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