]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blob - net/core/rtnetlink.c
projects / mirror_ubuntu-bionic-kernel.git / blob
? search:


fececbf6840a16d03156c8553e11ff0ed58cbc54
[mirror_ubuntu-bionic-kernel.git] / net / core / rtnetlink.c
1 /*
2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
5 *
6 * Routing netlink socket interface: protocol independent part.
7 *
8 * Authors: Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
9 *
10 * This program is free software; you can redistribute it and/or
11 * modify it under the terms of the GNU General Public License
12 * as published by the Free Software Foundation; either version
13 * 2 of the License, or (at your option) any later version.
14 *
15 * Fixes:
16 * Vitaly E. Lavrov RTA_OK arithmetics was wrong.
17 */
18
19 #include <linux/bitops.h>
20 #include <linux/errno.h>
21 #include <linux/module.h>
22 #include <linux/types.h>
23 #include <linux/socket.h>
24 #include <linux/kernel.h>
25 #include <linux/timer.h>
26 #include <linux/string.h>
27 #include <linux/sockios.h>
28 #include <linux/net.h>
29 #include <linux/fcntl.h>
30 #include <linux/mm.h>
31 #include <linux/slab.h>
32 #include <linux/interrupt.h>
33 #include <linux/capability.h>
34 #include <linux/skbuff.h>
35 #include <linux/init.h>
36 #include <linux/security.h>
37 #include <linux/mutex.h>
38 #include <linux/if_addr.h>
39 #include <linux/if_bridge.h>
40 #include <linux/if_vlan.h>
41 #include <linux/pci.h>
42 #include <linux/etherdevice.h>
43 #include <linux/bpf.h>
44
45 #include <linux/uaccess.h>
46
47 #include <linux/inet.h>
48 #include <linux/netdevice.h>
49 #include <net/switchdev.h>
50 #include <net/ip.h>
51 #include <net/protocol.h>
52 #include <net/arp.h>
53 #include <net/route.h>
54 #include <net/udp.h>
55 #include <net/tcp.h>
56 #include <net/sock.h>
57 #include <net/pkt_sched.h>
58 #include <net/fib_rules.h>
59 #include <net/rtnetlink.h>
60 #include <net/net_namespace.h>
61
62 struct rtnl_link {
63 rtnl_doit_func doit;
64 rtnl_dumpit_func dumpit;
65 unsigned int flags;
66 };
67
68 static DEFINE_MUTEX(rtnl_mutex);
69
70 void rtnl_lock(void)
71 {
72 mutex_lock(&rtnl_mutex);
73 }
74 EXPORT_SYMBOL(rtnl_lock);
75
76 static struct sk_buff *defer_kfree_skb_list;
77 void rtnl_kfree_skbs(struct sk_buff *head, struct sk_buff *tail)
78 {
79 if (head && tail) {
80 tail->next = defer_kfree_skb_list;
81 defer_kfree_skb_list = head;
82 }
83 }
84 EXPORT_SYMBOL(rtnl_kfree_skbs);
85
86 void __rtnl_unlock(void)
87 {
88 struct sk_buff *head = defer_kfree_skb_list;
89
90 defer_kfree_skb_list = NULL;
91
92 mutex_unlock(&rtnl_mutex);
93
94 while (head) {
95 struct sk_buff *next = head->next;
96
97 kfree_skb(head);
98 cond_resched();
99 head = next;
100 }
101 }
102
103 void rtnl_unlock(void)
104 {
105 /* This fellow will unlock it for us. */
106 netdev_run_todo();
107 }
108 EXPORT_SYMBOL(rtnl_unlock);
109
110 int rtnl_trylock(void)
111 {
112 return mutex_trylock(&rtnl_mutex);
113 }
114 EXPORT_SYMBOL(rtnl_trylock);
115
116 int rtnl_is_locked(void)
117 {
118 return mutex_is_locked(&rtnl_mutex);
119 }
120 EXPORT_SYMBOL(rtnl_is_locked);
121
122 #ifdef CONFIG_PROVE_LOCKING
123 bool lockdep_rtnl_is_held(void)
124 {
125 return lockdep_is_held(&rtnl_mutex);
126 }
127 EXPORT_SYMBOL(lockdep_rtnl_is_held);
128 #endif /* #ifdef CONFIG_PROVE_LOCKING */
129
130 static struct rtnl_link __rcu *rtnl_msg_handlers[RTNL_FAMILY_MAX + 1];
131 static refcount_t rtnl_msg_handlers_ref[RTNL_FAMILY_MAX + 1];
132
133 static inline int rtm_msgindex(int msgtype)
134 {
135 int msgindex = msgtype - RTM_BASE;
136
137 /*
138 * msgindex < 0 implies someone tried to register a netlink
139 * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
140 * the message type has not been added to linux/rtnetlink.h
141 */
142 BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
143
144 return msgindex;
145 }
146
147 /**
148 * __rtnl_register - Register a rtnetlink message type
149 * @protocol: Protocol family or PF_UNSPEC
150 * @msgtype: rtnetlink message type
151 * @doit: Function pointer called for each request message
152 * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
153 * @flags: rtnl_link_flags to modifiy behaviour of doit/dumpit functions
154 *
155 * Registers the specified function pointers (at least one of them has
156 * to be non-NULL) to be called whenever a request message for the
157 * specified protocol family and message type is received.
158 *
159 * The special protocol family PF_UNSPEC may be used to define fallback
160 * function pointers for the case when no entry for the specific protocol
161 * family exists.
162 *
163 * Returns 0 on success or a negative error code.
164 */
165 int __rtnl_register(int protocol, int msgtype,
166 rtnl_doit_func doit, rtnl_dumpit_func dumpit,
167 unsigned int flags)
168 {
169 struct rtnl_link *tab;
170 int msgindex;
171
172 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
173 msgindex = rtm_msgindex(msgtype);
174
175 tab = rcu_dereference_raw(rtnl_msg_handlers[protocol]);
176 if (tab == NULL) {
177 tab = kcalloc(RTM_NR_MSGTYPES, sizeof(*tab), GFP_KERNEL);
178 if (tab == NULL)
179 return -ENOBUFS;
180
181 rcu_assign_pointer(rtnl_msg_handlers[protocol], tab);
182 }
183
184 if (doit)
185 tab[msgindex].doit = doit;
186 if (dumpit)
187 tab[msgindex].dumpit = dumpit;
188 tab[msgindex].flags |= flags;
189
190 return 0;
191 }
192 EXPORT_SYMBOL_GPL(__rtnl_register);
193
194 /**
195 * rtnl_register - Register a rtnetlink message type
196 *
197 * Identical to __rtnl_register() but panics on failure. This is useful
198 * as failure of this function is very unlikely, it can only happen due
199 * to lack of memory when allocating the chain to store all message
200 * handlers for a protocol. Meant for use in init functions where lack
201 * of memory implies no sense in continuing.
202 */
203 void rtnl_register(int protocol, int msgtype,
204 rtnl_doit_func doit, rtnl_dumpit_func dumpit,
205 unsigned int flags)
206 {
207 if (__rtnl_register(protocol, msgtype, doit, dumpit, flags) < 0)
208 panic("Unable to register rtnetlink message handler, "
209 "protocol = %d, message type = %d\n",
210 protocol, msgtype);
211 }
212 EXPORT_SYMBOL_GPL(rtnl_register);
213
214 /**
215 * rtnl_unregister - Unregister a rtnetlink message type
216 * @protocol: Protocol family or PF_UNSPEC
217 * @msgtype: rtnetlink message type
218 *
219 * Returns 0 on success or a negative error code.
220 */
221 int rtnl_unregister(int protocol, int msgtype)
222 {
223 struct rtnl_link *handlers;
224 int msgindex;
225
226 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
227 msgindex = rtm_msgindex(msgtype);
228
229 rtnl_lock();
230 handlers = rtnl_dereference(rtnl_msg_handlers[protocol]);
231 if (!handlers) {
232 rtnl_unlock();
233 return -ENOENT;
234 }
235
236 handlers[msgindex].doit = NULL;
237 handlers[msgindex].dumpit = NULL;
238 handlers[msgindex].flags = 0;
239 rtnl_unlock();
240
241 return 0;
242 }
243 EXPORT_SYMBOL_GPL(rtnl_unregister);
244
245 /**
246 * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
247 * @protocol : Protocol family or PF_UNSPEC
248 *
249 * Identical to calling rtnl_unregster() for all registered message types
250 * of a certain protocol family.
251 */
252 void rtnl_unregister_all(int protocol)
253 {
254 struct rtnl_link *handlers;
255
256 BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
257
258 rtnl_lock();
259 handlers = rtnl_dereference(rtnl_msg_handlers[protocol]);
260 RCU_INIT_POINTER(rtnl_msg_handlers[protocol], NULL);
261 rtnl_unlock();
262
263 synchronize_net();
264
265 while (refcount_read(&rtnl_msg_handlers_ref[protocol]) > 1)
266 schedule();
267 kfree(handlers);
268 }
269 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
270
271 static LIST_HEAD(link_ops);
272
273 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
274 {
275 const struct rtnl_link_ops *ops;
276
277 list_for_each_entry(ops, &link_ops, list) {
278 if (!strcmp(ops->kind, kind))
279 return ops;
280 }
281 return NULL;
282 }
283
284 /**
285 * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
286 * @ops: struct rtnl_link_ops * to register
287 *
288 * The caller must hold the rtnl_mutex. This function should be used
289 * by drivers that create devices during module initialization. It
290 * must be called before registering the devices.
291 *
292 * Returns 0 on success or a negative error code.
293 */
294 int __rtnl_link_register(struct rtnl_link_ops *ops)
295 {
296 if (rtnl_link_ops_get(ops->kind))
297 return -EEXIST;
298
299 /* The check for setup is here because if ops
300 * does not have that filled up, it is not possible
301 * to use the ops for creating device. So do not
302 * fill up dellink as well. That disables rtnl_dellink.
303 */
304 if (ops->setup && !ops->dellink)
305 ops->dellink = unregister_netdevice_queue;
306
307 list_add_tail(&ops->list, &link_ops);
308 return 0;
309 }
310 EXPORT_SYMBOL_GPL(__rtnl_link_register);
311
312 /**
313 * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
314 * @ops: struct rtnl_link_ops * to register
315 *
316 * Returns 0 on success or a negative error code.
317 */
318 int rtnl_link_register(struct rtnl_link_ops *ops)
319 {
320 int err;
321
322 rtnl_lock();
323 err = __rtnl_link_register(ops);
324 rtnl_unlock();
325 return err;
326 }
327 EXPORT_SYMBOL_GPL(rtnl_link_register);
328
329 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
330 {
331 struct net_device *dev;
332 LIST_HEAD(list_kill);
333
334 for_each_netdev(net, dev) {
335 if (dev->rtnl_link_ops == ops)
336 ops->dellink(dev, &list_kill);
337 }
338 unregister_netdevice_many(&list_kill);
339 }
340
341 /**
342 * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
343 * @ops: struct rtnl_link_ops * to unregister
344 *
345 * The caller must hold the rtnl_mutex.
346 */
347 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
348 {
349 struct net *net;
350
351 for_each_net(net) {
352 __rtnl_kill_links(net, ops);
353 }
354 list_del(&ops->list);
355 }
356 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
357
358 /* Return with the rtnl_lock held when there are no network
359 * devices unregistering in any network namespace.
360 */
361 static void rtnl_lock_unregistering_all(void)
362 {
363 struct net *net;
364 bool unregistering;
365 DEFINE_WAIT_FUNC(wait, woken_wake_function);
366
367 add_wait_queue(&netdev_unregistering_wq, &wait);
368 for (;;) {
369 unregistering = false;
370 rtnl_lock();
371 for_each_net(net) {
372 if (net->dev_unreg_count > 0) {
373 unregistering = true;
374 break;
375 }
376 }
377 if (!unregistering)
378 break;
379 __rtnl_unlock();
380
381 wait_woken(&wait, TASK_UNINTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
382 }
383 remove_wait_queue(&netdev_unregistering_wq, &wait);
384 }
385
386 /**
387 * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
388 * @ops: struct rtnl_link_ops * to unregister
389 */
390 void rtnl_link_unregister(struct rtnl_link_ops *ops)
391 {
392 /* Close the race with cleanup_net() */
393 mutex_lock(&net_mutex);
394 rtnl_lock_unregistering_all();
395 __rtnl_link_unregister(ops);
396 rtnl_unlock();
397 mutex_unlock(&net_mutex);
398 }
399 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
400
401 static size_t rtnl_link_get_slave_info_data_size(const struct net_device *dev)
402 {
403 struct net_device *master_dev;
404 const struct rtnl_link_ops *ops;
405 size_t size = 0;
406
407 rcu_read_lock();
408
409 master_dev = netdev_master_upper_dev_get_rcu((struct net_device *)dev);
410 if (!master_dev)
411 goto out;
412
413 ops = master_dev->rtnl_link_ops;
414 if (!ops || !ops->get_slave_size)
415 goto out;
416 /* IFLA_INFO_SLAVE_DATA + nested data */
417 size = nla_total_size(sizeof(struct nlattr)) +
418 ops->get_slave_size(master_dev, dev);
419
420 out:
421 rcu_read_unlock();
422 return size;
423 }
424
425 static size_t rtnl_link_get_size(const struct net_device *dev)
426 {
427 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
428 size_t size;
429
430 if (!ops)
431 return 0;
432
433 size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
434 nla_total_size(strlen(ops->kind) + 1); /* IFLA_INFO_KIND */
435
436 if (ops->get_size)
437 /* IFLA_INFO_DATA + nested data */
438 size += nla_total_size(sizeof(struct nlattr)) +
439 ops->get_size(dev);
440
441 if (ops->get_xstats_size)
442 /* IFLA_INFO_XSTATS */
443 size += nla_total_size(ops->get_xstats_size(dev));
444
445 size += rtnl_link_get_slave_info_data_size(dev);
446
447 return size;
448 }
449
450 static LIST_HEAD(rtnl_af_ops);
451
452 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
453 {
454 const struct rtnl_af_ops *ops;
455
456 list_for_each_entry_rcu(ops, &rtnl_af_ops, list) {
457 if (ops->family == family)
458 return ops;
459 }
460
461 return NULL;
462 }
463
464 /**
465 * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
466 * @ops: struct rtnl_af_ops * to register
467 *
468 * Returns 0 on success or a negative error code.
469 */
470 void rtnl_af_register(struct rtnl_af_ops *ops)
471 {
472 rtnl_lock();
473 list_add_tail_rcu(&ops->list, &rtnl_af_ops);
474 rtnl_unlock();
475 }
476 EXPORT_SYMBOL_GPL(rtnl_af_register);
477
478 /**
479 * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
480 * @ops: struct rtnl_af_ops * to unregister
481 */
482 void rtnl_af_unregister(struct rtnl_af_ops *ops)
483 {
484 rtnl_lock();
485 list_del_rcu(&ops->list);
486 rtnl_unlock();
487
488 synchronize_rcu();
489 }
490 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
491
492 static size_t rtnl_link_get_af_size(const struct net_device *dev,
493 u32 ext_filter_mask)
494 {
495 struct rtnl_af_ops *af_ops;
496 size_t size;
497
498 /* IFLA_AF_SPEC */
499 size = nla_total_size(sizeof(struct nlattr));
500
501 rcu_read_lock();
502 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
503 if (af_ops->get_link_af_size) {
504 /* AF_* + nested data */
505 size += nla_total_size(sizeof(struct nlattr)) +
506 af_ops->get_link_af_size(dev, ext_filter_mask);
507 }
508 }
509 rcu_read_unlock();
510
511 return size;
512 }
513
514 static bool rtnl_have_link_slave_info(const struct net_device *dev)
515 {
516 struct net_device *master_dev;
517 bool ret = false;
518
519 rcu_read_lock();
520
521 master_dev = netdev_master_upper_dev_get_rcu((struct net_device *)dev);
522 if (master_dev && master_dev->rtnl_link_ops)
523 ret = true;
524 rcu_read_unlock();
525 return ret;
526 }
527
528 static int rtnl_link_slave_info_fill(struct sk_buff *skb,
529 const struct net_device *dev)
530 {
531 struct net_device *master_dev;
532 const struct rtnl_link_ops *ops;
533 struct nlattr *slave_data;
534 int err;
535
536 master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
537 if (!master_dev)
538 return 0;
539 ops = master_dev->rtnl_link_ops;
540 if (!ops)
541 return 0;
542 if (nla_put_string(skb, IFLA_INFO_SLAVE_KIND, ops->kind) < 0)
543 return -EMSGSIZE;
544 if (ops->fill_slave_info) {
545 slave_data = nla_nest_start(skb, IFLA_INFO_SLAVE_DATA);
546 if (!slave_data)
547 return -EMSGSIZE;
548 err = ops->fill_slave_info(skb, master_dev, dev);
549 if (err < 0)
550 goto err_cancel_slave_data;
551 nla_nest_end(skb, slave_data);
552 }
553 return 0;
554
555 err_cancel_slave_data:
556 nla_nest_cancel(skb, slave_data);
557 return err;
558 }
559
560 static int rtnl_link_info_fill(struct sk_buff *skb,
561 const struct net_device *dev)
562 {
563 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
564 struct nlattr *data;
565 int err;
566
567 if (!ops)
568 return 0;
569 if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
570 return -EMSGSIZE;
571 if (ops->fill_xstats) {
572 err = ops->fill_xstats(skb, dev);
573 if (err < 0)
574 return err;
575 }
576 if (ops->fill_info) {
577 data = nla_nest_start(skb, IFLA_INFO_DATA);
578 if (data == NULL)
579 return -EMSGSIZE;
580 err = ops->fill_info(skb, dev);
581 if (err < 0)
582 goto err_cancel_data;
583 nla_nest_end(skb, data);
584 }
585 return 0;
586
587 err_cancel_data:
588 nla_nest_cancel(skb, data);
589 return err;
590 }
591
592 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
593 {
594 struct nlattr *linkinfo;
595 int err = -EMSGSIZE;
596
597 linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
598 if (linkinfo == NULL)
599 goto out;
600
601 err = rtnl_link_info_fill(skb, dev);
602 if (err < 0)
603 goto err_cancel_link;
604
605 err = rtnl_link_slave_info_fill(skb, dev);
606 if (err < 0)
607 goto err_cancel_link;
608
609 nla_nest_end(skb, linkinfo);
610 return 0;
611
612 err_cancel_link:
613 nla_nest_cancel(skb, linkinfo);
614 out:
615 return err;
616 }
617
618 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
619 {
620 struct sock *rtnl = net->rtnl;
621 int err = 0;
622
623 NETLINK_CB(skb).dst_group = group;
624 if (echo)
625 refcount_inc(&skb->users);
626 netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
627 if (echo)
628 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
629 return err;
630 }
631
632 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
633 {
634 struct sock *rtnl = net->rtnl;
635
636 return nlmsg_unicast(rtnl, skb, pid);
637 }
638 EXPORT_SYMBOL(rtnl_unicast);
639
640 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
641 struct nlmsghdr *nlh, gfp_t flags)
642 {
643 struct sock *rtnl = net->rtnl;
644 int report = 0;
645
646 if (nlh)
647 report = nlmsg_report(nlh);
648
649 nlmsg_notify(rtnl, skb, pid, group, report, flags);
650 }
651 EXPORT_SYMBOL(rtnl_notify);
652
653 void rtnl_set_sk_err(struct net *net, u32 group, int error)
654 {
655 struct sock *rtnl = net->rtnl;
656
657 netlink_set_err(rtnl, 0, group, error);
658 }
659 EXPORT_SYMBOL(rtnl_set_sk_err);
660
661 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
662 {
663 struct nlattr *mx;
664 int i, valid = 0;
665
666 mx = nla_nest_start(skb, RTA_METRICS);
667 if (mx == NULL)
668 return -ENOBUFS;
669
670 for (i = 0; i < RTAX_MAX; i++) {
671 if (metrics[i]) {
672 if (i == RTAX_CC_ALGO - 1) {
673 char tmp[TCP_CA_NAME_MAX], *name;
674
675 name = tcp_ca_get_name_by_key(metrics[i], tmp);
676 if (!name)
677 continue;
678 if (nla_put_string(skb, i + 1, name))
679 goto nla_put_failure;
680 } else if (i == RTAX_FEATURES - 1) {
681 u32 user_features = metrics[i] & RTAX_FEATURE_MASK;
682
683 if (!user_features)
684 continue;
685 BUILD_BUG_ON(RTAX_FEATURE_MASK & DST_FEATURE_MASK);
686 if (nla_put_u32(skb, i + 1, user_features))
687 goto nla_put_failure;
688 } else {
689 if (nla_put_u32(skb, i + 1, metrics[i]))
690 goto nla_put_failure;
691 }
692 valid++;
693 }
694 }
695
696 if (!valid) {
697 nla_nest_cancel(skb, mx);
698 return 0;
699 }
700
701 return nla_nest_end(skb, mx);
702
703 nla_put_failure:
704 nla_nest_cancel(skb, mx);
705 return -EMSGSIZE;
706 }
707 EXPORT_SYMBOL(rtnetlink_put_metrics);
708
709 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
710 long expires, u32 error)
711 {
712 struct rta_cacheinfo ci = {
713 .rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse),
714 .rta_used = dst->__use,
715 .rta_clntref = atomic_read(&(dst->__refcnt)),
716 .rta_error = error,
717 .rta_id = id,
718 };
719
720 if (expires) {
721 unsigned long clock;
722
723 clock = jiffies_to_clock_t(abs(expires));
724 clock = min_t(unsigned long, clock, INT_MAX);
725 ci.rta_expires = (expires > 0) ? clock : -clock;
726 }
727 return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
728 }
729 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
730
731 static void set_operstate(struct net_device *dev, unsigned char transition)
732 {
733 unsigned char operstate = dev->operstate;
734
735 switch (transition) {
736 case IF_OPER_UP:
737 if ((operstate == IF_OPER_DORMANT ||
738 operstate == IF_OPER_UNKNOWN) &&
739 !netif_dormant(dev))
740 operstate = IF_OPER_UP;
741 break;
742
743 case IF_OPER_DORMANT:
744 if (operstate == IF_OPER_UP ||
745 operstate == IF_OPER_UNKNOWN)
746 operstate = IF_OPER_DORMANT;
747 break;
748 }
749
750 if (dev->operstate != operstate) {
751 write_lock_bh(&dev_base_lock);
752 dev->operstate = operstate;
753 write_unlock_bh(&dev_base_lock);
754 netdev_state_change(dev);
755 }
756 }
757
758 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
759 {
760 return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
761 (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
762 }
763
764 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
765 const struct ifinfomsg *ifm)
766 {
767 unsigned int flags = ifm->ifi_flags;
768
769 /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
770 if (ifm->ifi_change)
771 flags = (flags & ifm->ifi_change) |
772 (rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
773
774 return flags;
775 }
776
777 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
778 const struct rtnl_link_stats64 *b)
779 {
780 a->rx_packets = b->rx_packets;
781 a->tx_packets = b->tx_packets;
782 a->rx_bytes = b->rx_bytes;
783 a->tx_bytes = b->tx_bytes;
784 a->rx_errors = b->rx_errors;
785 a->tx_errors = b->tx_errors;
786 a->rx_dropped = b->rx_dropped;
787 a->tx_dropped = b->tx_dropped;
788
789 a->multicast = b->multicast;
790 a->collisions = b->collisions;
791
792 a->rx_length_errors = b->rx_length_errors;
793 a->rx_over_errors = b->rx_over_errors;
794 a->rx_crc_errors = b->rx_crc_errors;
795 a->rx_frame_errors = b->rx_frame_errors;
796 a->rx_fifo_errors = b->rx_fifo_errors;
797 a->rx_missed_errors = b->rx_missed_errors;
798
799 a->tx_aborted_errors = b->tx_aborted_errors;
800 a->tx_carrier_errors = b->tx_carrier_errors;
801 a->tx_fifo_errors = b->tx_fifo_errors;
802 a->tx_heartbeat_errors = b->tx_heartbeat_errors;
803 a->tx_window_errors = b->tx_window_errors;
804
805 a->rx_compressed = b->rx_compressed;
806 a->tx_compressed = b->tx_compressed;
807
808 a->rx_nohandler = b->rx_nohandler;
809 }
810
811 /* All VF info */
812 static inline int rtnl_vfinfo_size(const struct net_device *dev,
813 u32 ext_filter_mask)
814 {
815 if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF)) {
816 int num_vfs = dev_num_vf(dev->dev.parent);
817 size_t size = nla_total_size(0);
818 size += num_vfs *
819 (nla_total_size(0) +
820 nla_total_size(sizeof(struct ifla_vf_mac)) +
821 nla_total_size(sizeof(struct ifla_vf_vlan)) +
822 nla_total_size(0) + /* nest IFLA_VF_VLAN_LIST */
823 nla_total_size(MAX_VLAN_LIST_LEN *
824 sizeof(struct ifla_vf_vlan_info)) +
825 nla_total_size(sizeof(struct ifla_vf_spoofchk)) +
826 nla_total_size(sizeof(struct ifla_vf_tx_rate)) +
827 nla_total_size(sizeof(struct ifla_vf_rate)) +
828 nla_total_size(sizeof(struct ifla_vf_link_state)) +
829 nla_total_size(sizeof(struct ifla_vf_rss_query_en)) +
830 nla_total_size(0) + /* nest IFLA_VF_STATS */
831 /* IFLA_VF_STATS_RX_PACKETS */
832 nla_total_size_64bit(sizeof(__u64)) +
833 /* IFLA_VF_STATS_TX_PACKETS */
834 nla_total_size_64bit(sizeof(__u64)) +
835 /* IFLA_VF_STATS_RX_BYTES */
836 nla_total_size_64bit(sizeof(__u64)) +
837 /* IFLA_VF_STATS_TX_BYTES */
838 nla_total_size_64bit(sizeof(__u64)) +
839 /* IFLA_VF_STATS_BROADCAST */
840 nla_total_size_64bit(sizeof(__u64)) +
841 /* IFLA_VF_STATS_MULTICAST */
842 nla_total_size_64bit(sizeof(__u64)) +
843 nla_total_size(sizeof(struct ifla_vf_trust)));
844 return size;
845 } else
846 return 0;
847 }
848
849 static size_t rtnl_port_size(const struct net_device *dev,
850 u32 ext_filter_mask)
851 {
852 size_t port_size = nla_total_size(4) /* PORT_VF */
853 + nla_total_size(PORT_PROFILE_MAX) /* PORT_PROFILE */
854 + nla_total_size(PORT_UUID_MAX) /* PORT_INSTANCE_UUID */
855 + nla_total_size(PORT_UUID_MAX) /* PORT_HOST_UUID */
856 + nla_total_size(1) /* PROT_VDP_REQUEST */
857 + nla_total_size(2); /* PORT_VDP_RESPONSE */
858 size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
859 size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
860 + port_size;
861 size_t port_self_size = nla_total_size(sizeof(struct nlattr))
862 + port_size;
863
864 if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
865 !(ext_filter_mask & RTEXT_FILTER_VF))
866 return 0;
867 if (dev_num_vf(dev->dev.parent))
868 return port_self_size + vf_ports_size +
869 vf_port_size * dev_num_vf(dev->dev.parent);
870 else
871 return port_self_size;
872 }
873
874 static size_t rtnl_xdp_size(void)
875 {
876 size_t xdp_size = nla_total_size(0) + /* nest IFLA_XDP */
877 nla_total_size(1) + /* XDP_ATTACHED */
878 nla_total_size(4); /* XDP_PROG_ID */
879
880 return xdp_size;
881 }
882
883 static noinline size_t if_nlmsg_size(const struct net_device *dev,
884 u32 ext_filter_mask)
885 {
886 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
887 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
888 + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
889 + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
890 + nla_total_size_64bit(sizeof(struct rtnl_link_ifmap))
891 + nla_total_size(sizeof(struct rtnl_link_stats))
892 + nla_total_size_64bit(sizeof(struct rtnl_link_stats64))
893 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
894 + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
895 + nla_total_size(4) /* IFLA_TXQLEN */
896 + nla_total_size(4) /* IFLA_WEIGHT */
897 + nla_total_size(4) /* IFLA_MTU */
898 + nla_total_size(4) /* IFLA_LINK */
899 + nla_total_size(4) /* IFLA_MASTER */
900 + nla_total_size(1) /* IFLA_CARRIER */
901 + nla_total_size(4) /* IFLA_PROMISCUITY */
902 + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
903 + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
904 + nla_total_size(4) /* IFLA_GSO_MAX_SEGS */
905 + nla_total_size(4) /* IFLA_GSO_MAX_SIZE */
906 + nla_total_size(1) /* IFLA_OPERSTATE */
907 + nla_total_size(1) /* IFLA_LINKMODE */
908 + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
909 + nla_total_size(4) /* IFLA_LINK_NETNSID */
910 + nla_total_size(4) /* IFLA_GROUP */
911 + nla_total_size(ext_filter_mask
912 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
913 + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
914 + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
915 + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
916 + rtnl_link_get_af_size(dev, ext_filter_mask) /* IFLA_AF_SPEC */
917 + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */
918 + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_SWITCH_ID */
919 + nla_total_size(IFNAMSIZ) /* IFLA_PHYS_PORT_NAME */
920 + rtnl_xdp_size() /* IFLA_XDP */
921 + nla_total_size(4) /* IFLA_EVENT */
922 + nla_total_size(4) /* IFLA_NEW_NETNSID */
923 + nla_total_size(4) /* IFLA_NEW_IFINDEX */
924 + nla_total_size(1) /* IFLA_PROTO_DOWN */
925 + nla_total_size(4) /* IFLA_IF_NETNSID */
926 + nla_total_size(4) /* IFLA_CARRIER_UP_COUNT */
927 + nla_total_size(4) /* IFLA_CARRIER_DOWN_COUNT */
928 + 0;
929 }
930
931 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
932 {
933 struct nlattr *vf_ports;
934 struct nlattr *vf_port;
935 int vf;
936 int err;
937
938 vf_ports = nla_nest_start(skb, IFLA_VF_PORTS);
939 if (!vf_ports)
940 return -EMSGSIZE;
941
942 for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
943 vf_port = nla_nest_start(skb, IFLA_VF_PORT);
944 if (!vf_port)
945 goto nla_put_failure;
946 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
947 goto nla_put_failure;
948 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
949 if (err == -EMSGSIZE)
950 goto nla_put_failure;
951 if (err) {
952 nla_nest_cancel(skb, vf_port);
953 continue;
954 }
955 nla_nest_end(skb, vf_port);
956 }
957
958 nla_nest_end(skb, vf_ports);
959
960 return 0;
961
962 nla_put_failure:
963 nla_nest_cancel(skb, vf_ports);
964 return -EMSGSIZE;
965 }
966
967 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
968 {
969 struct nlattr *port_self;
970 int err;
971
972 port_self = nla_nest_start(skb, IFLA_PORT_SELF);
973 if (!port_self)
974 return -EMSGSIZE;
975
976 err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
977 if (err) {
978 nla_nest_cancel(skb, port_self);
979 return (err == -EMSGSIZE) ? err : 0;
980 }
981
982 nla_nest_end(skb, port_self);
983
984 return 0;
985 }
986
987 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
988 u32 ext_filter_mask)
989 {
990 int err;
991
992 if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
993 !(ext_filter_mask & RTEXT_FILTER_VF))
994 return 0;
995
996 err = rtnl_port_self_fill(skb, dev);
997 if (err)
998 return err;
999
1000 if (dev_num_vf(dev->dev.parent)) {
1001 err = rtnl_vf_ports_fill(skb, dev);
1002 if (err)
1003 return err;
1004 }
1005
1006 return 0;
1007 }
1008
1009 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
1010 {
1011 int err;
1012 struct netdev_phys_item_id ppid;
1013
1014 err = dev_get_phys_port_id(dev, &ppid);
1015 if (err) {
1016 if (err == -EOPNOTSUPP)
1017 return 0;
1018 return err;
1019 }
1020
1021 if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
1022 return -EMSGSIZE;
1023
1024 return 0;
1025 }
1026
1027 static int rtnl_phys_port_name_fill(struct sk_buff *skb, struct net_device *dev)
1028 {
1029 char name[IFNAMSIZ];
1030 int err;
1031
1032 err = dev_get_phys_port_name(dev, name, sizeof(name));
1033 if (err) {
1034 if (err == -EOPNOTSUPP)
1035 return 0;
1036 return err;
1037 }
1038
1039 if (nla_put_string(skb, IFLA_PHYS_PORT_NAME, name))
1040 return -EMSGSIZE;
1041
1042 return 0;
1043 }
1044
1045 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev)
1046 {
1047 int err;
1048 struct switchdev_attr attr = {
1049 .orig_dev = dev,
1050 .id = SWITCHDEV_ATTR_ID_PORT_PARENT_ID,
1051 .flags = SWITCHDEV_F_NO_RECURSE,
1052 };
1053
1054 err = switchdev_port_attr_get(dev, &attr);
1055 if (err) {
1056 if (err == -EOPNOTSUPP)
1057 return 0;
1058 return err;
1059 }
1060
1061 if (nla_put(skb, IFLA_PHYS_SWITCH_ID, attr.u.ppid.id_len,
1062 attr.u.ppid.id))
1063 return -EMSGSIZE;
1064
1065 return 0;
1066 }
1067
1068 static noinline_for_stack int rtnl_fill_stats(struct sk_buff *skb,
1069 struct net_device *dev)
1070 {
1071 struct rtnl_link_stats64 *sp;
1072 struct nlattr *attr;
1073
1074 attr = nla_reserve_64bit(skb, IFLA_STATS64,
1075 sizeof(struct rtnl_link_stats64), IFLA_PAD);
1076 if (!attr)
1077 return -EMSGSIZE;
1078
1079 sp = nla_data(attr);
1080 dev_get_stats(dev, sp);
1081
1082 attr = nla_reserve(skb, IFLA_STATS,
1083 sizeof(struct rtnl_link_stats));
1084 if (!attr)
1085 return -EMSGSIZE;
1086
1087 copy_rtnl_link_stats(nla_data(attr), sp);
1088
1089 return 0;
1090 }
1091
1092 static noinline_for_stack int rtnl_fill_vfinfo(struct sk_buff *skb,
1093 struct net_device *dev,
1094 int vfs_num,
1095 struct nlattr *vfinfo)
1096 {
1097 struct ifla_vf_rss_query_en vf_rss_query_en;
1098 struct nlattr *vf, *vfstats, *vfvlanlist;
1099 struct ifla_vf_link_state vf_linkstate;
1100 struct ifla_vf_vlan_info vf_vlan_info;
1101 struct ifla_vf_spoofchk vf_spoofchk;
1102 struct ifla_vf_tx_rate vf_tx_rate;
1103 struct ifla_vf_stats vf_stats;
1104 struct ifla_vf_trust vf_trust;
1105 struct ifla_vf_vlan vf_vlan;
1106 struct ifla_vf_rate vf_rate;
1107 struct ifla_vf_mac vf_mac;
1108 struct ifla_vf_info ivi;
1109
1110 memset(&ivi, 0, sizeof(ivi));
1111
1112 /* Not all SR-IOV capable drivers support the
1113 * spoofcheck and "RSS query enable" query. Preset to
1114 * -1 so the user space tool can detect that the driver
1115 * didn't report anything.
1116 */
1117 ivi.spoofchk = -1;
1118 ivi.rss_query_en = -1;
1119 ivi.trusted = -1;
1120 /* The default value for VF link state is "auto"
1121 * IFLA_VF_LINK_STATE_AUTO which equals zero
1122 */
1123 ivi.linkstate = 0;
1124 /* VLAN Protocol by default is 802.1Q */
1125 ivi.vlan_proto = htons(ETH_P_8021Q);
1126 if (dev->netdev_ops->ndo_get_vf_config(dev, vfs_num, &ivi))
1127 return 0;
1128
1129 memset(&vf_vlan_info, 0, sizeof(vf_vlan_info));
1130
1131 vf_mac.vf =
1132 vf_vlan.vf =
1133 vf_vlan_info.vf =
1134 vf_rate.vf =
1135 vf_tx_rate.vf =
1136 vf_spoofchk.vf =
1137 vf_linkstate.vf =
1138 vf_rss_query_en.vf =
1139 vf_trust.vf = ivi.vf;
1140
1141 memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1142 vf_vlan.vlan = ivi.vlan;
1143 vf_vlan.qos = ivi.qos;
1144 vf_vlan_info.vlan = ivi.vlan;
1145 vf_vlan_info.qos = ivi.qos;
1146 vf_vlan_info.vlan_proto = ivi.vlan_proto;
1147 vf_tx_rate.rate = ivi.max_tx_rate;
1148 vf_rate.min_tx_rate = ivi.min_tx_rate;
1149 vf_rate.max_tx_rate = ivi.max_tx_rate;
1150 vf_spoofchk.setting = ivi.spoofchk;
1151 vf_linkstate.link_state = ivi.linkstate;
1152 vf_rss_query_en.setting = ivi.rss_query_en;
1153 vf_trust.setting = ivi.trusted;
1154 vf = nla_nest_start(skb, IFLA_VF_INFO);
1155 if (!vf)
1156 goto nla_put_vfinfo_failure;
1157 if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1158 nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1159 nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1160 &vf_rate) ||
1161 nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1162 &vf_tx_rate) ||
1163 nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1164 &vf_spoofchk) ||
1165 nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1166 &vf_linkstate) ||
1167 nla_put(skb, IFLA_VF_RSS_QUERY_EN,
1168 sizeof(vf_rss_query_en),
1169 &vf_rss_query_en) ||
1170 nla_put(skb, IFLA_VF_TRUST,
1171 sizeof(vf_trust), &vf_trust))
1172 goto nla_put_vf_failure;
1173 vfvlanlist = nla_nest_start(skb, IFLA_VF_VLAN_LIST);
1174 if (!vfvlanlist)
1175 goto nla_put_vf_failure;
1176 if (nla_put(skb, IFLA_VF_VLAN_INFO, sizeof(vf_vlan_info),
1177 &vf_vlan_info)) {
1178 nla_nest_cancel(skb, vfvlanlist);
1179 goto nla_put_vf_failure;
1180 }
1181 nla_nest_end(skb, vfvlanlist);
1182 memset(&vf_stats, 0, sizeof(vf_stats));
1183 if (dev->netdev_ops->ndo_get_vf_stats)
1184 dev->netdev_ops->ndo_get_vf_stats(dev, vfs_num,
1185 &vf_stats);
1186 vfstats = nla_nest_start(skb, IFLA_VF_STATS);
1187 if (!vfstats)
1188 goto nla_put_vf_failure;
1189 if (nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_PACKETS,
1190 vf_stats.rx_packets, IFLA_VF_STATS_PAD) ||
1191 nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_PACKETS,
1192 vf_stats.tx_packets, IFLA_VF_STATS_PAD) ||
1193 nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_BYTES,
1194 vf_stats.rx_bytes, IFLA_VF_STATS_PAD) ||
1195 nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_BYTES,
1196 vf_stats.tx_bytes, IFLA_VF_STATS_PAD) ||
1197 nla_put_u64_64bit(skb, IFLA_VF_STATS_BROADCAST,
1198 vf_stats.broadcast, IFLA_VF_STATS_PAD) ||
1199 nla_put_u64_64bit(skb, IFLA_VF_STATS_MULTICAST,
1200 vf_stats.multicast, IFLA_VF_STATS_PAD)) {
1201 nla_nest_cancel(skb, vfstats);
1202 goto nla_put_vf_failure;
1203 }
1204 nla_nest_end(skb, vfstats);
1205 nla_nest_end(skb, vf);
1206 return 0;
1207
1208 nla_put_vf_failure:
1209 nla_nest_cancel(skb, vf);
1210 nla_put_vfinfo_failure:
1211 nla_nest_cancel(skb, vfinfo);
1212 return -EMSGSIZE;
1213 }
1214
1215 static noinline_for_stack int rtnl_fill_vf(struct sk_buff *skb,
1216 struct net_device *dev,
1217 u32 ext_filter_mask)
1218 {
1219 struct nlattr *vfinfo;
1220 int i, num_vfs;
1221
1222 if (!dev->dev.parent || ((ext_filter_mask & RTEXT_FILTER_VF) == 0))
1223 return 0;
1224
1225 num_vfs = dev_num_vf(dev->dev.parent);
1226 if (nla_put_u32(skb, IFLA_NUM_VF, num_vfs))
1227 return -EMSGSIZE;
1228
1229 if (!dev->netdev_ops->ndo_get_vf_config)
1230 return 0;
1231
1232 vfinfo = nla_nest_start(skb, IFLA_VFINFO_LIST);
1233 if (!vfinfo)
1234 return -EMSGSIZE;
1235
1236 for (i = 0; i < num_vfs; i++) {
1237 if (rtnl_fill_vfinfo(skb, dev, i, vfinfo))
1238 return -EMSGSIZE;
1239 }
1240
1241 nla_nest_end(skb, vfinfo);
1242 return 0;
1243 }
1244
1245 static int rtnl_fill_link_ifmap(struct sk_buff *skb, struct net_device *dev)
1246 {
1247 struct rtnl_link_ifmap map;
1248
1249 memset(&map, 0, sizeof(map));
1250 map.mem_start = dev->mem_start;
1251 map.mem_end = dev->mem_end;
1252 map.base_addr = dev->base_addr;
1253 map.irq = dev->irq;
1254 map.dma = dev->dma;
1255 map.port = dev->if_port;
1256
1257 if (nla_put_64bit(skb, IFLA_MAP, sizeof(map), &map, IFLA_PAD))
1258 return -EMSGSIZE;
1259
1260 return 0;
1261 }
1262
1263 static u8 rtnl_xdp_attached_mode(struct net_device *dev, u32 *prog_id)
1264 {
1265 const struct net_device_ops *ops = dev->netdev_ops;
1266 const struct bpf_prog *generic_xdp_prog;
1267
1268 ASSERT_RTNL();
1269
1270 *prog_id = 0;
1271 generic_xdp_prog = rtnl_dereference(dev->xdp_prog);
1272 if (generic_xdp_prog) {
1273 *prog_id = generic_xdp_prog->aux->id;
1274 return XDP_ATTACHED_SKB;
1275 }
1276 if (!ops->ndo_bpf)
1277 return XDP_ATTACHED_NONE;
1278
1279 return __dev_xdp_attached(dev, ops->ndo_bpf, prog_id);
1280 }
1281
1282 static int rtnl_xdp_fill(struct sk_buff *skb, struct net_device *dev)
1283 {
1284 struct nlattr *xdp;
1285 u32 prog_id;
1286 int err;
1287
1288 xdp = nla_nest_start(skb, IFLA_XDP);
1289 if (!xdp)
1290 return -EMSGSIZE;
1291
1292 err = nla_put_u8(skb, IFLA_XDP_ATTACHED,
1293 rtnl_xdp_attached_mode(dev, &prog_id));
1294 if (err)
1295 goto err_cancel;
1296
1297 if (prog_id) {
1298 err = nla_put_u32(skb, IFLA_XDP_PROG_ID, prog_id);
1299 if (err)
1300 goto err_cancel;
1301 }
1302
1303 nla_nest_end(skb, xdp);
1304 return 0;
1305
1306 err_cancel:
1307 nla_nest_cancel(skb, xdp);
1308 return err;
1309 }
1310
1311 static u32 rtnl_get_event(unsigned long event)
1312 {
1313 u32 rtnl_event_type = IFLA_EVENT_NONE;
1314
1315 switch (event) {
1316 case NETDEV_REBOOT:
1317 rtnl_event_type = IFLA_EVENT_REBOOT;
1318 break;
1319 case NETDEV_FEAT_CHANGE:
1320 rtnl_event_type = IFLA_EVENT_FEATURES;
1321 break;
1322 case NETDEV_BONDING_FAILOVER:
1323 rtnl_event_type = IFLA_EVENT_BONDING_FAILOVER;
1324 break;
1325 case NETDEV_NOTIFY_PEERS:
1326 rtnl_event_type = IFLA_EVENT_NOTIFY_PEERS;
1327 break;
1328 case NETDEV_RESEND_IGMP:
1329 rtnl_event_type = IFLA_EVENT_IGMP_RESEND;
1330 break;
1331 case NETDEV_CHANGEINFODATA:
1332 rtnl_event_type = IFLA_EVENT_BONDING_OPTIONS;
1333 break;
1334 default:
1335 break;
1336 }
1337
1338 return rtnl_event_type;
1339 }
1340
1341 static int put_master_ifindex(struct sk_buff *skb, struct net_device *dev)
1342 {
1343 const struct net_device *upper_dev;
1344 int ret = 0;
1345
1346 rcu_read_lock();
1347
1348 upper_dev = netdev_master_upper_dev_get_rcu(dev);
1349 if (upper_dev)
1350 ret = nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex);
1351
1352 rcu_read_unlock();
1353 return ret;
1354 }
1355
1356 static int nla_put_iflink(struct sk_buff *skb, const struct net_device *dev)
1357 {
1358 int ifindex = dev_get_iflink(dev);
1359
1360 if (dev->ifindex == ifindex)
1361 return 0;
1362
1363 return nla_put_u32(skb, IFLA_LINK, ifindex);
1364 }
1365
1366 static noinline_for_stack int nla_put_ifalias(struct sk_buff *skb,
1367 struct net_device *dev)
1368 {
1369 char buf[IFALIASZ];
1370 int ret;
1371
1372 ret = dev_get_alias(dev, buf, sizeof(buf));
1373 return ret > 0 ? nla_put_string(skb, IFLA_IFALIAS, buf) : 0;
1374 }
1375
1376 static int rtnl_fill_link_netnsid(struct sk_buff *skb,
1377 const struct net_device *dev,
1378 struct net *src_net)
1379 {
1380 if (dev->rtnl_link_ops && dev->rtnl_link_ops->get_link_net) {
1381 struct net *link_net = dev->rtnl_link_ops->get_link_net(dev);
1382
1383 if (!net_eq(dev_net(dev), link_net)) {
1384 int id = peernet2id_alloc(src_net, link_net);
1385
1386 if (nla_put_s32(skb, IFLA_LINK_NETNSID, id))
1387 return -EMSGSIZE;
1388 }
1389 }
1390
1391 return 0;
1392 }
1393
1394 static int rtnl_fill_link_af(struct sk_buff *skb,
1395 const struct net_device *dev,
1396 u32 ext_filter_mask)
1397 {
1398 const struct rtnl_af_ops *af_ops;
1399 struct nlattr *af_spec;
1400
1401 af_spec = nla_nest_start(skb, IFLA_AF_SPEC);
1402 if (!af_spec)
1403 return -EMSGSIZE;
1404
1405 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
1406 struct nlattr *af;
1407 int err;
1408
1409 if (!af_ops->fill_link_af)
1410 continue;
1411
1412 af = nla_nest_start(skb, af_ops->family);
1413 if (!af)
1414 return -EMSGSIZE;
1415
1416 err = af_ops->fill_link_af(skb, dev, ext_filter_mask);
1417 /*
1418 * Caller may return ENODATA to indicate that there
1419 * was no data to be dumped. This is not an error, it
1420 * means we should trim the attribute header and
1421 * continue.
1422 */
1423 if (err == -ENODATA)
1424 nla_nest_cancel(skb, af);
1425 else if (err < 0)
1426 return -EMSGSIZE;
1427
1428 nla_nest_end(skb, af);
1429 }
1430
1431 nla_nest_end(skb, af_spec);
1432 return 0;
1433 }
1434
1435 static int rtnl_fill_ifinfo(struct sk_buff *skb,
1436 struct net_device *dev, struct net *src_net,
1437 int type, u32 pid, u32 seq, u32 change,
1438 unsigned int flags, u32 ext_filter_mask,
1439 u32 event, int *new_nsid, int new_ifindex,
1440 int tgt_netnsid)
1441 {
1442 struct ifinfomsg *ifm;
1443 struct nlmsghdr *nlh;
1444
1445 ASSERT_RTNL();
1446 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
1447 if (nlh == NULL)
1448 return -EMSGSIZE;
1449
1450 ifm = nlmsg_data(nlh);
1451 ifm->ifi_family = AF_UNSPEC;
1452 ifm->__ifi_pad = 0;
1453 ifm->ifi_type = dev->type;
1454 ifm->ifi_index = dev->ifindex;
1455 ifm->ifi_flags = dev_get_flags(dev);
1456 ifm->ifi_change = change;
1457
1458 if (tgt_netnsid >= 0 && nla_put_s32(skb, IFLA_IF_NETNSID, tgt_netnsid))
1459 goto nla_put_failure;
1460
1461 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
1462 nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
1463 nla_put_u8(skb, IFLA_OPERSTATE,
1464 netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
1465 nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
1466 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
1467 nla_put_u32(skb, IFLA_GROUP, dev->group) ||
1468 nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
1469 nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
1470 nla_put_u32(skb, IFLA_GSO_MAX_SEGS, dev->gso_max_segs) ||
1471 nla_put_u32(skb, IFLA_GSO_MAX_SIZE, dev->gso_max_size) ||
1472 #ifdef CONFIG_RPS
1473 nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1474 #endif
1475 nla_put_iflink(skb, dev) ||
1476 put_master_ifindex(skb, dev) ||
1477 nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1478 (dev->qdisc &&
1479 nla_put_string(skb, IFLA_QDISC, dev->qdisc->ops->id)) ||
1480 nla_put_ifalias(skb, dev) ||
1481 nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1482 atomic_read(&dev->carrier_up_count) +
1483 atomic_read(&dev->carrier_down_count)) ||
1484 nla_put_u8(skb, IFLA_PROTO_DOWN, dev->proto_down) ||
1485 nla_put_u32(skb, IFLA_CARRIER_UP_COUNT,
1486 atomic_read(&dev->carrier_up_count)) ||
1487 nla_put_u32(skb, IFLA_CARRIER_DOWN_COUNT,
1488 atomic_read(&dev->carrier_down_count)))
1489 goto nla_put_failure;
1490
1491 if (event != IFLA_EVENT_NONE) {
1492 if (nla_put_u32(skb, IFLA_EVENT, event))
1493 goto nla_put_failure;
1494 }
1495
1496 if (rtnl_fill_link_ifmap(skb, dev))
1497 goto nla_put_failure;
1498
1499 if (dev->addr_len) {
1500 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1501 nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1502 goto nla_put_failure;
1503 }
1504
1505 if (rtnl_phys_port_id_fill(skb, dev))
1506 goto nla_put_failure;
1507
1508 if (rtnl_phys_port_name_fill(skb, dev))
1509 goto nla_put_failure;
1510
1511 if (rtnl_phys_switch_id_fill(skb, dev))
1512 goto nla_put_failure;
1513
1514 if (rtnl_fill_stats(skb, dev))
1515 goto nla_put_failure;
1516
1517 if (rtnl_fill_vf(skb, dev, ext_filter_mask))
1518 goto nla_put_failure;
1519
1520 if (rtnl_port_fill(skb, dev, ext_filter_mask))
1521 goto nla_put_failure;
1522
1523 if (rtnl_xdp_fill(skb, dev))
1524 goto nla_put_failure;
1525
1526 if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1527 if (rtnl_link_fill(skb, dev) < 0)
1528 goto nla_put_failure;
1529 }
1530
1531 if (rtnl_fill_link_netnsid(skb, dev, src_net))
1532 goto nla_put_failure;
1533
1534 if (new_nsid &&
1535 nla_put_s32(skb, IFLA_NEW_NETNSID, *new_nsid) < 0)
1536 goto nla_put_failure;
1537 if (new_ifindex &&
1538 nla_put_s32(skb, IFLA_NEW_IFINDEX, new_ifindex) < 0)
1539 goto nla_put_failure;
1540
1541
1542 rcu_read_lock();
1543 if (rtnl_fill_link_af(skb, dev, ext_filter_mask))
1544 goto nla_put_failure_rcu;
1545 rcu_read_unlock();
1546
1547 nlmsg_end(skb, nlh);
1548 return 0;
1549
1550 nla_put_failure_rcu:
1551 rcu_read_unlock();
1552 nla_put_failure:
1553 nlmsg_cancel(skb, nlh);
1554 return -EMSGSIZE;
1555 }
1556
1557 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1558 [IFLA_IFNAME] = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1559 [IFLA_ADDRESS] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1560 [IFLA_BROADCAST] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1561 [IFLA_MAP] = { .len = sizeof(struct rtnl_link_ifmap) },
1562 [IFLA_MTU] = { .type = NLA_U32 },
1563 [IFLA_LINK] = { .type = NLA_U32 },
1564 [IFLA_MASTER] = { .type = NLA_U32 },
1565 [IFLA_CARRIER] = { .type = NLA_U8 },
1566 [IFLA_TXQLEN] = { .type = NLA_U32 },
1567 [IFLA_WEIGHT] = { .type = NLA_U32 },
1568 [IFLA_OPERSTATE] = { .type = NLA_U8 },
1569 [IFLA_LINKMODE] = { .type = NLA_U8 },
1570 [IFLA_LINKINFO] = { .type = NLA_NESTED },
1571 [IFLA_NET_NS_PID] = { .type = NLA_U32 },
1572 [IFLA_NET_NS_FD] = { .type = NLA_U32 },
1573 /* IFLA_IFALIAS is a string, but policy is set to NLA_BINARY to
1574 * allow 0-length string (needed to remove an alias).
1575 */
1576 [IFLA_IFALIAS] = { .type = NLA_BINARY, .len = IFALIASZ - 1 },
1577 [IFLA_VFINFO_LIST] = {. type = NLA_NESTED },
1578 [IFLA_VF_PORTS] = { .type = NLA_NESTED },
1579 [IFLA_PORT_SELF] = { .type = NLA_NESTED },
1580 [IFLA_AF_SPEC] = { .type = NLA_NESTED },
1581 [IFLA_EXT_MASK] = { .type = NLA_U32 },
1582 [IFLA_PROMISCUITY] = { .type = NLA_U32 },
1583 [IFLA_NUM_TX_QUEUES] = { .type = NLA_U32 },
1584 [IFLA_NUM_RX_QUEUES] = { .type = NLA_U32 },
1585 [IFLA_PHYS_PORT_ID] = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1586 [IFLA_CARRIER_CHANGES] = { .type = NLA_U32 }, /* ignored */
1587 [IFLA_PHYS_SWITCH_ID] = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1588 [IFLA_LINK_NETNSID] = { .type = NLA_S32 },
1589 [IFLA_PROTO_DOWN] = { .type = NLA_U8 },
1590 [IFLA_XDP] = { .type = NLA_NESTED },
1591 [IFLA_EVENT] = { .type = NLA_U32 },
1592 [IFLA_GROUP] = { .type = NLA_U32 },
1593 [IFLA_IF_NETNSID] = { .type = NLA_S32 },
1594 [IFLA_CARRIER_UP_COUNT] = { .type = NLA_U32 },
1595 [IFLA_CARRIER_DOWN_COUNT] = { .type = NLA_U32 },
1596 };
1597
1598 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1599 [IFLA_INFO_KIND] = { .type = NLA_STRING },
1600 [IFLA_INFO_DATA] = { .type = NLA_NESTED },
1601 [IFLA_INFO_SLAVE_KIND] = { .type = NLA_STRING },
1602 [IFLA_INFO_SLAVE_DATA] = { .type = NLA_NESTED },
1603 };
1604
1605 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1606 [IFLA_VF_MAC] = { .len = sizeof(struct ifla_vf_mac) },
1607 [IFLA_VF_VLAN] = { .len = sizeof(struct ifla_vf_vlan) },
1608 [IFLA_VF_VLAN_LIST] = { .type = NLA_NESTED },
1609 [IFLA_VF_TX_RATE] = { .len = sizeof(struct ifla_vf_tx_rate) },
1610 [IFLA_VF_SPOOFCHK] = { .len = sizeof(struct ifla_vf_spoofchk) },
1611 [IFLA_VF_RATE] = { .len = sizeof(struct ifla_vf_rate) },
1612 [IFLA_VF_LINK_STATE] = { .len = sizeof(struct ifla_vf_link_state) },
1613 [IFLA_VF_RSS_QUERY_EN] = { .len = sizeof(struct ifla_vf_rss_query_en) },
1614 [IFLA_VF_STATS] = { .type = NLA_NESTED },
1615 [IFLA_VF_TRUST] = { .len = sizeof(struct ifla_vf_trust) },
1616 [IFLA_VF_IB_NODE_GUID] = { .len = sizeof(struct ifla_vf_guid) },
1617 [IFLA_VF_IB_PORT_GUID] = { .len = sizeof(struct ifla_vf_guid) },
1618 };
1619
1620 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1621 [IFLA_PORT_VF] = { .type = NLA_U32 },
1622 [IFLA_PORT_PROFILE] = { .type = NLA_STRING,
1623 .len = PORT_PROFILE_MAX },
1624 [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1625 .len = PORT_UUID_MAX },
1626 [IFLA_PORT_HOST_UUID] = { .type = NLA_STRING,
1627 .len = PORT_UUID_MAX },
1628 [IFLA_PORT_REQUEST] = { .type = NLA_U8, },
1629 [IFLA_PORT_RESPONSE] = { .type = NLA_U16, },
1630
1631 /* Unused, but we need to keep it here since user space could
1632 * fill it. It's also broken with regard to NLA_BINARY use in
1633 * combination with structs.
1634 */
1635 [IFLA_PORT_VSI_TYPE] = { .type = NLA_BINARY,
1636 .len = sizeof(struct ifla_port_vsi) },
1637 };
1638
1639 static const struct nla_policy ifla_xdp_policy[IFLA_XDP_MAX + 1] = {
1640 [IFLA_XDP_FD] = { .type = NLA_S32 },
1641 [IFLA_XDP_ATTACHED] = { .type = NLA_U8 },
1642 [IFLA_XDP_FLAGS] = { .type = NLA_U32 },
1643 [IFLA_XDP_PROG_ID] = { .type = NLA_U32 },
1644 };
1645
1646 static const struct rtnl_link_ops *linkinfo_to_kind_ops(const struct nlattr *nla)
1647 {
1648 const struct rtnl_link_ops *ops = NULL;
1649 struct nlattr *linfo[IFLA_INFO_MAX + 1];
1650
1651 if (nla_parse_nested(linfo, IFLA_INFO_MAX, nla,
1652 ifla_info_policy, NULL) < 0)
1653 return NULL;
1654
1655 if (linfo[IFLA_INFO_KIND]) {
1656 char kind[MODULE_NAME_LEN];
1657
1658 nla_strlcpy(kind, linfo[IFLA_INFO_KIND], sizeof(kind));
1659 ops = rtnl_link_ops_get(kind);
1660 }
1661
1662 return ops;
1663 }
1664
1665 static bool link_master_filtered(struct net_device *dev, int master_idx)
1666 {
1667 struct net_device *master;
1668
1669 if (!master_idx)
1670 return false;
1671
1672 master = netdev_master_upper_dev_get(dev);
1673 if (!master || master->ifindex != master_idx)
1674 return true;
1675
1676 return false;
1677 }
1678
1679 static bool link_kind_filtered(const struct net_device *dev,
1680 const struct rtnl_link_ops *kind_ops)
1681 {
1682 if (kind_ops && dev->rtnl_link_ops != kind_ops)
1683 return true;
1684
1685 return false;
1686 }
1687
1688 static bool link_dump_filtered(struct net_device *dev,
1689 int master_idx,
1690 const struct rtnl_link_ops *kind_ops)
1691 {
1692 if (link_master_filtered(dev, master_idx) ||
1693 link_kind_filtered(dev, kind_ops))
1694 return true;
1695
1696 return false;
1697 }
1698
1699 static struct net *get_target_net(struct sock *sk, int netnsid)
1700 {
1701 struct net *net;
1702
1703 net = get_net_ns_by_id(sock_net(sk), netnsid);
1704 if (!net)
1705 return ERR_PTR(-EINVAL);
1706
1707 /* For now, the caller is required to have CAP_NET_ADMIN in
1708 * the user namespace owning the target net ns.
1709 */
1710 if (!sk_ns_capable(sk, net->user_ns, CAP_NET_ADMIN)) {
1711 put_net(net);
1712 return ERR_PTR(-EACCES);
1713 }
1714 return net;
1715 }
1716
1717 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
1718 {
1719 struct net *net = sock_net(skb->sk);
1720 struct net *tgt_net = net;
1721 int h, s_h;
1722 int idx = 0, s_idx;
1723 struct net_device *dev;
1724 struct hlist_head *head;
1725 struct nlattr *tb[IFLA_MAX+1];
1726 u32 ext_filter_mask = 0;
1727 const struct rtnl_link_ops *kind_ops = NULL;
1728 unsigned int flags = NLM_F_MULTI;
1729 int master_idx = 0;
1730 int netnsid = -1;
1731 int err;
1732 int hdrlen;
1733
1734 s_h = cb->args[0];
1735 s_idx = cb->args[1];
1736
1737 /* A hack to preserve kernel<->userspace interface.
1738 * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
1739 * However, before Linux v3.9 the code here assumed rtgenmsg and that's
1740 * what iproute2 < v3.9.0 used.
1741 * We can detect the old iproute2. Even including the IFLA_EXT_MASK
1742 * attribute, its netlink message is shorter than struct ifinfomsg.
1743 */
1744 hdrlen = nlmsg_len(cb->nlh) < sizeof(struct ifinfomsg) ?
1745 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
1746
1747 if (nlmsg_parse(cb->nlh, hdrlen, tb, IFLA_MAX,
1748 ifla_policy, NULL) >= 0) {
1749 if (tb[IFLA_IF_NETNSID]) {
1750 netnsid = nla_get_s32(tb[IFLA_IF_NETNSID]);
1751 tgt_net = get_target_net(skb->sk, netnsid);
1752 if (IS_ERR(tgt_net)) {
1753 tgt_net = net;
1754 netnsid = -1;
1755 }
1756 }
1757
1758 if (tb[IFLA_EXT_MASK])
1759 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
1760
1761 if (tb[IFLA_MASTER])
1762 master_idx = nla_get_u32(tb[IFLA_MASTER]);
1763
1764 if (tb[IFLA_LINKINFO])
1765 kind_ops = linkinfo_to_kind_ops(tb[IFLA_LINKINFO]);
1766
1767 if (master_idx || kind_ops)
1768 flags |= NLM_F_DUMP_FILTERED;
1769 }
1770
1771 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1772 idx = 0;
1773 head = &tgt_net->dev_index_head[h];
1774 hlist_for_each_entry(dev, head, index_hlist) {
1775 if (link_dump_filtered(dev, master_idx, kind_ops))
1776 goto cont;
1777 if (idx < s_idx)
1778 goto cont;
1779 err = rtnl_fill_ifinfo(skb, dev, net,
1780 RTM_NEWLINK,
1781 NETLINK_CB(cb->skb).portid,
1782 cb->nlh->nlmsg_seq, 0,
1783 flags,
1784 ext_filter_mask, 0, NULL, 0,
1785 netnsid);
1786
1787 if (err < 0) {
1788 if (likely(skb->len))
1789 goto out;
1790
1791 goto out_err;
1792 }
1793 cont:
1794 idx++;
1795 }
1796 }
1797 out:
1798 err = skb->len;
1799 out_err:
1800 cb->args[1] = idx;
1801 cb->args[0] = h;
1802 cb->seq = net->dev_base_seq;
1803 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1804 if (netnsid >= 0)
1805 put_net(tgt_net);
1806
1807 return err;
1808 }
1809
1810 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len,
1811 struct netlink_ext_ack *exterr)
1812 {
1813 return nla_parse(tb, IFLA_MAX, head, len, ifla_policy, exterr);
1814 }
1815 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
1816
1817 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
1818 {
1819 struct net *net;
1820 /* Examine the link attributes and figure out which
1821 * network namespace we are talking about.
1822 */
1823 if (tb[IFLA_NET_NS_PID])
1824 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
1825 else if (tb[IFLA_NET_NS_FD])
1826 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
1827 else
1828 net = get_net(src_net);
1829 return net;
1830 }
1831 EXPORT_SYMBOL(rtnl_link_get_net);
1832
1833 /* Figure out which network namespace we are talking about by
1834 * examining the link attributes in the following order:
1835 *
1836 * 1. IFLA_NET_NS_PID
1837 * 2. IFLA_NET_NS_FD
1838 * 3. IFLA_IF_NETNSID
1839 */
1840 static struct net *rtnl_link_get_net_by_nlattr(struct net *src_net,
1841 struct nlattr *tb[])
1842 {
1843 struct net *net;
1844
1845 if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD])
1846 return rtnl_link_get_net(src_net, tb);
1847
1848 if (!tb[IFLA_IF_NETNSID])
1849 return get_net(src_net);
1850
1851 net = get_net_ns_by_id(src_net, nla_get_u32(tb[IFLA_IF_NETNSID]));
1852 if (!net)
1853 return ERR_PTR(-EINVAL);
1854
1855 return net;
1856 }
1857
1858 static struct net *rtnl_link_get_net_capable(const struct sk_buff *skb,
1859 struct net *src_net,
1860 struct nlattr *tb[], int cap)
1861 {
1862 struct net *net;
1863
1864 net = rtnl_link_get_net_by_nlattr(src_net, tb);
1865 if (IS_ERR(net))
1866 return net;
1867
1868 if (!netlink_ns_capable(skb, net->user_ns, cap)) {
1869 put_net(net);
1870 return ERR_PTR(-EPERM);
1871 }
1872
1873 return net;
1874 }
1875
1876 /* Verify that rtnetlink requests do not pass additional properties
1877 * potentially referring to different network namespaces.
1878 */
1879 static int rtnl_ensure_unique_netns(struct nlattr *tb[],
1880 struct netlink_ext_ack *extack,
1881 bool netns_id_only)
1882 {
1883
1884 if (netns_id_only) {
1885 if (!tb[IFLA_NET_NS_PID] && !tb[IFLA_NET_NS_FD])
1886 return 0;
1887
1888 NL_SET_ERR_MSG(extack, "specified netns attribute not supported");
1889 return -EOPNOTSUPP;
1890 }
1891
1892 if (tb[IFLA_IF_NETNSID] && (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]))
1893 goto invalid_attr;
1894
1895 if (tb[IFLA_NET_NS_PID] && (tb[IFLA_IF_NETNSID] || tb[IFLA_NET_NS_FD]))
1896 goto invalid_attr;
1897
1898 if (tb[IFLA_NET_NS_FD] && (tb[IFLA_IF_NETNSID] || tb[IFLA_NET_NS_PID]))
1899 goto invalid_attr;
1900
1901 return 0;
1902
1903 invalid_attr:
1904 NL_SET_ERR_MSG(extack, "multiple netns identifying attributes specified");
1905 return -EINVAL;
1906 }
1907
1908 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
1909 {
1910 if (dev) {
1911 if (tb[IFLA_ADDRESS] &&
1912 nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
1913 return -EINVAL;
1914
1915 if (tb[IFLA_BROADCAST] &&
1916 nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
1917 return -EINVAL;
1918 }
1919
1920 if (tb[IFLA_AF_SPEC]) {
1921 struct nlattr *af;
1922 int rem, err;
1923
1924 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
1925 const struct rtnl_af_ops *af_ops;
1926
1927 rcu_read_lock();
1928 af_ops = rtnl_af_lookup(nla_type(af));
1929 if (!af_ops) {
1930 rcu_read_unlock();
1931 return -EAFNOSUPPORT;
1932 }
1933
1934 if (!af_ops->set_link_af) {
1935 rcu_read_unlock();
1936 return -EOPNOTSUPP;
1937 }
1938
1939 if (af_ops->validate_link_af) {
1940 err = af_ops->validate_link_af(dev, af);
1941 if (err < 0) {
1942 rcu_read_unlock();
1943 return err;
1944 }
1945 }
1946
1947 rcu_read_unlock();
1948 }
1949 }
1950
1951 return 0;
1952 }
1953
1954 static int handle_infiniband_guid(struct net_device *dev, struct ifla_vf_guid *ivt,
1955 int guid_type)
1956 {
1957 const struct net_device_ops *ops = dev->netdev_ops;
1958
1959 return ops->ndo_set_vf_guid(dev, ivt->vf, ivt->guid, guid_type);
1960 }
1961
1962 static int handle_vf_guid(struct net_device *dev, struct ifla_vf_guid *ivt, int guid_type)
1963 {
1964 if (dev->type != ARPHRD_INFINIBAND)
1965 return -EOPNOTSUPP;
1966
1967 return handle_infiniband_guid(dev, ivt, guid_type);
1968 }
1969
1970 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb)
1971 {
1972 const struct net_device_ops *ops = dev->netdev_ops;
1973 int err = -EINVAL;
1974
1975 if (tb[IFLA_VF_MAC]) {
1976 struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]);
1977
1978 err = -EOPNOTSUPP;
1979 if (ops->ndo_set_vf_mac)
1980 err = ops->ndo_set_vf_mac(dev, ivm->vf,
1981 ivm->mac);
1982 if (err < 0)
1983 return err;
1984 }
1985
1986 if (tb[IFLA_VF_VLAN]) {
1987 struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]);
1988
1989 err = -EOPNOTSUPP;
1990 if (ops->ndo_set_vf_vlan)
1991 err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan,
1992 ivv->qos,
1993 htons(ETH_P_8021Q));
1994 if (err < 0)
1995 return err;
1996 }
1997
1998 if (tb[IFLA_VF_VLAN_LIST]) {
1999 struct ifla_vf_vlan_info *ivvl[MAX_VLAN_LIST_LEN];
2000 struct nlattr *attr;
2001 int rem, len = 0;
2002
2003 err = -EOPNOTSUPP;
2004 if (!ops->ndo_set_vf_vlan)
2005 return err;
2006
2007 nla_for_each_nested(attr, tb[IFLA_VF_VLAN_LIST], rem) {
2008 if (nla_type(attr) != IFLA_VF_VLAN_INFO ||
2009 nla_len(attr) < NLA_HDRLEN) {
2010 return -EINVAL;
2011 }
2012 if (len >= MAX_VLAN_LIST_LEN)
2013 return -EOPNOTSUPP;
2014 ivvl[len] = nla_data(attr);
2015
2016 len++;
2017 }
2018 if (len == 0)
2019 return -EINVAL;
2020
2021 err = ops->ndo_set_vf_vlan(dev, ivvl[0]->vf, ivvl[0]->vlan,
2022 ivvl[0]->qos, ivvl[0]->vlan_proto);
2023 if (err < 0)
2024 return err;
2025 }
2026
2027 if (tb[IFLA_VF_TX_RATE]) {
2028 struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]);
2029 struct ifla_vf_info ivf;
2030
2031 err = -EOPNOTSUPP;
2032 if (ops->ndo_get_vf_config)
2033 err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf);
2034 if (err < 0)
2035 return err;
2036
2037 err = -EOPNOTSUPP;
2038 if (ops->ndo_set_vf_rate)
2039 err = ops->ndo_set_vf_rate(dev, ivt->vf,
2040 ivf.min_tx_rate,
2041 ivt->rate);
2042 if (err < 0)
2043 return err;
2044 }
2045
2046 if (tb[IFLA_VF_RATE]) {
2047 struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]);
2048
2049 err = -EOPNOTSUPP;
2050 if (ops->ndo_set_vf_rate)
2051 err = ops->ndo_set_vf_rate(dev, ivt->vf,
2052 ivt->min_tx_rate,
2053 ivt->max_tx_rate);
2054 if (err < 0)
2055 return err;
2056 }
2057
2058 if (tb[IFLA_VF_SPOOFCHK]) {
2059 struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]);
2060
2061 err = -EOPNOTSUPP;
2062 if (ops->ndo_set_vf_spoofchk)
2063 err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
2064 ivs->setting);
2065 if (err < 0)
2066 return err;
2067 }
2068
2069 if (tb[IFLA_VF_LINK_STATE]) {
2070 struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]);
2071
2072 err = -EOPNOTSUPP;
2073 if (ops->ndo_set_vf_link_state)
2074 err = ops->ndo_set_vf_link_state(dev, ivl->vf,
2075 ivl->link_state);
2076 if (err < 0)
2077 return err;
2078 }
2079
2080 if (tb[IFLA_VF_RSS_QUERY_EN]) {
2081 struct ifla_vf_rss_query_en *ivrssq_en;
2082
2083 err = -EOPNOTSUPP;
2084 ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]);
2085 if (ops->ndo_set_vf_rss_query_en)
2086 err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf,
2087 ivrssq_en->setting);
2088 if (err < 0)
2089 return err;
2090 }
2091
2092 if (tb[IFLA_VF_TRUST]) {
2093 struct ifla_vf_trust *ivt = nla_data(tb[IFLA_VF_TRUST]);
2094
2095 err = -EOPNOTSUPP;
2096 if (ops->ndo_set_vf_trust)
2097 err = ops->ndo_set_vf_trust(dev, ivt->vf, ivt->setting);
2098 if (err < 0)
2099 return err;
2100 }
2101
2102 if (tb[IFLA_VF_IB_NODE_GUID]) {
2103 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_NODE_GUID]);
2104
2105 if (!ops->ndo_set_vf_guid)
2106 return -EOPNOTSUPP;
2107
2108 return handle_vf_guid(dev, ivt, IFLA_VF_IB_NODE_GUID);
2109 }
2110
2111 if (tb[IFLA_VF_IB_PORT_GUID]) {
2112 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_PORT_GUID]);
2113
2114 if (!ops->ndo_set_vf_guid)
2115 return -EOPNOTSUPP;
2116
2117 return handle_vf_guid(dev, ivt, IFLA_VF_IB_PORT_GUID);
2118 }
2119
2120 return err;
2121 }
2122
2123 static int do_set_master(struct net_device *dev, int ifindex,
2124 struct netlink_ext_ack *extack)
2125 {
2126 struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
2127 const struct net_device_ops *ops;
2128 int err;
2129
2130 if (upper_dev) {
2131 if (upper_dev->ifindex == ifindex)
2132 return 0;
2133 ops = upper_dev->netdev_ops;
2134 if (ops->ndo_del_slave) {
2135 err = ops->ndo_del_slave(upper_dev, dev);
2136 if (err)
2137 return err;
2138 } else {
2139 return -EOPNOTSUPP;
2140 }
2141 }
2142
2143 if (ifindex) {
2144 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
2145 if (!upper_dev)
2146 return -EINVAL;
2147 ops = upper_dev->netdev_ops;
2148 if (ops->ndo_add_slave) {
2149 err = ops->ndo_add_slave(upper_dev, dev, extack);
2150 if (err)
2151 return err;
2152 } else {
2153 return -EOPNOTSUPP;
2154 }
2155 }
2156 return 0;
2157 }
2158
2159 #define DO_SETLINK_MODIFIED 0x01
2160 /* notify flag means notify + modified. */
2161 #define DO_SETLINK_NOTIFY 0x03
2162 static int do_setlink(const struct sk_buff *skb,
2163 struct net_device *dev, struct ifinfomsg *ifm,
2164 struct netlink_ext_ack *extack,
2165 struct nlattr **tb, char *ifname, int status)
2166 {
2167 const struct net_device_ops *ops = dev->netdev_ops;
2168 int err;
2169
2170 err = validate_linkmsg(dev, tb);
2171 if (err < 0)
2172 return err;
2173
2174 if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD] || tb[IFLA_IF_NETNSID]) {
2175 struct net *net = rtnl_link_get_net_capable(skb, dev_net(dev),
2176 tb, CAP_NET_ADMIN);
2177 if (IS_ERR(net)) {
2178 err = PTR_ERR(net);
2179 goto errout;
2180 }
2181
2182 err = dev_change_net_namespace(dev, net, ifname);
2183 put_net(net);
2184 if (err)
2185 goto errout;
2186 status |= DO_SETLINK_MODIFIED;
2187 }
2188
2189 if (tb[IFLA_MAP]) {
2190 struct rtnl_link_ifmap *u_map;
2191 struct ifmap k_map;
2192
2193 if (!ops->ndo_set_config) {
2194 err = -EOPNOTSUPP;
2195 goto errout;
2196 }
2197
2198 if (!netif_device_present(dev)) {
2199 err = -ENODEV;
2200 goto errout;
2201 }
2202
2203 u_map = nla_data(tb[IFLA_MAP]);
2204 k_map.mem_start = (unsigned long) u_map->mem_start;
2205 k_map.mem_end = (unsigned long) u_map->mem_end;
2206 k_map.base_addr = (unsigned short) u_map->base_addr;
2207 k_map.irq = (unsigned char) u_map->irq;
2208 k_map.dma = (unsigned char) u_map->dma;
2209 k_map.port = (unsigned char) u_map->port;
2210
2211 err = ops->ndo_set_config(dev, &k_map);
2212 if (err < 0)
2213 goto errout;
2214
2215 status |= DO_SETLINK_NOTIFY;
2216 }
2217
2218 if (tb[IFLA_ADDRESS]) {
2219 struct sockaddr *sa;
2220 int len;
2221
2222 len = sizeof(sa_family_t) + max_t(size_t, dev->addr_len,
2223 sizeof(*sa));
2224 sa = kmalloc(len, GFP_KERNEL);
2225 if (!sa) {
2226 err = -ENOMEM;
2227 goto errout;
2228 }
2229 sa->sa_family = dev->type;
2230 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
2231 dev->addr_len);
2232 err = dev_set_mac_address(dev, sa);
2233 kfree(sa);
2234 if (err)
2235 goto errout;
2236 status |= DO_SETLINK_MODIFIED;
2237 }
2238
2239 if (tb[IFLA_MTU]) {
2240 err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
2241 if (err < 0)
2242 goto errout;
2243 status |= DO_SETLINK_MODIFIED;
2244 }
2245
2246 if (tb[IFLA_GROUP]) {
2247 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2248 status |= DO_SETLINK_NOTIFY;
2249 }
2250
2251 /*
2252 * Interface selected by interface index but interface
2253 * name provided implies that a name change has been
2254 * requested.
2255 */
2256 if (ifm->ifi_index > 0 && ifname[0]) {
2257 err = dev_change_name(dev, ifname);
2258 if (err < 0)
2259 goto errout;
2260 status |= DO_SETLINK_MODIFIED;
2261 }
2262
2263 if (tb[IFLA_IFALIAS]) {
2264 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
2265 nla_len(tb[IFLA_IFALIAS]));
2266 if (err < 0)
2267 goto errout;
2268 status |= DO_SETLINK_NOTIFY;
2269 }
2270
2271 if (tb[IFLA_BROADCAST]) {
2272 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
2273 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
2274 }
2275
2276 if (ifm->ifi_flags || ifm->ifi_change) {
2277 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
2278 if (err < 0)
2279 goto errout;
2280 }
2281
2282 if (tb[IFLA_MASTER]) {
2283 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]), extack);
2284 if (err)
2285 goto errout;
2286 status |= DO_SETLINK_MODIFIED;
2287 }
2288
2289 if (tb[IFLA_CARRIER]) {
2290 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
2291 if (err)
2292 goto errout;
2293 status |= DO_SETLINK_MODIFIED;
2294 }
2295
2296 if (tb[IFLA_TXQLEN]) {
2297 unsigned int value = nla_get_u32(tb[IFLA_TXQLEN]);
2298 unsigned int orig_len = dev->tx_queue_len;
2299
2300 if (dev->tx_queue_len ^ value) {
2301 dev->tx_queue_len = value;
2302 err = call_netdevice_notifiers(
2303 NETDEV_CHANGE_TX_QUEUE_LEN, dev);
2304 err = notifier_to_errno(err);
2305 if (err) {
2306 dev->tx_queue_len = orig_len;
2307 goto errout;
2308 }
2309 status |= DO_SETLINK_MODIFIED;
2310 }
2311 }
2312
2313 if (tb[IFLA_OPERSTATE])
2314 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2315
2316 if (tb[IFLA_LINKMODE]) {
2317 unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
2318
2319 write_lock_bh(&dev_base_lock);
2320 if (dev->link_mode ^ value)
2321 status |= DO_SETLINK_NOTIFY;
2322 dev->link_mode = value;
2323 write_unlock_bh(&dev_base_lock);
2324 }
2325
2326 if (tb[IFLA_VFINFO_LIST]) {
2327 struct nlattr *vfinfo[IFLA_VF_MAX + 1];
2328 struct nlattr *attr;
2329 int rem;
2330
2331 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
2332 if (nla_type(attr) != IFLA_VF_INFO ||
2333 nla_len(attr) < NLA_HDRLEN) {
2334 err = -EINVAL;
2335 goto errout;
2336 }
2337 err = nla_parse_nested(vfinfo, IFLA_VF_MAX, attr,
2338 ifla_vf_policy, NULL);
2339 if (err < 0)
2340 goto errout;
2341 err = do_setvfinfo(dev, vfinfo);
2342 if (err < 0)
2343 goto errout;
2344 status |= DO_SETLINK_NOTIFY;
2345 }
2346 }
2347 err = 0;
2348
2349 if (tb[IFLA_VF_PORTS]) {
2350 struct nlattr *port[IFLA_PORT_MAX+1];
2351 struct nlattr *attr;
2352 int vf;
2353 int rem;
2354
2355 err = -EOPNOTSUPP;
2356 if (!ops->ndo_set_vf_port)
2357 goto errout;
2358
2359 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
2360 if (nla_type(attr) != IFLA_VF_PORT ||
2361 nla_len(attr) < NLA_HDRLEN) {
2362 err = -EINVAL;
2363 goto errout;
2364 }
2365 err = nla_parse_nested(port, IFLA_PORT_MAX, attr,
2366 ifla_port_policy, NULL);
2367 if (err < 0)
2368 goto errout;
2369 if (!port[IFLA_PORT_VF]) {
2370 err = -EOPNOTSUPP;
2371 goto errout;
2372 }
2373 vf = nla_get_u32(port[IFLA_PORT_VF]);
2374 err = ops->ndo_set_vf_port(dev, vf, port);
2375 if (err < 0)
2376 goto errout;
2377 status |= DO_SETLINK_NOTIFY;
2378 }
2379 }
2380 err = 0;
2381
2382 if (tb[IFLA_PORT_SELF]) {
2383 struct nlattr *port[IFLA_PORT_MAX+1];
2384
2385 err = nla_parse_nested(port, IFLA_PORT_MAX,
2386 tb[IFLA_PORT_SELF], ifla_port_policy,
2387 NULL);
2388 if (err < 0)
2389 goto errout;
2390
2391 err = -EOPNOTSUPP;
2392 if (ops->ndo_set_vf_port)
2393 err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
2394 if (err < 0)
2395 goto errout;
2396 status |= DO_SETLINK_NOTIFY;
2397 }
2398
2399 if (tb[IFLA_AF_SPEC]) {
2400 struct nlattr *af;
2401 int rem;
2402
2403 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2404 const struct rtnl_af_ops *af_ops;
2405
2406 rcu_read_lock();
2407
2408 BUG_ON(!(af_ops = rtnl_af_lookup(nla_type(af))));
2409
2410 err = af_ops->set_link_af(dev, af);
2411 if (err < 0) {
2412 rcu_read_unlock();
2413 goto errout;
2414 }
2415
2416 rcu_read_unlock();
2417 status |= DO_SETLINK_NOTIFY;
2418 }
2419 }
2420 err = 0;
2421
2422 if (tb[IFLA_PROTO_DOWN]) {
2423 err = dev_change_proto_down(dev,
2424 nla_get_u8(tb[IFLA_PROTO_DOWN]));
2425 if (err)
2426 goto errout;
2427 status |= DO_SETLINK_NOTIFY;
2428 }
2429
2430 if (tb[IFLA_XDP]) {
2431 struct nlattr *xdp[IFLA_XDP_MAX + 1];
2432 u32 xdp_flags = 0;
2433
2434 err = nla_parse_nested(xdp, IFLA_XDP_MAX, tb[IFLA_XDP],
2435 ifla_xdp_policy, NULL);
2436 if (err < 0)
2437 goto errout;
2438
2439 if (xdp[IFLA_XDP_ATTACHED] || xdp[IFLA_XDP_PROG_ID]) {
2440 err = -EINVAL;
2441 goto errout;
2442 }
2443
2444 if (xdp[IFLA_XDP_FLAGS]) {
2445 xdp_flags = nla_get_u32(xdp[IFLA_XDP_FLAGS]);
2446 if (xdp_flags & ~XDP_FLAGS_MASK) {
2447 err = -EINVAL;
2448 goto errout;
2449 }
2450 if (hweight32(xdp_flags & XDP_FLAGS_MODES) > 1) {
2451 err = -EINVAL;
2452 goto errout;
2453 }
2454 }
2455
2456 if (xdp[IFLA_XDP_FD]) {
2457 err = dev_change_xdp_fd(dev, extack,
2458 nla_get_s32(xdp[IFLA_XDP_FD]),
2459 xdp_flags);
2460 if (err)
2461 goto errout;
2462 status |= DO_SETLINK_NOTIFY;
2463 }
2464 }
2465
2466 errout:
2467 if (status & DO_SETLINK_MODIFIED) {
2468 if ((status & DO_SETLINK_NOTIFY) == DO_SETLINK_NOTIFY)
2469 netdev_state_change(dev);
2470
2471 if (err < 0)
2472 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",
2473 dev->name);
2474 }
2475
2476 return err;
2477 }
2478
2479 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
2480 struct netlink_ext_ack *extack)
2481 {
2482 struct net *net = sock_net(skb->sk);
2483 struct ifinfomsg *ifm;
2484 struct net_device *dev;
2485 int err;
2486 struct nlattr *tb[IFLA_MAX+1];
2487 char ifname[IFNAMSIZ];
2488
2489 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy,
2490 extack);
2491 if (err < 0)
2492 goto errout;
2493
2494 err = rtnl_ensure_unique_netns(tb, extack, false);
2495 if (err < 0)
2496 goto errout;
2497
2498 if (tb[IFLA_IFNAME])
2499 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2500 else
2501 ifname[0] = '\0';
2502
2503 err = -EINVAL;
2504 ifm = nlmsg_data(nlh);
2505 if (ifm->ifi_index > 0)
2506 dev = __dev_get_by_index(net, ifm->ifi_index);
2507 else if (tb[IFLA_IFNAME])
2508 dev = __dev_get_by_name(net, ifname);
2509 else
2510 goto errout;
2511
2512 if (dev == NULL) {
2513 err = -ENODEV;
2514 goto errout;
2515 }
2516
2517 err = do_setlink(skb, dev, ifm, extack, tb, ifname, 0);
2518 errout:
2519 return err;
2520 }
2521
2522 static int rtnl_group_dellink(const struct net *net, int group)
2523 {
2524 struct net_device *dev, *aux;
2525 LIST_HEAD(list_kill);
2526 bool found = false;
2527
2528 if (!group)
2529 return -EPERM;
2530
2531 for_each_netdev(net, dev) {
2532 if (dev->group == group) {
2533 const struct rtnl_link_ops *ops;
2534
2535 found = true;
2536 ops = dev->rtnl_link_ops;
2537 if (!ops || !ops->dellink)
2538 return -EOPNOTSUPP;
2539 }
2540 }
2541
2542 if (!found)
2543 return -ENODEV;
2544
2545 for_each_netdev_safe(net, dev, aux) {
2546 if (dev->group == group) {
2547 const struct rtnl_link_ops *ops;
2548
2549 ops = dev->rtnl_link_ops;
2550 ops->dellink(dev, &list_kill);
2551 }
2552 }
2553 unregister_netdevice_many(&list_kill);
2554
2555 return 0;
2556 }
2557
2558 int rtnl_delete_link(struct net_device *dev)
2559 {
2560 const struct rtnl_link_ops *ops;
2561 LIST_HEAD(list_kill);
2562
2563 ops = dev->rtnl_link_ops;
2564 if (!ops || !ops->dellink)
2565 return -EOPNOTSUPP;
2566
2567 ops->dellink(dev, &list_kill);
2568 unregister_netdevice_many(&list_kill);
2569
2570 return 0;
2571 }
2572 EXPORT_SYMBOL_GPL(rtnl_delete_link);
2573
2574 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
2575 struct netlink_ext_ack *extack)
2576 {
2577 struct net *net = sock_net(skb->sk);
2578 struct net *tgt_net = net;
2579 struct net_device *dev = NULL;
2580 struct ifinfomsg *ifm;
2581 char ifname[IFNAMSIZ];
2582 struct nlattr *tb[IFLA_MAX+1];
2583 int err;
2584 int netnsid = -1;
2585
2586 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
2587 if (err < 0)
2588 return err;
2589
2590 err = rtnl_ensure_unique_netns(tb, extack, true);
2591 if (err < 0)
2592 return err;
2593
2594 if (tb[IFLA_IFNAME])
2595 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2596
2597 if (tb[IFLA_IF_NETNSID]) {
2598 netnsid = nla_get_s32(tb[IFLA_IF_NETNSID]);
2599 tgt_net = get_target_net(NETLINK_CB(skb).sk, netnsid);
2600 if (IS_ERR(tgt_net))
2601 return PTR_ERR(tgt_net);
2602 }
2603
2604 err = -EINVAL;
2605 ifm = nlmsg_data(nlh);
2606 if (ifm->ifi_index > 0)
2607 dev = __dev_get_by_index(tgt_net, ifm->ifi_index);
2608 else if (tb[IFLA_IFNAME])
2609 dev = __dev_get_by_name(tgt_net, ifname);
2610 else if (tb[IFLA_GROUP])
2611 err = rtnl_group_dellink(tgt_net, nla_get_u32(tb[IFLA_GROUP]));
2612 else
2613 goto out;
2614
2615 if (!dev) {
2616 if (tb[IFLA_IFNAME] || ifm->ifi_index > 0)
2617 err = -ENODEV;
2618
2619 goto out;
2620 }
2621
2622 err = rtnl_delete_link(dev);
2623
2624 out:
2625 if (netnsid >= 0)
2626 put_net(tgt_net);
2627
2628 return err;
2629 }
2630
2631 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
2632 {
2633 unsigned int old_flags;
2634 int err;
2635
2636 old_flags = dev->flags;
2637 if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
2638 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm));
2639 if (err < 0)
2640 return err;
2641 }
2642
2643 dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
2644
2645 __dev_notify_flags(dev, old_flags, ~0U);
2646 return 0;
2647 }
2648 EXPORT_SYMBOL(rtnl_configure_link);
2649
2650 struct net_device *rtnl_create_link(struct net *net,
2651 const char *ifname, unsigned char name_assign_type,
2652 const struct rtnl_link_ops *ops, struct nlattr *tb[])
2653 {
2654 struct net_device *dev;
2655 unsigned int num_tx_queues = 1;
2656 unsigned int num_rx_queues = 1;
2657
2658 if (tb[IFLA_NUM_TX_QUEUES])
2659 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
2660 else if (ops->get_num_tx_queues)
2661 num_tx_queues = ops->get_num_tx_queues();
2662
2663 if (tb[IFLA_NUM_RX_QUEUES])
2664 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
2665 else if (ops->get_num_rx_queues)
2666 num_rx_queues = ops->get_num_rx_queues();
2667
2668 dev = alloc_netdev_mqs(ops->priv_size, ifname, name_assign_type,
2669 ops->setup, num_tx_queues, num_rx_queues);
2670 if (!dev)
2671 return ERR_PTR(-ENOMEM);
2672
2673 dev_net_set(dev, net);
2674 dev->rtnl_link_ops = ops;
2675 dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
2676
2677 if (tb[IFLA_MTU])
2678 dev->mtu = nla_get_u32(tb[IFLA_MTU]);
2679 if (tb[IFLA_ADDRESS]) {
2680 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
2681 nla_len(tb[IFLA_ADDRESS]));
2682 dev->addr_assign_type = NET_ADDR_SET;
2683 }
2684 if (tb[IFLA_BROADCAST])
2685 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
2686 nla_len(tb[IFLA_BROADCAST]));
2687 if (tb[IFLA_TXQLEN])
2688 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
2689 if (tb[IFLA_OPERSTATE])
2690 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2691 if (tb[IFLA_LINKMODE])
2692 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
2693 if (tb[IFLA_GROUP])
2694 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2695
2696 return dev;
2697 }
2698 EXPORT_SYMBOL(rtnl_create_link);
2699
2700 static int rtnl_group_changelink(const struct sk_buff *skb,
2701 struct net *net, int group,
2702 struct ifinfomsg *ifm,
2703 struct netlink_ext_ack *extack,
2704 struct nlattr **tb)
2705 {
2706 struct net_device *dev, *aux;
2707 int err;
2708
2709 for_each_netdev_safe(net, dev, aux) {
2710 if (dev->group == group) {
2711 err = do_setlink(skb, dev, ifm, extack, tb, NULL, 0);
2712 if (err < 0)
2713 return err;
2714 }
2715 }
2716
2717 return 0;
2718 }
2719
2720 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
2721 struct netlink_ext_ack *extack)
2722 {
2723 struct net *net = sock_net(skb->sk);
2724 const struct rtnl_link_ops *ops;
2725 const struct rtnl_link_ops *m_ops = NULL;
2726 struct net_device *dev;
2727 struct net_device *master_dev = NULL;
2728 struct ifinfomsg *ifm;
2729 char kind[MODULE_NAME_LEN];
2730 char ifname[IFNAMSIZ];
2731 struct nlattr *tb[IFLA_MAX+1];
2732 struct nlattr *linkinfo[IFLA_INFO_MAX+1];
2733 unsigned char name_assign_type = NET_NAME_USER;
2734 int err;
2735
2736 #ifdef CONFIG_MODULES
2737 replay:
2738 #endif
2739 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
2740 if (err < 0)
2741 return err;
2742
2743 err = rtnl_ensure_unique_netns(tb, extack, false);
2744 if (err < 0)
2745 return err;
2746
2747 if (tb[IFLA_IFNAME])
2748 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2749 else
2750 ifname[0] = '\0';
2751
2752 ifm = nlmsg_data(nlh);
2753 if (ifm->ifi_index > 0)
2754 dev = __dev_get_by_index(net, ifm->ifi_index);
2755 else {
2756 if (ifname[0])
2757 dev = __dev_get_by_name(net, ifname);
2758 else
2759 dev = NULL;
2760 }
2761
2762 if (dev) {
2763 master_dev = netdev_master_upper_dev_get(dev);
2764 if (master_dev)
2765 m_ops = master_dev->rtnl_link_ops;
2766 }
2767
2768 err = validate_linkmsg(dev, tb);
2769 if (err < 0)
2770 return err;
2771
2772 if (tb[IFLA_LINKINFO]) {
2773 err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
2774 tb[IFLA_LINKINFO], ifla_info_policy,
2775 NULL);
2776 if (err < 0)
2777 return err;
2778 } else
2779 memset(linkinfo, 0, sizeof(linkinfo));
2780
2781 if (linkinfo[IFLA_INFO_KIND]) {
2782 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
2783 ops = rtnl_link_ops_get(kind);
2784 } else {
2785 kind[0] = '\0';
2786 ops = NULL;
2787 }
2788
2789 if (1) {
2790 struct nlattr *attr[ops ? ops->maxtype + 1 : 1];
2791 struct nlattr *slave_attr[m_ops ? m_ops->slave_maxtype + 1 : 1];
2792 struct nlattr **data = NULL;
2793 struct nlattr **slave_data = NULL;
2794 struct net *dest_net, *link_net = NULL;
2795
2796 if (ops) {
2797 if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
2798 err = nla_parse_nested(attr, ops->maxtype,
2799 linkinfo[IFLA_INFO_DATA],
2800 ops->policy, NULL);
2801 if (err < 0)
2802 return err;
2803 data = attr;
2804 }
2805 if (ops->validate) {
2806 err = ops->validate(tb, data, extack);
2807 if (err < 0)
2808 return err;
2809 }
2810 }
2811
2812 if (m_ops) {
2813 if (m_ops->slave_maxtype &&
2814 linkinfo[IFLA_INFO_SLAVE_DATA]) {
2815 err = nla_parse_nested(slave_attr,
2816 m_ops->slave_maxtype,
2817 linkinfo[IFLA_INFO_SLAVE_DATA],
2818 m_ops->slave_policy,
2819 NULL);
2820 if (err < 0)
2821 return err;
2822 slave_data = slave_attr;
2823 }
2824 }
2825
2826 if (dev) {
2827 int status = 0;
2828
2829 if (nlh->nlmsg_flags & NLM_F_EXCL)
2830 return -EEXIST;
2831 if (nlh->nlmsg_flags & NLM_F_REPLACE)
2832 return -EOPNOTSUPP;
2833
2834 if (linkinfo[IFLA_INFO_DATA]) {
2835 if (!ops || ops != dev->rtnl_link_ops ||
2836 !ops->changelink)
2837 return -EOPNOTSUPP;
2838
2839 err = ops->changelink(dev, tb, data, extack);
2840 if (err < 0)
2841 return err;
2842 status |= DO_SETLINK_NOTIFY;
2843 }
2844
2845 if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
2846 if (!m_ops || !m_ops->slave_changelink)
2847 return -EOPNOTSUPP;
2848
2849 err = m_ops->slave_changelink(master_dev, dev,
2850 tb, slave_data,
2851 extack);
2852 if (err < 0)
2853 return err;
2854 status |= DO_SETLINK_NOTIFY;
2855 }
2856
2857 return do_setlink(skb, dev, ifm, extack, tb, ifname,
2858 status);
2859 }
2860
2861 if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
2862 if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
2863 return rtnl_group_changelink(skb, net,
2864 nla_get_u32(tb[IFLA_GROUP]),
2865 ifm, extack, tb);
2866 return -ENODEV;
2867 }
2868
2869 if (tb[IFLA_MAP] || tb[IFLA_PROTINFO])
2870 return -EOPNOTSUPP;
2871
2872 if (!ops) {
2873 #ifdef CONFIG_MODULES
2874 if (kind[0]) {
2875 __rtnl_unlock();
2876 request_module("rtnl-link-%s", kind);
2877 rtnl_lock();
2878 ops = rtnl_link_ops_get(kind);
2879 if (ops)
2880 goto replay;
2881 }
2882 #endif
2883 return -EOPNOTSUPP;
2884 }
2885
2886 if (!ops->setup)
2887 return -EOPNOTSUPP;
2888
2889 if (!ifname[0]) {
2890 snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
2891 name_assign_type = NET_NAME_ENUM;
2892 }
2893
2894 dest_net = rtnl_link_get_net_capable(skb, net, tb, CAP_NET_ADMIN);
2895 if (IS_ERR(dest_net))
2896 return PTR_ERR(dest_net);
2897
2898 if (tb[IFLA_LINK_NETNSID]) {
2899 int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
2900
2901 link_net = get_net_ns_by_id(dest_net, id);
2902 if (!link_net) {
2903 err = -EINVAL;
2904 goto out;
2905 }
2906 err = -EPERM;
2907 if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
2908 goto out;
2909 }
2910
2911 dev = rtnl_create_link(link_net ? : dest_net, ifname,
2912 name_assign_type, ops, tb);
2913 if (IS_ERR(dev)) {
2914 err = PTR_ERR(dev);
2915 goto out;
2916 }
2917
2918 dev->ifindex = ifm->ifi_index;
2919
2920 if (ops->newlink) {
2921 err = ops->newlink(link_net ? : net, dev, tb, data,
2922 extack);
2923 /* Drivers should call free_netdev() in ->destructor
2924 * and unregister it on failure after registration
2925 * so that device could be finally freed in rtnl_unlock.
2926 */
2927 if (err < 0) {
2928 /* If device is not registered at all, free it now */
2929 if (dev->reg_state == NETREG_UNINITIALIZED)
2930 free_netdev(dev);
2931 goto out;
2932 }
2933 } else {
2934 err = register_netdevice(dev);
2935 if (err < 0) {
2936 free_netdev(dev);
2937 goto out;
2938 }
2939 }
2940 err = rtnl_configure_link(dev, ifm);
2941 if (err < 0)
2942 goto out_unregister;
2943 if (link_net) {
2944 err = dev_change_net_namespace(dev, dest_net, ifname);
2945 if (err < 0)
2946 goto out_unregister;
2947 }
2948 if (tb[IFLA_MASTER]) {
2949 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]),
2950 extack);
2951 if (err)
2952 goto out_unregister;
2953 }
2954 out:
2955 if (link_net)
2956 put_net(link_net);
2957 put_net(dest_net);
2958 return err;
2959 out_unregister:
2960 if (ops->newlink) {
2961 LIST_HEAD(list_kill);
2962
2963 ops->dellink(dev, &list_kill);
2964 unregister_netdevice_many(&list_kill);
2965 } else {
2966 unregister_netdevice(dev);
2967 }
2968 goto out;
2969 }
2970 }
2971
2972 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr *nlh,
2973 struct netlink_ext_ack *extack)
2974 {
2975 struct net *net = sock_net(skb->sk);
2976 struct net *tgt_net = net;
2977 struct ifinfomsg *ifm;
2978 char ifname[IFNAMSIZ];
2979 struct nlattr *tb[IFLA_MAX+1];
2980 struct net_device *dev = NULL;
2981 struct sk_buff *nskb;
2982 int netnsid = -1;
2983 int err;
2984 u32 ext_filter_mask = 0;
2985
2986 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
2987 if (err < 0)
2988 return err;
2989
2990 err = rtnl_ensure_unique_netns(tb, extack, true);
2991 if (err < 0)
2992 return err;
2993
2994 if (tb[IFLA_IF_NETNSID]) {
2995 netnsid = nla_get_s32(tb[IFLA_IF_NETNSID]);
2996 tgt_net = get_target_net(NETLINK_CB(skb).sk, netnsid);
2997 if (IS_ERR(tgt_net))
2998 return PTR_ERR(tgt_net);
2999 }
3000
3001 if (tb[IFLA_IFNAME])
3002 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
3003
3004 if (tb[IFLA_EXT_MASK])
3005 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
3006
3007 err = -EINVAL;
3008 ifm = nlmsg_data(nlh);
3009 if (ifm->ifi_index > 0)
3010 dev = __dev_get_by_index(tgt_net, ifm->ifi_index);
3011 else if (tb[IFLA_IFNAME])
3012 dev = __dev_get_by_name(tgt_net, ifname);
3013 else
3014 goto out;
3015
3016 err = -ENODEV;
3017 if (dev == NULL)
3018 goto out;
3019
3020 err = -ENOBUFS;
3021 nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
3022 if (nskb == NULL)
3023 goto out;
3024
3025 err = rtnl_fill_ifinfo(nskb, dev, net,
3026 RTM_NEWLINK, NETLINK_CB(skb).portid,
3027 nlh->nlmsg_seq, 0, 0, ext_filter_mask,
3028 0, NULL, 0, netnsid);
3029 if (err < 0) {
3030 /* -EMSGSIZE implies BUG in if_nlmsg_size */
3031 WARN_ON(err == -EMSGSIZE);
3032 kfree_skb(nskb);
3033 } else
3034 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
3035 out:
3036 if (netnsid >= 0)
3037 put_net(tgt_net);
3038
3039 return err;
3040 }
3041
3042 static u16 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
3043 {
3044 struct net *net = sock_net(skb->sk);
3045 struct net_device *dev;
3046 struct nlattr *tb[IFLA_MAX+1];
3047 u32 ext_filter_mask = 0;
3048 u16 min_ifinfo_dump_size = 0;
3049 int hdrlen;
3050
3051 /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
3052 hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
3053 sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
3054
3055 if (nlmsg_parse(nlh, hdrlen, tb, IFLA_MAX, ifla_policy, NULL) >= 0) {
3056 if (tb[IFLA_EXT_MASK])
3057 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
3058 }
3059
3060 if (!ext_filter_mask)
3061 return NLMSG_GOODSIZE;
3062 /*
3063 * traverse the list of net devices and compute the minimum
3064 * buffer size based upon the filter mask.
3065 */
3066 rcu_read_lock();
3067 for_each_netdev_rcu(net, dev) {
3068 min_ifinfo_dump_size = max_t(u16, min_ifinfo_dump_size,
3069 if_nlmsg_size(dev,
3070 ext_filter_mask));
3071 }
3072 rcu_read_unlock();
3073
3074 return nlmsg_total_size(min_ifinfo_dump_size);
3075 }
3076
3077 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
3078 {
3079 int idx;
3080 int s_idx = cb->family;
3081
3082 if (s_idx == 0)
3083 s_idx = 1;
3084
3085 for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
3086 int type = cb->nlh->nlmsg_type-RTM_BASE;
3087 struct rtnl_link *handlers;
3088 rtnl_dumpit_func dumpit;
3089
3090 if (idx < s_idx || idx == PF_PACKET)
3091 continue;
3092
3093 handlers = rtnl_dereference(rtnl_msg_handlers[idx]);
3094 if (!handlers)
3095 continue;
3096
3097 dumpit = READ_ONCE(handlers[type].dumpit);
3098 if (!dumpit)
3099 continue;
3100
3101 if (idx > s_idx) {
3102 memset(&cb->args[0], 0, sizeof(cb->args));
3103 cb->prev_seq = 0;
3104 cb->seq = 0;
3105 }
3106 if (dumpit(skb, cb))
3107 break;
3108 }
3109 cb->family = idx;
3110
3111 return skb->len;
3112 }
3113
3114 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
3115 unsigned int change,
3116 u32 event, gfp_t flags, int *new_nsid,
3117 int new_ifindex)
3118 {
3119 struct net *net = dev_net(dev);
3120 struct sk_buff *skb;
3121 int err = -ENOBUFS;
3122 size_t if_info_size;
3123
3124 skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
3125 if (skb == NULL)
3126 goto errout;
3127
3128 err = rtnl_fill_ifinfo(skb, dev, dev_net(dev),
3129 type, 0, 0, change, 0, 0, event,
3130 new_nsid, new_ifindex, -1);
3131 if (err < 0) {
3132 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
3133 WARN_ON(err == -EMSGSIZE);
3134 kfree_skb(skb);
3135 goto errout;
3136 }
3137 return skb;
3138 errout:
3139 if (err < 0)
3140 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
3141 return NULL;
3142 }
3143
3144 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags)
3145 {
3146 struct net *net = dev_net(dev);
3147
3148 rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
3149 }
3150
3151 static void rtmsg_ifinfo_event(int type, struct net_device *dev,
3152 unsigned int change, u32 event,
3153 gfp_t flags, int *new_nsid, int new_ifindex)
3154 {
3155 struct sk_buff *skb;
3156
3157 if (dev->reg_state != NETREG_REGISTERED)
3158 return;
3159
3160 skb = rtmsg_ifinfo_build_skb(type, dev, change, event, flags, new_nsid,
3161 new_ifindex);
3162 if (skb)
3163 rtmsg_ifinfo_send(skb, dev, flags);
3164 }
3165
3166 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
3167 gfp_t flags)
3168 {
3169 rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
3170 NULL, 0);
3171 }
3172
3173 void rtmsg_ifinfo_newnet(int type, struct net_device *dev, unsigned int change,
3174 gfp_t flags, int *new_nsid, int new_ifindex)
3175 {
3176 rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
3177 new_nsid, new_ifindex);
3178 }
3179
3180 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
3181 struct net_device *dev,
3182 u8 *addr, u16 vid, u32 pid, u32 seq,
3183 int type, unsigned int flags,
3184 int nlflags, u16 ndm_state)
3185 {
3186 struct nlmsghdr *nlh;
3187 struct ndmsg *ndm;
3188
3189 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
3190 if (!nlh)
3191 return -EMSGSIZE;
3192
3193 ndm = nlmsg_data(nlh);
3194 ndm->ndm_family = AF_BRIDGE;
3195 ndm->ndm_pad1 = 0;
3196 ndm->ndm_pad2 = 0;
3197 ndm->ndm_flags = flags;
3198 ndm->ndm_type = 0;
3199 ndm->ndm_ifindex = dev->ifindex;
3200 ndm->ndm_state = ndm_state;
3201
3202 if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
3203 goto nla_put_failure;
3204 if (vid)
3205 if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
3206 goto nla_put_failure;
3207
3208 nlmsg_end(skb, nlh);
3209 return 0;
3210
3211 nla_put_failure:
3212 nlmsg_cancel(skb, nlh);
3213 return -EMSGSIZE;
3214 }
3215
3216 static inline size_t rtnl_fdb_nlmsg_size(void)
3217 {
3218 return NLMSG_ALIGN(sizeof(struct ndmsg)) +
3219 nla_total_size(ETH_ALEN) + /* NDA_LLADDR */
3220 nla_total_size(sizeof(u16)) + /* NDA_VLAN */
3221 0;
3222 }
3223
3224 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type,
3225 u16 ndm_state)
3226 {
3227 struct net *net = dev_net(dev);
3228 struct sk_buff *skb;
3229 int err = -ENOBUFS;
3230
3231 skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
3232 if (!skb)
3233 goto errout;
3234
3235 err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
3236 0, 0, type, NTF_SELF, 0, ndm_state);
3237 if (err < 0) {
3238 kfree_skb(skb);
3239 goto errout;
3240 }
3241
3242 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
3243 return;
3244 errout:
3245 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
3246 }
3247
3248 /**
3249 * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
3250 */
3251 int ndo_dflt_fdb_add(struct ndmsg *ndm,
3252 struct nlattr *tb[],
3253 struct net_device *dev,
3254 const unsigned char *addr, u16 vid,
3255 u16 flags)
3256 {
3257 int err = -EINVAL;
3258
3259 /* If aging addresses are supported device will need to
3260 * implement its own handler for this.
3261 */
3262 if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
3263 pr_info("%s: FDB only supports static addresses\n", dev->name);
3264 return err;
3265 }
3266
3267 if (vid) {
3268 pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
3269 return err;
3270 }
3271
3272 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
3273 err = dev_uc_add_excl(dev, addr);
3274 else if (is_multicast_ether_addr(addr))
3275 err = dev_mc_add_excl(dev, addr);
3276
3277 /* Only return duplicate errors if NLM_F_EXCL is set */
3278 if (err == -EEXIST && !(flags & NLM_F_EXCL))
3279 err = 0;
3280
3281 return err;
3282 }
3283 EXPORT_SYMBOL(ndo_dflt_fdb_add);
3284
3285 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid,
3286 struct netlink_ext_ack *extack)
3287 {
3288 u16 vid = 0;
3289
3290 if (vlan_attr) {
3291 if (nla_len(vlan_attr) != sizeof(u16)) {
3292 NL_SET_ERR_MSG(extack, "invalid vlan attribute size");
3293 return -EINVAL;
3294 }
3295
3296 vid = nla_get_u16(vlan_attr);
3297
3298 if (!vid || vid >= VLAN_VID_MASK) {
3299 NL_SET_ERR_MSG(extack, "invalid vlan id");
3300 return -EINVAL;
3301 }
3302 }
3303 *p_vid = vid;
3304 return 0;
3305 }
3306
3307 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh,
3308 struct netlink_ext_ack *extack)
3309 {
3310 struct net *net = sock_net(skb->sk);
3311 struct ndmsg *ndm;
3312 struct nlattr *tb[NDA_MAX+1];
3313 struct net_device *dev;
3314 u8 *addr;
3315 u16 vid;
3316 int err;
3317
3318 err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL, extack);
3319 if (err < 0)
3320 return err;
3321
3322 ndm = nlmsg_data(nlh);
3323 if (ndm->ndm_ifindex == 0) {
3324 NL_SET_ERR_MSG(extack, "invalid ifindex");
3325 return -EINVAL;
3326 }
3327
3328 dev = __dev_get_by_index(net, ndm->ndm_ifindex);
3329 if (dev == NULL) {
3330 NL_SET_ERR_MSG(extack, "unknown ifindex");
3331 return -ENODEV;
3332 }
3333
3334 if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
3335 NL_SET_ERR_MSG(extack, "invalid address");
3336 return -EINVAL;
3337 }
3338
3339 addr = nla_data(tb[NDA_LLADDR]);
3340
3341 err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
3342 if (err)
3343 return err;
3344
3345 err = -EOPNOTSUPP;
3346
3347 /* Support fdb on master device the net/bridge default case */
3348 if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
3349 (dev->priv_flags & IFF_BRIDGE_PORT)) {
3350 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3351 const struct net_device_ops *ops = br_dev->netdev_ops;
3352
3353 err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
3354 nlh->nlmsg_flags);
3355 if (err)
3356 goto out;
3357 else
3358 ndm->ndm_flags &= ~NTF_MASTER;
3359 }
3360
3361 /* Embedded bridge, macvlan, and any other device support */
3362 if ((ndm->ndm_flags & NTF_SELF)) {
3363 if (dev->netdev_ops->ndo_fdb_add)
3364 err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
3365 vid,
3366 nlh->nlmsg_flags);
3367 else
3368 err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
3369 nlh->nlmsg_flags);
3370
3371 if (!err) {
3372 rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH,
3373 ndm->ndm_state);
3374 ndm->ndm_flags &= ~NTF_SELF;
3375 }
3376 }
3377 out:
3378 return err;
3379 }
3380
3381 /**
3382 * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
3383 */
3384 int ndo_dflt_fdb_del(struct ndmsg *ndm,
3385 struct nlattr *tb[],
3386 struct net_device *dev,
3387 const unsigned char *addr, u16 vid)
3388 {
3389 int err = -EINVAL;
3390
3391 /* If aging addresses are supported device will need to
3392 * implement its own handler for this.
3393 */
3394 if (!(ndm->ndm_state & NUD_PERMANENT)) {
3395 pr_info("%s: FDB only supports static addresses\n", dev->name);
3396 return err;
3397 }
3398
3399 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
3400 err = dev_uc_del(dev, addr);
3401 else if (is_multicast_ether_addr(addr))
3402 err = dev_mc_del(dev, addr);
3403
3404 return err;
3405 }
3406 EXPORT_SYMBOL(ndo_dflt_fdb_del);
3407
3408 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh,
3409 struct netlink_ext_ack *extack)
3410 {
3411 struct net *net = sock_net(skb->sk);
3412 struct ndmsg *ndm;
3413 struct nlattr *tb[NDA_MAX+1];
3414 struct net_device *dev;
3415 int err = -EINVAL;
3416 __u8 *addr;
3417 u16 vid;
3418
3419 if (!netlink_capable(skb, CAP_NET_ADMIN))
3420 return -EPERM;
3421
3422 err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, NULL, extack);
3423 if (err < 0)
3424 return err;
3425
3426 ndm = nlmsg_data(nlh);
3427 if (ndm->ndm_ifindex == 0) {
3428 NL_SET_ERR_MSG(extack, "invalid ifindex");
3429 return -EINVAL;
3430 }
3431
3432 dev = __dev_get_by_index(net, ndm->ndm_ifindex);
3433 if (dev == NULL) {
3434 NL_SET_ERR_MSG(extack, "unknown ifindex");
3435 return -ENODEV;
3436 }
3437
3438 if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
3439 NL_SET_ERR_MSG(extack, "invalid address");
3440 return -EINVAL;
3441 }
3442
3443 addr = nla_data(tb[NDA_LLADDR]);
3444
3445 err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
3446 if (err)
3447 return err;
3448
3449 err = -EOPNOTSUPP;
3450
3451 /* Support fdb on master device the net/bridge default case */
3452 if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
3453 (dev->priv_flags & IFF_BRIDGE_PORT)) {
3454 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3455 const struct net_device_ops *ops = br_dev->netdev_ops;
3456
3457 if (ops->ndo_fdb_del)
3458 err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid);
3459
3460 if (err)
3461 goto out;
3462 else
3463 ndm->ndm_flags &= ~NTF_MASTER;
3464 }
3465
3466 /* Embedded bridge, macvlan, and any other device support */
3467 if (ndm->ndm_flags & NTF_SELF) {
3468 if (dev->netdev_ops->ndo_fdb_del)
3469 err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr,
3470 vid);
3471 else
3472 err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
3473
3474 if (!err) {
3475 rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH,
3476 ndm->ndm_state);
3477 ndm->ndm_flags &= ~NTF_SELF;
3478 }
3479 }
3480 out:
3481 return err;
3482 }
3483
3484 static int nlmsg_populate_fdb(struct sk_buff *skb,
3485 struct netlink_callback *cb,
3486 struct net_device *dev,
3487 int *idx,
3488 struct netdev_hw_addr_list *list)
3489 {
3490 struct netdev_hw_addr *ha;
3491 int err;
3492 u32 portid, seq;
3493
3494 portid = NETLINK_CB(cb->skb).portid;
3495 seq = cb->nlh->nlmsg_seq;
3496
3497 list_for_each_entry(ha, &list->list, list) {
3498 if (*idx < cb->args[2])
3499 goto skip;
3500
3501 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
3502 portid, seq,
3503 RTM_NEWNEIGH, NTF_SELF,
3504 NLM_F_MULTI, NUD_PERMANENT);
3505 if (err < 0)
3506 return err;
3507 skip:
3508 *idx += 1;
3509 }
3510 return 0;
3511 }
3512
3513 /**
3514 * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
3515 * @nlh: netlink message header
3516 * @dev: netdevice
3517 *
3518 * Default netdevice operation to dump the existing unicast address list.
3519 * Returns number of addresses from list put in skb.
3520 */
3521 int ndo_dflt_fdb_dump(struct sk_buff *skb,
3522 struct netlink_callback *cb,
3523 struct net_device *dev,
3524 struct net_device *filter_dev,
3525 int *idx)
3526 {
3527 int err;
3528
3529 netif_addr_lock_bh(dev);
3530 err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->uc);
3531 if (err)
3532 goto out;
3533 err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->mc);
3534 out:
3535 netif_addr_unlock_bh(dev);
3536 return err;
3537 }
3538 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
3539
3540 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
3541 {
3542 struct net_device *dev;
3543 struct nlattr *tb[IFLA_MAX+1];
3544 struct net_device *br_dev = NULL;
3545 const struct net_device_ops *ops = NULL;
3546 const struct net_device_ops *cops = NULL;
3547 struct ifinfomsg *ifm = nlmsg_data(cb->nlh);
3548 struct net *net = sock_net(skb->sk);
3549 struct hlist_head *head;
3550 int brport_idx = 0;
3551 int br_idx = 0;
3552 int h, s_h;
3553 int idx = 0, s_idx;
3554 int err = 0;
3555 int fidx = 0;
3556
3557 err = nlmsg_parse(cb->nlh, sizeof(struct ifinfomsg), tb,
3558 IFLA_MAX, ifla_policy, NULL);
3559 if (err < 0) {
3560 return -EINVAL;
3561 } else if (err == 0) {
3562 if (tb[IFLA_MASTER])
3563 br_idx = nla_get_u32(tb[IFLA_MASTER]);
3564 }
3565
3566 brport_idx = ifm->ifi_index;
3567
3568 if (br_idx) {
3569 br_dev = __dev_get_by_index(net, br_idx);
3570 if (!br_dev)
3571 return -ENODEV;
3572
3573 ops = br_dev->netdev_ops;
3574 }
3575
3576 s_h = cb->args[0];
3577 s_idx = cb->args[1];
3578
3579 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3580 idx = 0;
3581 head = &net->dev_index_head[h];
3582 hlist_for_each_entry(dev, head, index_hlist) {
3583
3584 if (brport_idx && (dev->ifindex != brport_idx))
3585 continue;
3586
3587 if (!br_idx) { /* user did not specify a specific bridge */
3588 if (dev->priv_flags & IFF_BRIDGE_PORT) {
3589 br_dev = netdev_master_upper_dev_get(dev);
3590 cops = br_dev->netdev_ops;
3591 }
3592 } else {
3593 if (dev != br_dev &&
3594 !(dev->priv_flags & IFF_BRIDGE_PORT))
3595 continue;
3596
3597 if (br_dev != netdev_master_upper_dev_get(dev) &&
3598 !(dev->priv_flags & IFF_EBRIDGE))
3599 continue;
3600 cops = ops;
3601 }
3602
3603 if (idx < s_idx)
3604 goto cont;
3605
3606 if (dev->priv_flags & IFF_BRIDGE_PORT) {
3607 if (cops && cops->ndo_fdb_dump) {
3608 err = cops->ndo_fdb_dump(skb, cb,
3609 br_dev, dev,
3610 &fidx);
3611 if (err == -EMSGSIZE)
3612 goto out;
3613 }
3614 }
3615
3616 if (dev->netdev_ops->ndo_fdb_dump)
3617 err = dev->netdev_ops->ndo_fdb_dump(skb, cb,
3618 dev, NULL,
3619 &fidx);
3620 else
3621 err = ndo_dflt_fdb_dump(skb, cb, dev, NULL,
3622 &fidx);
3623 if (err == -EMSGSIZE)
3624 goto out;
3625
3626 cops = NULL;
3627
3628 /* reset fdb offset to 0 for rest of the interfaces */
3629 cb->args[2] = 0;
3630 fidx = 0;
3631 cont:
3632 idx++;
3633 }
3634 }
3635
3636 out:
3637 cb->args[0] = h;
3638 cb->args[1] = idx;
3639 cb->args[2] = fidx;
3640
3641 return skb->len;
3642 }
3643
3644 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
3645 unsigned int attrnum, unsigned int flag)
3646 {
3647 if (mask & flag)
3648 return nla_put_u8(skb, attrnum, !!(flags & flag));
3649 return 0;
3650 }
3651
3652 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
3653 struct net_device *dev, u16 mode,
3654 u32 flags, u32 mask, int nlflags,
3655 u32 filter_mask,
3656 int (*vlan_fill)(struct sk_buff *skb,
3657 struct net_device *dev,
3658 u32 filter_mask))
3659 {
3660 struct nlmsghdr *nlh;
3661 struct ifinfomsg *ifm;
3662 struct nlattr *br_afspec;
3663 struct nlattr *protinfo;
3664 u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
3665 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3666 int err = 0;
3667
3668 nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
3669 if (nlh == NULL)
3670 return -EMSGSIZE;
3671
3672 ifm = nlmsg_data(nlh);
3673 ifm->ifi_family = AF_BRIDGE;
3674 ifm->__ifi_pad = 0;
3675 ifm->ifi_type = dev->type;
3676 ifm->ifi_index = dev->ifindex;
3677 ifm->ifi_flags = dev_get_flags(dev);
3678 ifm->ifi_change = 0;
3679
3680
3681 if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
3682 nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
3683 nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
3684 (br_dev &&
3685 nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
3686 (dev->addr_len &&
3687 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
3688 (dev->ifindex != dev_get_iflink(dev) &&
3689 nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
3690 goto nla_put_failure;
3691
3692 br_afspec = nla_nest_start(skb, IFLA_AF_SPEC);
3693 if (!br_afspec)
3694 goto nla_put_failure;
3695
3696 if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
3697 nla_nest_cancel(skb, br_afspec);
3698 goto nla_put_failure;
3699 }
3700
3701 if (mode != BRIDGE_MODE_UNDEF) {
3702 if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
3703 nla_nest_cancel(skb, br_afspec);
3704 goto nla_put_failure;
3705 }
3706 }
3707 if (vlan_fill) {
3708 err = vlan_fill(skb, dev, filter_mask);
3709 if (err) {
3710 nla_nest_cancel(skb, br_afspec);
3711 goto nla_put_failure;
3712 }
3713 }
3714 nla_nest_end(skb, br_afspec);
3715
3716 protinfo = nla_nest_start(skb, IFLA_PROTINFO | NLA_F_NESTED);
3717 if (!protinfo)
3718 goto nla_put_failure;
3719
3720 if (brport_nla_put_flag(skb, flags, mask,
3721 IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
3722 brport_nla_put_flag(skb, flags, mask,
3723 IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
3724 brport_nla_put_flag(skb, flags, mask,
3725 IFLA_BRPORT_FAST_LEAVE,
3726 BR_MULTICAST_FAST_LEAVE) ||
3727 brport_nla_put_flag(skb, flags, mask,
3728 IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
3729 brport_nla_put_flag(skb, flags, mask,
3730 IFLA_BRPORT_LEARNING, BR_LEARNING) ||
3731 brport_nla_put_flag(skb, flags, mask,
3732 IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
3733 brport_nla_put_flag(skb, flags, mask,
3734 IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
3735 brport_nla_put_flag(skb, flags, mask,
3736 IFLA_BRPORT_PROXYARP, BR_PROXYARP)) {
3737 nla_nest_cancel(skb, protinfo);
3738 goto nla_put_failure;
3739 }
3740
3741 nla_nest_end(skb, protinfo);
3742
3743 nlmsg_end(skb, nlh);
3744 return 0;
3745 nla_put_failure:
3746 nlmsg_cancel(skb, nlh);
3747 return err ? err : -EMSGSIZE;
3748 }
3749 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink);
3750
3751 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
3752 {
3753 struct net *net = sock_net(skb->sk);
3754 struct net_device *dev;
3755 int idx = 0;
3756 u32 portid = NETLINK_CB(cb->skb).portid;
3757 u32 seq = cb->nlh->nlmsg_seq;
3758 u32 filter_mask = 0;
3759 int err;
3760
3761 if (nlmsg_len(cb->nlh) > sizeof(struct ifinfomsg)) {
3762 struct nlattr *extfilt;
3763
3764 extfilt = nlmsg_find_attr(cb->nlh, sizeof(struct ifinfomsg),
3765 IFLA_EXT_MASK);
3766 if (extfilt) {
3767 if (nla_len(extfilt) < sizeof(filter_mask))
3768 return -EINVAL;
3769
3770 filter_mask = nla_get_u32(extfilt);
3771 }
3772 }
3773
3774 rcu_read_lock();
3775 for_each_netdev_rcu(net, dev) {
3776 const struct net_device_ops *ops = dev->netdev_ops;
3777 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3778
3779 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
3780 if (idx >= cb->args[0]) {
3781 err = br_dev->netdev_ops->ndo_bridge_getlink(
3782 skb, portid, seq, dev,
3783 filter_mask, NLM_F_MULTI);
3784 if (err < 0 && err != -EOPNOTSUPP) {
3785 if (likely(skb->len))
3786 break;
3787
3788 goto out_err;
3789 }
3790 }
3791 idx++;
3792 }
3793
3794 if (ops->ndo_bridge_getlink) {
3795 if (idx >= cb->args[0]) {
3796 err = ops->ndo_bridge_getlink(skb, portid,
3797 seq, dev,
3798 filter_mask,
3799 NLM_F_MULTI);
3800 if (err < 0 && err != -EOPNOTSUPP) {
3801 if (likely(skb->len))
3802 break;
3803
3804 goto out_err;
3805 }
3806 }
3807 idx++;
3808 }
3809 }
3810 err = skb->len;
3811 out_err:
3812 rcu_read_unlock();
3813 cb->args[0] = idx;
3814
3815 return err;
3816 }
3817
3818 static inline size_t bridge_nlmsg_size(void)
3819 {
3820 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3821 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3822 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3823 + nla_total_size(sizeof(u32)) /* IFLA_MASTER */
3824 + nla_total_size(sizeof(u32)) /* IFLA_MTU */
3825 + nla_total_size(sizeof(u32)) /* IFLA_LINK */
3826 + nla_total_size(sizeof(u32)) /* IFLA_OPERSTATE */
3827 + nla_total_size(sizeof(u8)) /* IFLA_PROTINFO */
3828 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
3829 + nla_total_size(sizeof(u16)) /* IFLA_BRIDGE_FLAGS */
3830 + nla_total_size(sizeof(u16)); /* IFLA_BRIDGE_MODE */
3831 }
3832
3833 static int rtnl_bridge_notify(struct net_device *dev)
3834 {
3835 struct net *net = dev_net(dev);
3836 struct sk_buff *skb;
3837 int err = -EOPNOTSUPP;
3838
3839 if (!dev->netdev_ops->ndo_bridge_getlink)
3840 return 0;
3841
3842 skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
3843 if (!skb) {
3844 err = -ENOMEM;
3845 goto errout;
3846 }
3847
3848 err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
3849 if (err < 0)
3850 goto errout;
3851
3852 if (!skb->len)
3853 goto errout;
3854
3855 rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
3856 return 0;
3857 errout:
3858 WARN_ON(err == -EMSGSIZE);
3859 kfree_skb(skb);
3860 if (err)
3861 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
3862 return err;
3863 }
3864
3865 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3866 struct netlink_ext_ack *extack)
3867 {
3868 struct net *net = sock_net(skb->sk);
3869 struct ifinfomsg *ifm;
3870 struct net_device *dev;
3871 struct nlattr *br_spec, *attr = NULL;
3872 int rem, err = -EOPNOTSUPP;
3873 u16 flags = 0;
3874 bool have_flags = false;
3875
3876 if (nlmsg_len(nlh) < sizeof(*ifm))
3877 return -EINVAL;
3878
3879 ifm = nlmsg_data(nlh);
3880 if (ifm->ifi_family != AF_BRIDGE)
3881 return -EPFNOSUPPORT;
3882
3883 dev = __dev_get_by_index(net, ifm->ifi_index);
3884 if (!dev) {
3885 NL_SET_ERR_MSG(extack, "unknown ifindex");
3886 return -ENODEV;
3887 }
3888
3889 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3890 if (br_spec) {
3891 nla_for_each_nested(attr, br_spec, rem) {
3892 if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3893 if (nla_len(attr) < sizeof(flags))
3894 return -EINVAL;
3895
3896 have_flags = true;
3897 flags = nla_get_u16(attr);
3898 break;
3899 }
3900 }
3901 }
3902
3903 if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3904 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3905
3906 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
3907 err = -EOPNOTSUPP;
3908 goto out;
3909 }
3910
3911 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags);
3912 if (err)
3913 goto out;
3914
3915 flags &= ~BRIDGE_FLAGS_MASTER;
3916 }
3917
3918 if ((flags & BRIDGE_FLAGS_SELF)) {
3919 if (!dev->netdev_ops->ndo_bridge_setlink)
3920 err = -EOPNOTSUPP;
3921 else
3922 err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
3923 flags);
3924 if (!err) {
3925 flags &= ~BRIDGE_FLAGS_SELF;
3926
3927 /* Generate event to notify upper layer of bridge
3928 * change
3929 */
3930 err = rtnl_bridge_notify(dev);
3931 }
3932 }
3933
3934 if (have_flags)
3935 memcpy(nla_data(attr), &flags, sizeof(flags));
3936 out:
3937 return err;
3938 }
3939
3940 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
3941 struct netlink_ext_ack *extack)
3942 {
3943 struct net *net = sock_net(skb->sk);
3944 struct ifinfomsg *ifm;
3945 struct net_device *dev;
3946 struct nlattr *br_spec, *attr = NULL;
3947 int rem, err = -EOPNOTSUPP;
3948 u16 flags = 0;
3949 bool have_flags = false;
3950
3951 if (nlmsg_len(nlh) < sizeof(*ifm))
3952 return -EINVAL;
3953
3954 ifm = nlmsg_data(nlh);
3955 if (ifm->ifi_family != AF_BRIDGE)
3956 return -EPFNOSUPPORT;
3957
3958 dev = __dev_get_by_index(net, ifm->ifi_index);
3959 if (!dev) {
3960 NL_SET_ERR_MSG(extack, "unknown ifindex");
3961 return -ENODEV;
3962 }
3963
3964 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3965 if (br_spec) {
3966 nla_for_each_nested(attr, br_spec, rem) {
3967 if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
3968 if (nla_len(attr) < sizeof(flags))
3969 return -EINVAL;
3970
3971 have_flags = true;
3972 flags = nla_get_u16(attr);
3973 break;
3974 }
3975 }
3976 }
3977
3978 if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
3979 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
3980
3981 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
3982 err = -EOPNOTSUPP;
3983 goto out;
3984 }
3985
3986 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
3987 if (err)
3988 goto out;
3989
3990 flags &= ~BRIDGE_FLAGS_MASTER;
3991 }
3992
3993 if ((flags & BRIDGE_FLAGS_SELF)) {
3994 if (!dev->netdev_ops->ndo_bridge_dellink)
3995 err = -EOPNOTSUPP;
3996 else
3997 err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
3998 flags);
3999
4000 if (!err) {
4001 flags &= ~BRIDGE_FLAGS_SELF;
4002
4003 /* Generate event to notify upper layer of bridge
4004 * change
4005 */
4006 err = rtnl_bridge_notify(dev);
4007 }
4008 }
4009
4010 if (have_flags)
4011 memcpy(nla_data(attr), &flags, sizeof(flags));
4012 out:
4013 return err;
4014 }
4015
4016 static bool stats_attr_valid(unsigned int mask, int attrid, int idxattr)
4017 {
4018 return (mask & IFLA_STATS_FILTER_BIT(attrid)) &&
4019 (!idxattr || idxattr == attrid);
4020 }
4021
4022 #define IFLA_OFFLOAD_XSTATS_FIRST (IFLA_OFFLOAD_XSTATS_UNSPEC + 1)
4023 static int rtnl_get_offload_stats_attr_size(int attr_id)
4024 {
4025 switch (attr_id) {
4026 case IFLA_OFFLOAD_XSTATS_CPU_HIT:
4027 return sizeof(struct rtnl_link_stats64);
4028 }
4029
4030 return 0;
4031 }
4032
4033 static int rtnl_get_offload_stats(struct sk_buff *skb, struct net_device *dev,
4034 int *prividx)
4035 {
4036 struct nlattr *attr = NULL;
4037 int attr_id, size;
4038 void *attr_data;
4039 int err;
4040
4041 if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
4042 dev->netdev_ops->ndo_get_offload_stats))
4043 return -ENODATA;
4044
4045 for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
4046 attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
4047 if (attr_id < *prividx)
4048 continue;
4049
4050 size = rtnl_get_offload_stats_attr_size(attr_id);
4051 if (!size)
4052 continue;
4053
4054 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
4055 continue;
4056
4057 attr = nla_reserve_64bit(skb, attr_id, size,
4058 IFLA_OFFLOAD_XSTATS_UNSPEC);
4059 if (!attr)
4060 goto nla_put_failure;
4061
4062 attr_data = nla_data(attr);
4063 memset(attr_data, 0, size);
4064 err = dev->netdev_ops->ndo_get_offload_stats(attr_id, dev,
4065 attr_data);
4066 if (err)
4067 goto get_offload_stats_failure;
4068 }
4069
4070 if (!attr)
4071 return -ENODATA;
4072
4073 *prividx = 0;
4074 return 0;
4075
4076 nla_put_failure:
4077 err = -EMSGSIZE;
4078 get_offload_stats_failure:
4079 *prividx = attr_id;
4080 return err;
4081 }
4082
4083 static int rtnl_get_offload_stats_size(const struct net_device *dev)
4084 {
4085 int nla_size = 0;
4086 int attr_id;
4087 int size;
4088
4089 if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
4090 dev->netdev_ops->ndo_get_offload_stats))
4091 return 0;
4092
4093 for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
4094 attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
4095 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
4096 continue;
4097 size = rtnl_get_offload_stats_attr_size(attr_id);
4098 nla_size += nla_total_size_64bit(size);
4099 }
4100
4101 if (nla_size != 0)
4102 nla_size += nla_total_size(0);
4103
4104 return nla_size;
4105 }
4106
4107 static int rtnl_fill_statsinfo(struct sk_buff *skb, struct net_device *dev,
4108 int type, u32 pid, u32 seq, u32 change,
4109 unsigned int flags, unsigned int filter_mask,
4110 int *idxattr, int *prividx)
4111 {
4112 struct if_stats_msg *ifsm;
4113 struct nlmsghdr *nlh;
4114 struct nlattr *attr;
4115 int s_prividx = *prividx;
4116 int err;
4117
4118 ASSERT_RTNL();
4119
4120 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifsm), flags);
4121 if (!nlh)
4122 return -EMSGSIZE;
4123
4124 ifsm = nlmsg_data(nlh);
4125 ifsm->family = PF_UNSPEC;
4126 ifsm->pad1 = 0;
4127 ifsm->pad2 = 0;
4128 ifsm->ifindex = dev->ifindex;
4129 ifsm->filter_mask = filter_mask;
4130
4131 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, *idxattr)) {
4132 struct rtnl_link_stats64 *sp;
4133
4134 attr = nla_reserve_64bit(skb, IFLA_STATS_LINK_64,
4135 sizeof(struct rtnl_link_stats64),
4136 IFLA_STATS_UNSPEC);
4137 if (!attr)
4138 goto nla_put_failure;
4139
4140 sp = nla_data(attr);
4141 dev_get_stats(dev, sp);
4142 }
4143
4144 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, *idxattr)) {
4145 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
4146
4147 if (ops && ops->fill_linkxstats) {
4148 *idxattr = IFLA_STATS_LINK_XSTATS;
4149 attr = nla_nest_start(skb,
4150 IFLA_STATS_LINK_XSTATS);
4151 if (!attr)
4152 goto nla_put_failure;
4153
4154 err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
4155 nla_nest_end(skb, attr);
4156 if (err)
4157 goto nla_put_failure;
4158 *idxattr = 0;
4159 }
4160 }
4161
4162 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE,
4163 *idxattr)) {
4164 const struct rtnl_link_ops *ops = NULL;
4165 const struct net_device *master;
4166
4167 master = netdev_master_upper_dev_get(dev);
4168 if (master)
4169 ops = master->rtnl_link_ops;
4170 if (ops && ops->fill_linkxstats) {
4171 *idxattr = IFLA_STATS_LINK_XSTATS_SLAVE;
4172 attr = nla_nest_start(skb,
4173 IFLA_STATS_LINK_XSTATS_SLAVE);
4174 if (!attr)
4175 goto nla_put_failure;
4176
4177 err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
4178 nla_nest_end(skb, attr);
4179 if (err)
4180 goto nla_put_failure;
4181 *idxattr = 0;
4182 }
4183 }
4184
4185 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS,
4186 *idxattr)) {
4187 *idxattr = IFLA_STATS_LINK_OFFLOAD_XSTATS;
4188 attr = nla_nest_start(skb, IFLA_STATS_LINK_OFFLOAD_XSTATS);
4189 if (!attr)
4190 goto nla_put_failure;
4191
4192 err = rtnl_get_offload_stats(skb, dev, prividx);
4193 if (err == -ENODATA)
4194 nla_nest_cancel(skb, attr);
4195 else
4196 nla_nest_end(skb, attr);
4197
4198 if (err && err != -ENODATA)
4199 goto nla_put_failure;
4200 *idxattr = 0;
4201 }
4202
4203 if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, *idxattr)) {
4204 struct rtnl_af_ops *af_ops;
4205
4206 *idxattr = IFLA_STATS_AF_SPEC;
4207 attr = nla_nest_start(skb, IFLA_STATS_AF_SPEC);
4208 if (!attr)
4209 goto nla_put_failure;
4210
4211 rcu_read_lock();
4212 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
4213 if (af_ops->fill_stats_af) {
4214 struct nlattr *af;
4215 int err;
4216
4217 af = nla_nest_start(skb, af_ops->family);
4218 if (!af) {
4219 rcu_read_unlock();
4220 goto nla_put_failure;
4221 }
4222 err = af_ops->fill_stats_af(skb, dev);
4223
4224 if (err == -ENODATA) {
4225 nla_nest_cancel(skb, af);
4226 } else if (err < 0) {
4227 rcu_read_unlock();
4228 goto nla_put_failure;
4229 }
4230
4231 nla_nest_end(skb, af);
4232 }
4233 }
4234 rcu_read_unlock();
4235
4236 nla_nest_end(skb, attr);
4237
4238 *idxattr = 0;
4239 }
4240
4241 nlmsg_end(skb, nlh);
4242
4243 return 0;
4244
4245 nla_put_failure:
4246 /* not a multi message or no progress mean a real error */
4247 if (!(flags & NLM_F_MULTI) || s_prividx == *prividx)
4248 nlmsg_cancel(skb, nlh);
4249 else
4250 nlmsg_end(skb, nlh);
4251
4252 return -EMSGSIZE;
4253 }
4254
4255 static size_t if_nlmsg_stats_size(const struct net_device *dev,
4256 u32 filter_mask)
4257 {
4258 size_t size = 0;
4259
4260 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, 0))
4261 size += nla_total_size_64bit(sizeof(struct rtnl_link_stats64));
4262
4263 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, 0)) {
4264 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
4265 int attr = IFLA_STATS_LINK_XSTATS;
4266
4267 if (ops && ops->get_linkxstats_size) {
4268 size += nla_total_size(ops->get_linkxstats_size(dev,
4269 attr));
4270 /* for IFLA_STATS_LINK_XSTATS */
4271 size += nla_total_size(0);
4272 }
4273 }
4274
4275 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE, 0)) {
4276 struct net_device *_dev = (struct net_device *)dev;
4277 const struct rtnl_link_ops *ops = NULL;
4278 const struct net_device *master;
4279
4280 /* netdev_master_upper_dev_get can't take const */
4281 master = netdev_master_upper_dev_get(_dev);
4282 if (master)
4283 ops = master->rtnl_link_ops;
4284 if (ops && ops->get_linkxstats_size) {
4285 int attr = IFLA_STATS_LINK_XSTATS_SLAVE;
4286
4287 size += nla_total_size(ops->get_linkxstats_size(dev,
4288 attr));
4289 /* for IFLA_STATS_LINK_XSTATS_SLAVE */
4290 size += nla_total_size(0);
4291 }
4292 }
4293
4294 if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS, 0))
4295 size += rtnl_get_offload_stats_size(dev);
4296
4297 if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, 0)) {
4298 struct rtnl_af_ops *af_ops;
4299
4300 /* for IFLA_STATS_AF_SPEC */
4301 size += nla_total_size(0);
4302
4303 rcu_read_lock();
4304 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
4305 if (af_ops->get_stats_af_size) {
4306 size += nla_total_size(
4307 af_ops->get_stats_af_size(dev));
4308
4309 /* for AF_* */
4310 size += nla_total_size(0);
4311 }
4312 }
4313 rcu_read_unlock();
4314 }
4315
4316 return size;
4317 }
4318
4319 static int rtnl_stats_get(struct sk_buff *skb, struct nlmsghdr *nlh,
4320 struct netlink_ext_ack *extack)
4321 {
4322 struct net *net = sock_net(skb->sk);
4323 struct net_device *dev = NULL;
4324 int idxattr = 0, prividx = 0;
4325 struct if_stats_msg *ifsm;
4326 struct sk_buff *nskb;
4327 u32 filter_mask;
4328 int err;
4329
4330 if (nlmsg_len(nlh) < sizeof(*ifsm))
4331 return -EINVAL;
4332
4333 ifsm = nlmsg_data(nlh);
4334 if (ifsm->ifindex > 0)
4335 dev = __dev_get_by_index(net, ifsm->ifindex);
4336 else
4337 return -EINVAL;
4338
4339 if (!dev)
4340 return -ENODEV;
4341
4342 filter_mask = ifsm->filter_mask;
4343 if (!filter_mask)
4344 return -EINVAL;
4345
4346 nskb = nlmsg_new(if_nlmsg_stats_size(dev, filter_mask), GFP_KERNEL);
4347 if (!nskb)
4348 return -ENOBUFS;
4349
4350 err = rtnl_fill_statsinfo(nskb, dev, RTM_NEWSTATS,
4351 NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
4352 0, filter_mask, &idxattr, &prividx);
4353 if (err < 0) {
4354 /* -EMSGSIZE implies BUG in if_nlmsg_stats_size */
4355 WARN_ON(err == -EMSGSIZE);
4356 kfree_skb(nskb);
4357 } else {
4358 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
4359 }
4360
4361 return err;
4362 }
4363
4364 static int rtnl_stats_dump(struct sk_buff *skb, struct netlink_callback *cb)
4365 {
4366 int h, s_h, err, s_idx, s_idxattr, s_prividx;
4367 struct net *net = sock_net(skb->sk);
4368 unsigned int flags = NLM_F_MULTI;
4369 struct if_stats_msg *ifsm;
4370 struct hlist_head *head;
4371 struct net_device *dev;
4372 u32 filter_mask = 0;
4373 int idx = 0;
4374
4375 s_h = cb->args[0];
4376 s_idx = cb->args[1];
4377 s_idxattr = cb->args[2];
4378 s_prividx = cb->args[3];
4379
4380 cb->seq = net->dev_base_seq;
4381
4382 if (nlmsg_len(cb->nlh) < sizeof(*ifsm))
4383 return -EINVAL;
4384
4385 ifsm = nlmsg_data(cb->nlh);
4386 filter_mask = ifsm->filter_mask;
4387 if (!filter_mask)
4388 return -EINVAL;
4389
4390 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4391 idx = 0;
4392 head = &net->dev_index_head[h];
4393 hlist_for_each_entry(dev, head, index_hlist) {
4394 if (idx < s_idx)
4395 goto cont;
4396 err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS,
4397 NETLINK_CB(cb->skb).portid,
4398 cb->nlh->nlmsg_seq, 0,
4399 flags, filter_mask,
4400 &s_idxattr, &s_prividx);
4401 /* If we ran out of room on the first message,
4402 * we're in trouble
4403 */
4404 WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
4405
4406 if (err < 0)
4407 goto out;
4408 s_prividx = 0;
4409 s_idxattr = 0;
4410 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
4411 cont:
4412 idx++;
4413 }
4414 }
4415 out:
4416 cb->args[3] = s_prividx;
4417 cb->args[2] = s_idxattr;
4418 cb->args[1] = idx;
4419 cb->args[0] = h;
4420
4421 return skb->len;
4422 }
4423
4424 /* Process one rtnetlink message. */
4425
4426 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh,
4427 struct netlink_ext_ack *extack)
4428 {
4429 struct net *net = sock_net(skb->sk);
4430 struct rtnl_link *handlers;
4431 int err = -EOPNOTSUPP;
4432 rtnl_doit_func doit;
4433 unsigned int flags;
4434 int kind;
4435 int family;
4436 int type;
4437
4438 type = nlh->nlmsg_type;
4439 if (type > RTM_MAX)
4440 return -EOPNOTSUPP;
4441
4442 type -= RTM_BASE;
4443
4444 /* All the messages must have at least 1 byte length */
4445 if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
4446 return 0;
4447
4448 family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
4449 kind = type&3;
4450
4451 if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
4452 return -EPERM;
4453
4454 if (family >= ARRAY_SIZE(rtnl_msg_handlers))
4455 family = PF_UNSPEC;
4456
4457 rcu_read_lock();
4458 handlers = rcu_dereference(rtnl_msg_handlers[family]);
4459 if (!handlers) {
4460 family = PF_UNSPEC;
4461 handlers = rcu_dereference(rtnl_msg_handlers[family]);
4462 }
4463
4464 if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
4465 struct sock *rtnl;
4466 rtnl_dumpit_func dumpit;
4467 u16 min_dump_alloc = 0;
4468
4469 dumpit = READ_ONCE(handlers[type].dumpit);
4470 if (!dumpit) {
4471 family = PF_UNSPEC;
4472 handlers = rcu_dereference(rtnl_msg_handlers[PF_UNSPEC]);
4473 if (!handlers)
4474 goto err_unlock;
4475
4476 dumpit = READ_ONCE(handlers[type].dumpit);
4477 if (!dumpit)
4478 goto err_unlock;
4479 }
4480
4481 refcount_inc(&rtnl_msg_handlers_ref[family]);
4482
4483 if (type == RTM_GETLINK - RTM_BASE)
4484 min_dump_alloc = rtnl_calcit(skb, nlh);
4485
4486 rcu_read_unlock();
4487
4488 rtnl = net->rtnl;
4489 {
4490 struct netlink_dump_control c = {
4491 .dump = dumpit,
4492 .min_dump_alloc = min_dump_alloc,
4493 };
4494 err = netlink_dump_start(rtnl, skb, nlh, &c);
4495 }
4496 refcount_dec(&rtnl_msg_handlers_ref[family]);
4497 return err;
4498 }
4499
4500 doit = READ_ONCE(handlers[type].doit);
4501 if (!doit) {
4502 family = PF_UNSPEC;
4503 handlers = rcu_dereference(rtnl_msg_handlers[family]);
4504 }
4505
4506 flags = READ_ONCE(handlers[type].flags);
4507 if (flags & RTNL_FLAG_DOIT_UNLOCKED) {
4508 refcount_inc(&rtnl_msg_handlers_ref[family]);
4509 doit = READ_ONCE(handlers[type].doit);
4510 rcu_read_unlock();
4511 if (doit)
4512 err = doit(skb, nlh, extack);
4513 refcount_dec(&rtnl_msg_handlers_ref[family]);
4514 return err;
4515 }
4516
4517 rcu_read_unlock();
4518
4519 rtnl_lock();
4520 handlers = rtnl_dereference(rtnl_msg_handlers[family]);
4521 if (handlers) {
4522 doit = READ_ONCE(handlers[type].doit);
4523 if (doit)
4524 err = doit(skb, nlh, extack);
4525 }
4526 rtnl_unlock();
4527 return err;
4528
4529 err_unlock:
4530 rcu_read_unlock();
4531 return -EOPNOTSUPP;
4532 }
4533
4534 static void rtnetlink_rcv(struct sk_buff *skb)
4535 {
4536 netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
4537 }
4538
4539 static int rtnetlink_bind(struct net *net, int group)
4540 {
4541 switch (group) {
4542 case RTNLGRP_IPV4_MROUTE_R:
4543 case RTNLGRP_IPV6_MROUTE_R:
4544 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
4545 return -EPERM;
4546 break;
4547 }
4548 return 0;
4549 }
4550
4551 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
4552 {
4553 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
4554
4555 switch (event) {
4556 case NETDEV_REBOOT:
4557 case NETDEV_CHANGEMTU:
4558 case NETDEV_CHANGEADDR:
4559 case NETDEV_CHANGENAME:
4560 case NETDEV_FEAT_CHANGE:
4561 case NETDEV_BONDING_FAILOVER:
4562 case NETDEV_POST_TYPE_CHANGE:
4563 case NETDEV_NOTIFY_PEERS:
4564 case NETDEV_CHANGEUPPER:
4565 case NETDEV_RESEND_IGMP:
4566 case NETDEV_CHANGEINFODATA:
4567 case NETDEV_CHANGELOWERSTATE:
4568 case NETDEV_CHANGE_TX_QUEUE_LEN:
4569 rtmsg_ifinfo_event(RTM_NEWLINK, dev, 0, rtnl_get_event(event),
4570 GFP_KERNEL, NULL, 0);
4571 break;
4572 default:
4573 break;
4574 }
4575 return NOTIFY_DONE;
4576 }
4577
4578 static struct notifier_block rtnetlink_dev_notifier = {
4579 .notifier_call = rtnetlink_event,
4580 };
4581
4582
4583 static int __net_init rtnetlink_net_init(struct net *net)
4584 {
4585 struct sock *sk;
4586 struct netlink_kernel_cfg cfg = {
4587 .groups = RTNLGRP_MAX,
4588 .input = rtnetlink_rcv,
4589 .cb_mutex = &rtnl_mutex,
4590 .flags = NL_CFG_F_NONROOT_RECV,
4591 .bind = rtnetlink_bind,
4592 };
4593
4594 sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
4595 if (!sk)
4596 return -ENOMEM;
4597 net->rtnl = sk;
4598 return 0;
4599 }
4600
4601 static void __net_exit rtnetlink_net_exit(struct net *net)
4602 {
4603 netlink_kernel_release(net->rtnl);
4604 net->rtnl = NULL;
4605 }
4606
4607 static struct pernet_operations rtnetlink_net_ops = {
4608 .init = rtnetlink_net_init,
4609 .exit = rtnetlink_net_exit,
4610 };
4611
4612 void __init rtnetlink_init(void)
4613 {
4614 int i;
4615
4616 for (i = 0; i < ARRAY_SIZE(rtnl_msg_handlers_ref); i++)
4617 refcount_set(&rtnl_msg_handlers_ref[i], 1);
4618
4619 if (register_pernet_subsys(&rtnetlink_net_ops))
4620 panic("rtnetlink_init: cannot initialize rtnetlink\n");
4621
4622 register_netdevice_notifier(&rtnetlink_dev_notifier);
4623
4624 rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
4625 rtnl_dump_ifinfo, 0);
4626 rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, 0);
4627 rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, 0);
4628 rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, 0);
4629
4630 rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, 0);
4631 rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, 0);
4632 rtnl_register(PF_UNSPEC, RTM_GETNETCONF, NULL, rtnl_dump_all, 0);
4633
4634 rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, 0);
4635 rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, 0);
4636 rtnl_register(PF_BRIDGE, RTM_GETNEIGH, NULL, rtnl_fdb_dump, 0);
4637
4638 rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, 0);
4639 rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, 0);
4640 rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, 0);
4641
4642 rtnl_register(PF_UNSPEC, RTM_GETSTATS, rtnl_stats_get, rtnl_stats_dump,
4643 0);
4644 }
ubuntu-bionic-kernel mirror