, nlh);
3515 return err ? err : -EMSGSIZE;
3516 }
3517 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink);
3518
3519 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
3520 {
3521 struct net *net = sock_net(skb->sk);
3522 struct net_device *dev;
3523 int idx = 0;
3524 u32 portid = NETLINK_CB(cb->skb).portid;
3525 u32 seq = cb->nlh->nlmsg_seq;
3526 u32 filter_mask = 0;
3527 int err;
3528
3529 if (nlmsg_len(cb->nlh) > sizeof(struct ifinfomsg)) {
3530 struct nlattr *extfilt;
3531
3532 extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct ifinfomsg),
3533 IFLA_EXT_MASK);
3534 if (extfilt) {
3535 if (nla_len(extfilt) < sizeof(filter_mask))
3536 return -EINVAL;
3537
3538 filter_mask = nla_get_u32(extfilt);
3539 }
3540 }
3541
3542 rcu_read_lock();
3543 for_each_netdev_rcu(net, dev) {
3544 const struct net_device_ops *ops = dev->netdev_ops;
3545 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3546
3547 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
3548 if (idx >= cb->args[0]) {
3549 err = br_dev->netdev_ops->ndo_bridge_getlink(
3550 skb, portid, seq, dev,
3551 filter_mask, NLM_F_MULTI);
3552 if (err < 0 && err != -EOPNOTSUPP) {
3553 if (likely(skb->len))
3554 break;
3555
3556 goto out_err;
3557 }
3558 }
3559 idx++;
3560 }
3561
3562 if (ops->ndo_bridge_getlink) {
3563 if (idx >= cb->args[0]) {
3564 err = ops->ndo_bridge_getlink(skb, portid,
3565 seq, dev,
3566 filter_mask,
3567 NLM_F_MULTI);
3568 if (err < 0 && err != -EOPNOTSUPP) {
3569 if (likely(skb->len))
3570 break;
3571
3572 goto out_err;
3573 }
3574 }
3575 idx++;
3576 }
3577 }
3578 err = skb->len;
3579 out_err:
3580 rcu_read_unlock();
3581 cb->args[0] = idx;
3582
3583 return err;
3584 }
3585
3586 static inline size_t bridge_nlmsg_size(void)
3587 {
3588 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3589 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3590 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3591 + nla_total_size(sizeof(u32)) /* IFLA_MASTER */
3592 + nla_total_size(sizeof(u32)) /* IFLA_MTU */
3593 + nla_total_size(sizeof(u32)) /* IFLA_LINK */
3594 + nla_total_size(sizeof(u32)) /* IFLA_OPERSTATE */
3595 + nla_total_size(sizeof(u8)) /* IFLA_PROTINFO */
3596 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
3597 + nla_total_size(sizeof(u16)) /* IFLA_BRIDGE_FLAGS */
3598 + nla_total_size(sizeof(u16)); /* IFLA_BRIDGE_MODE */
3599 }
3600
3601 static int rtnl_bridge_notify(struct net_device *dev)
3602 {
3603 struct net *net = dev_net(dev);
3604 struct sk_buff *skb;
3605 int err = -EOPNOTSUPP;
3606
3607 if (!dev->netdev_ops->ndo_bridge_getlink)
3608 return 0;
3609
3610 skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
3611 if (!skb) {
3612 err = -ENOMEM;
3613 goto errout;
3614 }
3615
3616 err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
3617 if (err < 0)
3618 goto errout;
3619
3620 if (!skb->len)
3621 goto errout;
3622
3623 rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
3624 return 0;
3625 errout:
3626 WARN_ON(err == -EMSGSIZE);
3627 kfree_skb(skb);
3628 if (err)
3629 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
3630 return err;
3631 }
3632
3633 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3634 struct netlink_ext_ack *extack)
3635 {
3636 struct net *net = sock_net(skb->sk);
3637 struct ifinfomsg *ifm;
3638 struct net_device *dev;
3639 struct nlattr *br_spec, *attr = NULL;
3640 int rem, err = -EOPNOTSUPP;
3641 u16 flags = 0;
3642 bool have_flags = false;
3643
3644 if (nlmsg_len(nlh) < sizeof(*ifm))
3645 return -EINVAL;
3646
3647 ifm = nlmsg_data(nlh);
3648 if (ifm->ifi_family != AF_BRIDGE)
3649 return -EPFNOSUPPORT;
3650
3651 dev = __dev_get_by_index(net, ifm->ifi_index);
3652 if (!dev) {
3653 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3654 return -ENODEV;
3655 }
3656
3657 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3658 if (br_spec) {
3659 nla_for_each_nested(attr, br_spec, rem) {
3660 if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3661 if (nla_len(attr) < sizeof(flags))
3662 return -EINVAL;
3663
3664 have_flags = true;
3665 flags = nla_get_u16(attr);
3666 break;
3667 }
3668 }
3669 }
3670
3671 if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3672 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3673
3674 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
3675 err = -EOPNOTSUPP;
3676 goto out;
3677 }
3678
3679 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags);
3680 if (err)
3681 goto out;
3682
3683 flags &= ~BRIDGE_FLAGS_MASTER;
3684 }
3685
3686 if ((flags & BRIDGE_FLAGS_SELF)) {
3687 if (!dev->netdev_ops->ndo_bridge_setlink)
3688 err = -EOPNOTSUPP;
3689 else
3690 err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
3691 flags);
3692 if (!err) {
3693 flags &= ~BRIDGE_FLAGS_SELF;
3694
3695 /* Generate event to notify upper layer of bridge
3696 * change
3697 */
3698 err = rtnl_bridge_notify(dev);
3699 }
3700 }
3701
3702 if (have_flags)
3703 memcpy(nla_data(attr), &flags, sizeof(flags));
3704 out:
3705 return err;
3706 }
3707
3708 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
3709 struct netlink_ext_ack *extack)
3710 {
3711 struct net *net = sock_net(skb->sk);
3712 struct ifinfomsg *ifm;
3713 struct net_device *dev;
3714 struct nlattr *br_spec, *attr = NULL;
3715 int rem, err = -EOPNOTSUPP;
3716 u16 flags = 0;
3717 bool have_flags = false;
3718
3719 if (nlmsg_len(nlh) < sizeof(*ifm))
3720 return -EINVAL;
3721
3722 ifm = nlmsg_data(nlh);
3723 if (ifm->ifi_family != AF_BRIDGE)
3724 return -EPFNOSUPPORT;
3725
3726 dev = __dev_get_by_index(net, ifm->ifi_index);
3727 if (!dev) {
3728 pr_info("PF_BRIDGE: RTM_SETLINK with unknown ifindex\n");
3729 return -ENODEV;
3730 }
3731
3732 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3733 if (br_spec) {
3734 nla_for_each_nested(attr, br_spec, rem) {
3735 if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3736 if (nla_len(attr) < sizeof(flags))
3737 return -EINVAL;
3738
3739 have_flags = true;
3740 flags = nla_get_u16(attr);
3741 break;
3742 }
3743 }
3744 }
3745
3746 if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3747 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3748
3749 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
3750 err = -EOPNOTSUPP;
3751 goto out;
3752 }
3753
3754 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
3755 if (err)
3756 goto out;
3757
3758 flags &= ~BRIDGE_FLAGS_MASTER;
3759 }
3760
3761 if ((flags & BRIDGE_FLAGS_SELF)) {
3762 if (!dev->netdev_ops->ndo_bridge_dellink)
3763 err = -EOPNOTSUPP;
3764 else
3765 err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
3766 flags);
3767
3768 if (!err) {
3769 flags &= ~BRIDGE_FLAGS_SELF;
3770
3771 /* Generate event to notify upper layer of bridge
3772 * change
3773 */
3774 err = rtnl_bridge_notify(dev);
3775 }
3776 }
3777
3778 if (have_flags)
3779 memcpy(nla_data(attr), &flags, sizeof(flags));
3780 out:
3781 return err;
3782 }
3783
3784 static bool stats_attr_valid(unsigned int mask, int attrid, int idxattr)
3785 {
3786 return (mask & IFLA_STATS_FILTER_BIT(attrid)) &&
3787 (!idxattr || idxattr == attrid);
3788 }
3789
3790 #define IFLA_OFFLOAD_XSTATS_FIRST (IFLA_OFFLOAD_XSTATS_UNSPEC + 1)
3791 static int rtnl_get_offload_stats_attr_size(int attr_id)
3792 {
3793 switch (attr_id) {
3794 case IFLA_OFFLOAD_XSTATS_CPU_HIT:
3795 return sizeof(struct rtnl_link_stats64);
3796 }
3797
3798 return 0;
3799 }
3800
3801 static int rtnl_get_offload_stats(struct sk_buff *skb, struct net_device *dev,
3802 int *prividx)
3803 {
3804 struct nlattr *attr = NULL;
3805 int attr_id, size;
3806 void *attr_data;
3807 int err;
3808
3809 if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
3810 dev->netdev_ops->ndo_get_offload_stats))
3811 return -ENODATA;
3812
3813 for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
3814 attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
3815 if (attr_id < *prividx)
3816 continue;
3817
3818 size = rtnl_get_offload_stats_attr_size(attr_id);
3819 if (!size)
3820 continue;
3821
3822 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
3823 continue;
3824
3825 attr = nla_reserve_64bit(skb, attr_id, size,
3826 IFLA_OFFLOAD_XSTATS_UNSPEC);
3827 if (!attr)
3828 goto nla_put_failure;
3829
3830 attr_data = nla_data(attr);
3831 memset(attr_data, 0, size);
3832 err = dev->netdev_ops->ndo_get_offload_stats(attr_id, dev,
3833 attr_data);
3834 if (err)
3835 goto get_offload_stats_failure;
3836 }
3837
3838 if (!attr)
3839 return -ENODATA;
3840
3841 *prividx = 0;
3842 return 0;
3843
3844 nla_put_failure:
3845 err = -EMSGSIZE;
3846 get_offload_stats_failure:
3847 *prividx = attr_id;
3848 return err;
3849 }
3850
3851 static int rtnl_get_offload_stats_size(const struct net_device *dev)
3852 {
3853 int nla_size = 0;
3854 int attr_id;
3855 int size;
3856
3857 if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
3858 dev->netdev_ops->ndo_get_offload_stats))
3859 return 0;
3860
3861 for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
3862 attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
3863 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
3864 continue;
3865 size = rtnl_get_offload_stats_attr_size(attr_id);
3866 nla_size += nla_total_size_64bit(size);
3867 }
3868
3869 if (nla_size != 0)
3870 nla_size += nla_total_size(0);
3871
3872 return nla_size;
3873 }
3874
3875 static int rtnl_fill_statsinfo(struct sk_buff *skb, struct net_device *dev,
3876 int type, u32 pid, u32 seq, u32 change,
3877 unsigned int flags, unsigned int filter_mask,
3878 int *idxattr, int *prividx)
3879 {
3880 struct if_stats_msg *ifsm;
3881 struct nlmsghdr *nlh;
3882 struct nlattr *attr;
3883 int s_prividx = *prividx;
3884 int err;
3885
3886 ASSERT_RTNL();
3887
3888 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifsm), flags);
3889 if (!nlh)
3890 return -EMSGSIZE;
3891
3892 ifsm = nlmsg_data(nlh);
3893 ifsm->family = PF_UNSPEC;
3894 ifsm->pad1 = 0;
3895 ifsm->pad2 = 0;
3896 ifsm->ifindex = dev->ifindex;
3897 ifsm->filter_mask = filter_mask;
3898
3899 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, *idxattr)) {
3900 struct rtnl_link_stats64 *sp;
3901
3902 attr = nla_reserve_64bit(skb, IFLA_STATS_LINK_64,
3903 sizeof(struct rtnl_link_stats64),
3904 IFLA_STATS_UNSPEC);
3905 if (!attr)
3906 goto nla_put_failure;
3907
3908 sp = nla_data(attr);
3909 dev_get_stats(dev, sp);
3910 }
3911
3912 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, *idxattr)) {
3913 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
3914
3915 if (ops && ops->fill_linkxstats) {
3916 *idxattr = IFLA_STATS_LINK_XSTATS;
3917 attr = nla_nest_start(skb,
3918 IFLA_STATS_LINK_XSTATS);
3919 if (!attr)
3920 goto nla_put_failure;
3921
3922 err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
3923 nla_nest_end(skb, attr);
3924 if (err)
3925 goto nla_put_failure;
3926 *idxattr = 0;
3927 }
3928 }
3929
3930 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE,
3931 *idxattr)) {
3932 const struct rtnl_link_ops *ops = NULL;
3933 const struct net_device *master;
3934
3935 master = netdev_master_upper_dev_get(dev);
3936 if (master)
3937 ops = master->rtnl_link_ops;
3938 if (ops && ops->fill_linkxstats) {
3939 *idxattr = IFLA_STATS_LINK_XSTATS_SLAVE;
3940 attr = nla_nest_start(skb,
3941 IFLA_STATS_LINK_XSTATS_SLAVE);
3942 if (!attr)
3943 goto nla_put_failure;
3944
3945 err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
3946 nla_nest_end(skb, attr);
3947 if (err)
3948 goto nla_put_failure;
3949 *idxattr = 0;
3950 }
3951 }
3952
3953 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS,
3954 *idxattr)) {
3955 *idxattr = IFLA_STATS_LINK_OFFLOAD_XSTATS;
3956 attr = nla_nest_start(skb, IFLA_STATS_LINK_OFFLOAD_XSTATS);
3957 if (!attr)
3958 goto nla_put_failure;
3959
3960 err = rtnl_get_offload_stats(skb, dev, prividx);
3961 if (err == -ENODATA)
3962 nla_nest_cancel(skb, attr);
3963 else
3964 nla_nest_end(skb, attr);
3965
3966 if (err && err != -ENODATA)
3967 goto nla_put_failure;
3968 *idxattr = 0;
3969 }
3970
3971 if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, *idxattr)) {
3972 struct rtnl_af_ops *af_ops;
3973
3974 *idxattr = IFLA_STATS_AF_SPEC;
3975 attr = nla_nest_start(skb, IFLA_STATS_AF_SPEC);
3976 if (!attr)
3977 goto nla_put_failure;
3978
3979 list_for_each_entry(af_ops, &rtnl_af_ops, list) {
3980 if (af_ops->fill_stats_af) {
3981 struct nlattr *af;
3982 int err;
3983
3984 af = nla_nest_start(skb, af_ops->family);
3985 if (!af)
3986 goto nla_put_failure;
3987
3988 err = af_ops->fill_stats_af(skb, dev);
3989
3990 if (err == -ENODATA)
3991 nla_nest_cancel(skb, af);
3992 else if (err < 0)
3993 goto nla_put_failure;
3994
3995 nla_nest_end(skb, af);
3996 }
3997 }
3998
3999 nla_nest_end(skb, attr);
4000
4001 *idxattr = 0;
4002 }
4003
4004 nlmsg_end(skb, nlh);
4005
4006 return 0;
4007
4008 nla_put_failure:
4009 /* not a multi message or no progress mean a real error */
4010 if (!(flags & NLM_F_MULTI) || s_prividx == *prividx)
4011 nlmsg_cancel(skb, nlh);
4012 else
4013 nlmsg_end(skb, nlh);
4014
4015 return -EMSGSIZE;
4016 }
4017
4018 static size_t if_nlmsg_stats_size(const struct net_device *dev,
4019 u32 filter_mask)
4020 {
4021 size_t size = 0;
4022
4023 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, 0))
4024 size += nla_total_size_64bit(sizeof(struct rtnl_link_stats64));
4025
4026 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, 0)) {
4027 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
4028 int attr = IFLA_STATS_LINK_XSTATS;
4029
4030 if (ops && ops->get_linkxstats_size) {
4031 size += nla_total_size(ops->get_linkxstats_size(dev,
4032 attr));
4033 /* for IFLA_STATS_LINK_XSTATS */
4034 size += nla_total_size(0);
4035 }
4036 }
4037
4038 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE, 0)) {
4039 struct net_device *_dev = (struct net_device *)dev;
4040 const struct rtnl_link_ops *ops = NULL;
4041 const struct net_device *master;
4042
4043 /* netdev_master_upper_dev_get can't take const */
4044 master = netdev_master_upper_dev_get(_dev);
4045 if (master)
4046 ops = master->rtnl_link_ops;
4047 if (ops && ops->get_linkxstats_size) {
4048 int attr = IFLA_STATS_LINK_XSTATS_SLAVE;
4049
4050 size += nla_total_size(ops->get_linkxstats_size(dev,
4051 attr));
4052 /* for IFLA_STATS_LINK_XSTATS_SLAVE */
4053 size += nla_total_size(0);
4054 }
4055 }
4056
4057 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS, 0))
4058 size += rtnl_get_offload_stats_size(dev);
4059
4060 if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, 0)) {
4061 struct rtnl_af_ops *af_ops;
4062
4063 /* for IFLA_STATS_AF_SPEC */
4064 size += nla_total_size(0);
4065
4066 list_for_each_entry(af_ops, &rtnl_af_ops, list) {
4067 if (af_ops->get_stats_af_size) {
4068 size += nla_total_size(
4069 af_ops->get_stats_af_size(dev));
4070
4071 /* for AF_* */
4072 size += nla_total_size(0);
4073 }
4074 }
4075 }
4076
4077 return size;
4078 }
4079
4080 static int rtnl_stats_get(struct sk_buff *skb, struct nlmsghdr *nlh,
4081 struct netlink_ext_ack *extack)
4082 {
4083 struct net *net = sock_net(skb->sk);
4084 struct net_device *dev = NULL;
4085 int idxattr = 0, prividx = 0;
4086 struct if_stats_msg *ifsm;
4087 struct sk_buff *nskb;
4088 u32 filter_mask;
4089 int err;
4090
4091 if (nlmsg_len(nlh) < sizeof(*ifsm))
4092 return -EINVAL;
4093
4094 ifsm = nlmsg_data(nlh);
4095 if (ifsm->ifindex > 0)
4096 dev = __dev_get_by_index(net, ifsm->ifindex);
4097 else
4098 return -EINVAL;
4099
4100 if (!dev)
4101 return -ENODEV;
4102
4103 filter_mask = ifsm->filter_mask;
4104 if (!filter_mask)
4105 return -EINVAL;
4106
4107 nskb = nlmsg_new(if_nlmsg_stats_size(dev, filter_mask), GFP_KERNEL);
4108 if (!nskb)
4109 return -ENOBUFS;
4110
4111 err = rtnl_fill_statsinfo(nskb, dev, RTM_NEWSTATS,
4112 NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
4113 0, filter_mask, &idxattr, &prividx);
4114 if (err < 0) {
4115 /* -EMSGSIZE implies BUG in if_nlmsg_stats_size */
4116 WARN_ON(err == -EMSGSIZE);
4117 kfree_skb(nskb);
4118 } else {
4119 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
4120 }
4121
4122 return err;
4123 }
4124
4125 static int rtnl_stats_dump(struct sk_buff *skb, struct netlink_callback *cb)
4126 {
4127 int h, s_h, err, s_idx, s_idxattr, s_prividx;
4128 struct net *net = sock_net(skb->sk);
4129 unsigned int flags = NLM_F_MULTI;
4130 struct if_stats_msg *ifsm;
4131 struct hlist_head *head;
4132 struct net_device *dev;
4133 u32 filter_mask = 0;
4134 int idx = 0;
4135
4136 s_h = cb->args[0];
4137 s_idx = cb->args[1];
4138 s_idxattr = cb->args[2];
4139 s_prividx = cb->args[3];
4140
4141 cb->seq = net->dev_base_seq;
4142
4143 if (nlmsg_len(cb->nlh) < sizeof(*ifsm))
4144 return -EINVAL;
4145
4146 ifsm = nlmsg_data(cb->nlh);
4147 filter_mask = ifsm->filter_mask;
4148 if (!filter_mask)
4149 return -EINVAL;
4150
4151 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4152 idx = 0;
4153 head = &net->dev_index_head[h];
4154 hlist_for_each_entry(dev, head, index_hlist) {
4155 if (idx < s_idx)
4156 goto cont;
4157 err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS,
4158 NETLINK_CB(cb->skb).portid,
4159 cb->nlh->nlmsg_seq, 0,
4160 flags, filter_mask,
4161 &s_idxattr, &s_prividx);
4162 /* If we ran out of room on the first message,
4163 * we're in trouble
4164 */
4165 WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
4166
4167 if (err < 0)
4168 goto out;
4169 s_prividx = 0;
4170 s_idxattr = 0;
4171 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4172 cont:
4173 idx++;
4174 }
4175 }
4176 out:
4177 cb->args[3] = s_prividx;
4178 cb->args[2] = s_idxattr;
4179 cb->args[1] = idx;
4180 cb->args[0] = h;
4181
4182 return skb->len;
4183 }
4184
4185 /* Process one rtnetlink message. */
4186
4187 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh,
4188 struct netlink_ext_ack *extack)
4189 {
4190 struct net *net = sock_net(skb->sk);
4191 struct rtnl_link *handlers;
4192 int err = -EOPNOTSUPP;
4193 rtnl_doit_func doit;
4194 unsigned int flags;
4195 int kind;
4196 int family;
4197 int type;
4198
4199 type = nlh->nlmsg_type;
4200 if (type > RTM_MAX)
4201 return -EOPNOTSUPP;
4202
4203 type -= RTM_BASE;
4204
4205 /* All the messages must have at least 1 byte length */
4206 if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
4207 return 0;
4208
4209 family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
4210 kind = type&3;
4211
4212 if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
4213 return -EPERM;
4214
4215 if (family >= ARRAY_SIZE(rtnl_msg_handlers))
4216 family = PF_UNSPEC;
4217
4218 rcu_read_lock();
4219 handlers = rcu_dereference(rtnl_msg_handlers[family]);
4220 if (!handlers) {
4221 family = PF_UNSPEC;
4222 handlers = rcu_dereference(rtnl_msg_handlers[family]);
4223 }
4224
4225 if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
4226 struct sock *rtnl;
4227 rtnl_dumpit_func dumpit;
4228 u16 min_dump_alloc = 0;
4229
4230 dumpit = READ_ONCE(handlers[type].dumpit);
4231 if (!dumpit) {
4232 family = PF_UNSPEC;
4233 handlers = rcu_dereference(rtnl_msg_handlers[PF_UNSPEC]);
4234 if (!handlers)
4235 goto err_unlock;
4236
4237 dumpit = READ_ONCE(handlers[type].dumpit);
4238 if (!dumpit)
4239 goto err_unlock;
4240 }
4241
4242 refcount_inc(&rtnl_msg_handlers_ref[family]);
4243
4244 if (type == RTM_GETLINK - RTM_BASE)
4245 min_dump_alloc = rtnl_calcit(skb, nlh);
4246
4247 rcu_read_unlock();
4248
4249 rtnl = net->rtnl;
4250 {
4251 struct netlink_dump_control c = {
4252 .dump = dumpit,
4253 .min_dump_alloc = min_dump_alloc,
4254 };
4255 err = netlink_dump_start(rtnl, skb, nlh, &c);
4256 }
4257 refcount_dec(&rtnl_msg_handlers_ref[family]);
4258 return err;
4259 }
4260
4261 doit = READ_ONCE(handlers[type].doit);
4262 if (!doit) {
4263 family = PF_UNSPEC;
4264 handlers = rcu_dereference(rtnl_msg_handlers[family]);
4265 }
4266
4267 flags = READ_ONCE(handlers[type].flags);
4268 if (flags & RTNL_FLAG_DOIT_UNLOCKED) {
4269 refcount_inc(&rtnl_msg_handlers_ref[family]);
4270 doit = READ_ONCE(handlers[type].doit);
4271 rcu_read_unlock();
4272 if (doit)
4273 err = doit(skb, nlh, extack);
4274 refcount_dec(&rtnl_msg_handlers_ref[family]);
4275 return err;
4276 }
4277
4278 rcu_read_unlock();
4279
4280 rtnl_lock();
4281 handlers = rtnl_dereference(rtnl_msg_handlers[family]);
4282 if (handlers) {
4283 doit = READ_ONCE(handlers[type].doit);
4284 if (doit)
4285 err = doit(skb, nlh, extack);
4286 }
4287 rtnl_unlock();
4288 return err;
4289
4290 err_unlock:
4291 rcu_read_unlock();
4292 return -EOPNOTSUPP;
4293 }
4294
4295 static void rtnetlink_rcv(struct sk_buff *skb)
4296 {
4297 netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
4298 }
4299
4300 static int rtnetlink_bind(struct net *net, int group)
4301 {
4302 switch (group) {
4303 case RTNLGRP_IPV4_MROUTE_R:
4304 case RTNLGRP_IPV6_MROUTE_R:
4305 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
4306 return -EPERM;
4307 break;
4308 }
4309 return 0;
4310 }
4311
4312 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
4313 {
4314 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
4315
4316 switch (event) {
4317 case NETDEV_REBOOT:
4318 case NETDEV_CHANGEMTU:
4319 case NETDEV_CHANGEADDR:
4320 case NETDEV_CHANGENAME:
4321 case NETDEV_FEAT_CHANGE:
4322 case NETDEV_BONDING_FAILOVER:
4323 case NETDEV_POST_TYPE_CHANGE:
4324 case NETDEV_NOTIFY_PEERS:
4325 case NETDEV_CHANGEUPPER:
4326 case NETDEV_RESEND_IGMP:
4327 case NETDEV_CHANGEINFODATA:
4328 case NETDEV_CHANGE_TX_QUEUE_LEN:
4329 rtmsg_ifinfo_event(RTM_NEWLINK, dev, 0, rtnl_get_event(event),
4330 GFP_KERNEL);
4331 break;
4332 default:
4333 break;
4334 }
4335 return NOTIFY_DONE;
4336 }
4337
4338 static struct notifier_block rtnetlink_dev_notifier = {
4339 .notifier_call = rtnetlink_event,
4340 };
4341
4342
4343 static int __net_init rtnetlink_net_init(struct net *net)
4344 {
4345 struct sock *sk;
4346 struct netlink_kernel_cfg cfg = {
4347 .groups = RTNLGRP_MAX,
4348 .input = rtnetlink_rcv,
4349 .cb_mutex = &rtnl_mutex,
4350 .flags = NL_CFG_F_NONROOT_RECV,
4351 .bind = rtnetlink_bind,
4352 };
4353
4354 sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
4355 if (!sk)
4356 return -ENOMEM;
4357 net->rtnl = sk;
4358 return 0;
4359 }
4360
4361 static void __net_exit rtnetlink_net_exit(struct net *net)
4362 {
4363 netlink_kernel_release(net->rtnl);
4364 net->rtnl = NULL;
4365 }
4366
4367 static struct pernet_operations rtnetlink_net_ops = {
4368 .init = rtnetlink_net_init,
4369 .exit = rtnetlink_net_exit,
4370 };
4371
4372 void __init rtnetlink_init(void)
4373 {
4374 int i;
4375
4376 for (i = 0; i < ARRAY_SIZE(rtnl_msg_handlers_ref); i++)
4377 refcount_set(&rtnl_msg_handlers_ref[i], 1);
4378
4379 if (register_pernet_subsys(&rtnetlink_net_ops))
4380 panic("rtnetlink_init: cannot initialize rtnetlink\n");
4381
4382 register_netdevice_notifier(&rtnetlink_dev_notifier);
4383
4384 rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
4385 rtnl_dump_ifinfo, 0);
4386 rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, 0);
4387 rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, 0);
4388 rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, 0);
4389
4390 rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, 0);
4391 rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, 0);
4392 rtnl_register(PF_UNSPEC, RTM_GETNETCONF, NULL, rtnl_dump_all, 0);
4393
4394 rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, 0);
4395 rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, 0);
4396 rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, 0);
4397
4398 rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, 0);
4399 rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, 0);
4400 rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, 0);
4401
4402 rtnl_register(PF_UNSPEC, RTM_GETSTATS, rtnl_stats_get, rtnl_stats_dump,
4403 0);
4404 }
Linux with added FPC-III support