Loading...
1// SPDX-License-Identifier: GPL-2.0-or-later
2/*
3 * NET3 IP device support routines.
4 *
5 * Derived from the IP parts of dev.c 1.0.19
6 * Authors: Ross Biro
7 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
8 * Mark Evans, <evansmp@uhura.aston.ac.uk>
9 *
10 * Additional Authors:
11 * Alan Cox, <gw4pts@gw4pts.ampr.org>
12 * Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
13 *
14 * Changes:
15 * Alexey Kuznetsov: pa_* fields are replaced with ifaddr
16 * lists.
17 * Cyrus Durgin: updated for kmod
18 * Matthias Andree: in devinet_ioctl, compare label and
19 * address (4.4BSD alias style support),
20 * fall back to comparing just the label
21 * if no match found.
22 */
23
24
25#include <linux/uaccess.h>
26#include <linux/bitops.h>
27#include <linux/capability.h>
28#include <linux/module.h>
29#include <linux/types.h>
30#include <linux/kernel.h>
31#include <linux/sched/signal.h>
32#include <linux/string.h>
33#include <linux/mm.h>
34#include <linux/socket.h>
35#include <linux/sockios.h>
36#include <linux/in.h>
37#include <linux/errno.h>
38#include <linux/interrupt.h>
39#include <linux/if_addr.h>
40#include <linux/if_ether.h>
41#include <linux/inet.h>
42#include <linux/netdevice.h>
43#include <linux/etherdevice.h>
44#include <linux/skbuff.h>
45#include <linux/init.h>
46#include <linux/notifier.h>
47#include <linux/inetdevice.h>
48#include <linux/igmp.h>
49#include <linux/slab.h>
50#include <linux/hash.h>
51#ifdef CONFIG_SYSCTL
52#include <linux/sysctl.h>
53#endif
54#include <linux/kmod.h>
55#include <linux/netconf.h>
56
57#include <net/arp.h>
58#include <net/ip.h>
59#include <net/route.h>
60#include <net/ip_fib.h>
61#include <net/rtnetlink.h>
62#include <net/net_namespace.h>
63#include <net/addrconf.h>
64
65#define IPV6ONLY_FLAGS \
66 (IFA_F_NODAD | IFA_F_OPTIMISTIC | IFA_F_DADFAILED | \
67 IFA_F_HOMEADDRESS | IFA_F_TENTATIVE | \
68 IFA_F_MANAGETEMPADDR | IFA_F_STABLE_PRIVACY)
69
70static struct ipv4_devconf ipv4_devconf = {
71 .data = {
72 [IPV4_DEVCONF_ACCEPT_REDIRECTS - 1] = 1,
73 [IPV4_DEVCONF_SEND_REDIRECTS - 1] = 1,
74 [IPV4_DEVCONF_SECURE_REDIRECTS - 1] = 1,
75 [IPV4_DEVCONF_SHARED_MEDIA - 1] = 1,
76 [IPV4_DEVCONF_IGMPV2_UNSOLICITED_REPORT_INTERVAL - 1] = 10000 /*ms*/,
77 [IPV4_DEVCONF_IGMPV3_UNSOLICITED_REPORT_INTERVAL - 1] = 1000 /*ms*/,
78 [IPV4_DEVCONF_ARP_EVICT_NOCARRIER - 1] = 1,
79 },
80};
81
82static struct ipv4_devconf ipv4_devconf_dflt = {
83 .data = {
84 [IPV4_DEVCONF_ACCEPT_REDIRECTS - 1] = 1,
85 [IPV4_DEVCONF_SEND_REDIRECTS - 1] = 1,
86 [IPV4_DEVCONF_SECURE_REDIRECTS - 1] = 1,
87 [IPV4_DEVCONF_SHARED_MEDIA - 1] = 1,
88 [IPV4_DEVCONF_ACCEPT_SOURCE_ROUTE - 1] = 1,
89 [IPV4_DEVCONF_IGMPV2_UNSOLICITED_REPORT_INTERVAL - 1] = 10000 /*ms*/,
90 [IPV4_DEVCONF_IGMPV3_UNSOLICITED_REPORT_INTERVAL - 1] = 1000 /*ms*/,
91 [IPV4_DEVCONF_ARP_EVICT_NOCARRIER - 1] = 1,
92 },
93};
94
95#define IPV4_DEVCONF_DFLT(net, attr) \
96 IPV4_DEVCONF((*net->ipv4.devconf_dflt), attr)
97
98static const struct nla_policy ifa_ipv4_policy[IFA_MAX+1] = {
99 [IFA_LOCAL] = { .type = NLA_U32 },
100 [IFA_ADDRESS] = { .type = NLA_U32 },
101 [IFA_BROADCAST] = { .type = NLA_U32 },
102 [IFA_LABEL] = { .type = NLA_STRING, .len = IFNAMSIZ - 1 },
103 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
104 [IFA_FLAGS] = { .type = NLA_U32 },
105 [IFA_RT_PRIORITY] = { .type = NLA_U32 },
106 [IFA_TARGET_NETNSID] = { .type = NLA_S32 },
107 [IFA_PROTO] = { .type = NLA_U8 },
108};
109
110struct inet_fill_args {
111 u32 portid;
112 u32 seq;
113 int event;
114 unsigned int flags;
115 int netnsid;
116 int ifindex;
117};
118
119#define IN4_ADDR_HSIZE_SHIFT 8
120#define IN4_ADDR_HSIZE (1U << IN4_ADDR_HSIZE_SHIFT)
121
122static u32 inet_addr_hash(const struct net *net, __be32 addr)
123{
124 u32 val = __ipv4_addr_hash(addr, net_hash_mix(net));
125
126 return hash_32(val, IN4_ADDR_HSIZE_SHIFT);
127}
128
129static void inet_hash_insert(struct net *net, struct in_ifaddr *ifa)
130{
131 u32 hash = inet_addr_hash(net, ifa->ifa_local);
132
133 ASSERT_RTNL();
134 hlist_add_head_rcu(&ifa->addr_lst, &net->ipv4.inet_addr_lst[hash]);
135}
136
137static void inet_hash_remove(struct in_ifaddr *ifa)
138{
139 ASSERT_RTNL();
140 hlist_del_init_rcu(&ifa->addr_lst);
141}
142
143/**
144 * __ip_dev_find - find the first device with a given source address.
145 * @net: the net namespace
146 * @addr: the source address
147 * @devref: if true, take a reference on the found device
148 *
149 * If a caller uses devref=false, it should be protected by RCU, or RTNL
150 */
151struct net_device *__ip_dev_find(struct net *net, __be32 addr, bool devref)
152{
153 struct net_device *result = NULL;
154 struct in_ifaddr *ifa;
155
156 rcu_read_lock();
157 ifa = inet_lookup_ifaddr_rcu(net, addr);
158 if (!ifa) {
159 struct flowi4 fl4 = { .daddr = addr };
160 struct fib_result res = { 0 };
161 struct fib_table *local;
162
163 /* Fallback to FIB local table so that communication
164 * over loopback subnets work.
165 */
166 local = fib_get_table(net, RT_TABLE_LOCAL);
167 if (local &&
168 !fib_table_lookup(local, &fl4, &res, FIB_LOOKUP_NOREF) &&
169 res.type == RTN_LOCAL)
170 result = FIB_RES_DEV(res);
171 } else {
172 result = ifa->ifa_dev->dev;
173 }
174 if (result && devref)
175 dev_hold(result);
176 rcu_read_unlock();
177 return result;
178}
179EXPORT_SYMBOL(__ip_dev_find);
180
181/* called under RCU lock */
182struct in_ifaddr *inet_lookup_ifaddr_rcu(struct net *net, __be32 addr)
183{
184 u32 hash = inet_addr_hash(net, addr);
185 struct in_ifaddr *ifa;
186
187 hlist_for_each_entry_rcu(ifa, &net->ipv4.inet_addr_lst[hash], addr_lst)
188 if (ifa->ifa_local == addr)
189 return ifa;
190
191 return NULL;
192}
193
194static void rtmsg_ifa(int event, struct in_ifaddr *, struct nlmsghdr *, u32);
195
196static BLOCKING_NOTIFIER_HEAD(inetaddr_chain);
197static BLOCKING_NOTIFIER_HEAD(inetaddr_validator_chain);
198static void inet_del_ifa(struct in_device *in_dev,
199 struct in_ifaddr __rcu **ifap,
200 int destroy);
201#ifdef CONFIG_SYSCTL
202static int devinet_sysctl_register(struct in_device *idev);
203static void devinet_sysctl_unregister(struct in_device *idev);
204#else
205static int devinet_sysctl_register(struct in_device *idev)
206{
207 return 0;
208}
209static void devinet_sysctl_unregister(struct in_device *idev)
210{
211}
212#endif
213
214/* Locks all the inet devices. */
215
216static struct in_ifaddr *inet_alloc_ifa(struct in_device *in_dev)
217{
218 struct in_ifaddr *ifa;
219
220 ifa = kzalloc(sizeof(*ifa), GFP_KERNEL_ACCOUNT);
221 if (!ifa)
222 return NULL;
223
224 in_dev_hold(in_dev);
225 ifa->ifa_dev = in_dev;
226
227 INIT_HLIST_NODE(&ifa->addr_lst);
228
229 return ifa;
230}
231
232static void inet_rcu_free_ifa(struct rcu_head *head)
233{
234 struct in_ifaddr *ifa = container_of(head, struct in_ifaddr, rcu_head);
235
236 in_dev_put(ifa->ifa_dev);
237 kfree(ifa);
238}
239
240static void inet_free_ifa(struct in_ifaddr *ifa)
241{
242 /* Our reference to ifa->ifa_dev must be freed ASAP
243 * to release the reference to the netdev the same way.
244 * in_dev_put() -> in_dev_finish_destroy() -> netdev_put()
245 */
246 call_rcu_hurry(&ifa->rcu_head, inet_rcu_free_ifa);
247}
248
249static void in_dev_free_rcu(struct rcu_head *head)
250{
251 struct in_device *idev = container_of(head, struct in_device, rcu_head);
252
253 kfree(rcu_dereference_protected(idev->mc_hash, 1));
254 kfree(idev);
255}
256
257void in_dev_finish_destroy(struct in_device *idev)
258{
259 struct net_device *dev = idev->dev;
260
261 WARN_ON(idev->ifa_list);
262 WARN_ON(idev->mc_list);
263#ifdef NET_REFCNT_DEBUG
264 pr_debug("%s: %p=%s\n", __func__, idev, dev ? dev->name : "NIL");
265#endif
266 netdev_put(dev, &idev->dev_tracker);
267 if (!idev->dead)
268 pr_err("Freeing alive in_device %p\n", idev);
269 else
270 call_rcu(&idev->rcu_head, in_dev_free_rcu);
271}
272EXPORT_SYMBOL(in_dev_finish_destroy);
273
274static struct in_device *inetdev_init(struct net_device *dev)
275{
276 struct in_device *in_dev;
277 int err = -ENOMEM;
278
279 ASSERT_RTNL();
280
281 in_dev = kzalloc(sizeof(*in_dev), GFP_KERNEL);
282 if (!in_dev)
283 goto out;
284 memcpy(&in_dev->cnf, dev_net(dev)->ipv4.devconf_dflt,
285 sizeof(in_dev->cnf));
286 in_dev->cnf.sysctl = NULL;
287 in_dev->dev = dev;
288 in_dev->arp_parms = neigh_parms_alloc(dev, &arp_tbl);
289 if (!in_dev->arp_parms)
290 goto out_kfree;
291 if (IPV4_DEVCONF(in_dev->cnf, FORWARDING))
292 dev_disable_lro(dev);
293 /* Reference in_dev->dev */
294 netdev_hold(dev, &in_dev->dev_tracker, GFP_KERNEL);
295 /* Account for reference dev->ip_ptr (below) */
296 refcount_set(&in_dev->refcnt, 1);
297
298 if (dev != blackhole_netdev) {
299 err = devinet_sysctl_register(in_dev);
300 if (err) {
301 in_dev->dead = 1;
302 neigh_parms_release(&arp_tbl, in_dev->arp_parms);
303 in_dev_put(in_dev);
304 in_dev = NULL;
305 goto out;
306 }
307 ip_mc_init_dev(in_dev);
308 if (dev->flags & IFF_UP)
309 ip_mc_up(in_dev);
310 }
311
312 /* we can receive as soon as ip_ptr is set -- do this last */
313 rcu_assign_pointer(dev->ip_ptr, in_dev);
314out:
315 return in_dev ?: ERR_PTR(err);
316out_kfree:
317 kfree(in_dev);
318 in_dev = NULL;
319 goto out;
320}
321
322static void inetdev_destroy(struct in_device *in_dev)
323{
324 struct net_device *dev;
325 struct in_ifaddr *ifa;
326
327 ASSERT_RTNL();
328
329 dev = in_dev->dev;
330
331 in_dev->dead = 1;
332
333 ip_mc_destroy_dev(in_dev);
334
335 while ((ifa = rtnl_dereference(in_dev->ifa_list)) != NULL) {
336 inet_del_ifa(in_dev, &in_dev->ifa_list, 0);
337 inet_free_ifa(ifa);
338 }
339
340 RCU_INIT_POINTER(dev->ip_ptr, NULL);
341
342 devinet_sysctl_unregister(in_dev);
343 neigh_parms_release(&arp_tbl, in_dev->arp_parms);
344 arp_ifdown(dev);
345
346 in_dev_put(in_dev);
347}
348
349static int __init inet_blackhole_dev_init(void)
350{
351 int err = 0;
352
353 rtnl_lock();
354 if (!inetdev_init(blackhole_netdev))
355 err = -ENOMEM;
356 rtnl_unlock();
357
358 return err;
359}
360late_initcall(inet_blackhole_dev_init);
361
362int inet_addr_onlink(struct in_device *in_dev, __be32 a, __be32 b)
363{
364 const struct in_ifaddr *ifa;
365
366 rcu_read_lock();
367 in_dev_for_each_ifa_rcu(ifa, in_dev) {
368 if (inet_ifa_match(a, ifa)) {
369 if (!b || inet_ifa_match(b, ifa)) {
370 rcu_read_unlock();
371 return 1;
372 }
373 }
374 }
375 rcu_read_unlock();
376 return 0;
377}
378
379static void __inet_del_ifa(struct in_device *in_dev,
380 struct in_ifaddr __rcu **ifap,
381 int destroy, struct nlmsghdr *nlh, u32 portid)
382{
383 struct in_ifaddr *promote = NULL;
384 struct in_ifaddr *ifa, *ifa1;
385 struct in_ifaddr __rcu **last_prim;
386 struct in_ifaddr *prev_prom = NULL;
387 int do_promote = IN_DEV_PROMOTE_SECONDARIES(in_dev);
388
389 ASSERT_RTNL();
390
391 ifa1 = rtnl_dereference(*ifap);
392 last_prim = ifap;
393 if (in_dev->dead)
394 goto no_promotions;
395
396 /* 1. Deleting primary ifaddr forces deletion all secondaries
397 * unless alias promotion is set
398 **/
399
400 if (!(ifa1->ifa_flags & IFA_F_SECONDARY)) {
401 struct in_ifaddr __rcu **ifap1 = &ifa1->ifa_next;
402
403 while ((ifa = rtnl_dereference(*ifap1)) != NULL) {
404 if (!(ifa->ifa_flags & IFA_F_SECONDARY) &&
405 ifa1->ifa_scope <= ifa->ifa_scope)
406 last_prim = &ifa->ifa_next;
407
408 if (!(ifa->ifa_flags & IFA_F_SECONDARY) ||
409 ifa1->ifa_mask != ifa->ifa_mask ||
410 !inet_ifa_match(ifa1->ifa_address, ifa)) {
411 ifap1 = &ifa->ifa_next;
412 prev_prom = ifa;
413 continue;
414 }
415
416 if (!do_promote) {
417 inet_hash_remove(ifa);
418 *ifap1 = ifa->ifa_next;
419
420 rtmsg_ifa(RTM_DELADDR, ifa, nlh, portid);
421 blocking_notifier_call_chain(&inetaddr_chain,
422 NETDEV_DOWN, ifa);
423 inet_free_ifa(ifa);
424 } else {
425 promote = ifa;
426 break;
427 }
428 }
429 }
430
431 /* On promotion all secondaries from subnet are changing
432 * the primary IP, we must remove all their routes silently
433 * and later to add them back with new prefsrc. Do this
434 * while all addresses are on the device list.
435 */
436 for (ifa = promote; ifa; ifa = rtnl_dereference(ifa->ifa_next)) {
437 if (ifa1->ifa_mask == ifa->ifa_mask &&
438 inet_ifa_match(ifa1->ifa_address, ifa))
439 fib_del_ifaddr(ifa, ifa1);
440 }
441
442no_promotions:
443 /* 2. Unlink it */
444
445 *ifap = ifa1->ifa_next;
446 inet_hash_remove(ifa1);
447
448 /* 3. Announce address deletion */
449
450 /* Send message first, then call notifier.
451 At first sight, FIB update triggered by notifier
452 will refer to already deleted ifaddr, that could confuse
453 netlink listeners. It is not true: look, gated sees
454 that route deleted and if it still thinks that ifaddr
455 is valid, it will try to restore deleted routes... Grr.
456 So that, this order is correct.
457 */
458 rtmsg_ifa(RTM_DELADDR, ifa1, nlh, portid);
459 blocking_notifier_call_chain(&inetaddr_chain, NETDEV_DOWN, ifa1);
460
461 if (promote) {
462 struct in_ifaddr *next_sec;
463
464 next_sec = rtnl_dereference(promote->ifa_next);
465 if (prev_prom) {
466 struct in_ifaddr *last_sec;
467
468 rcu_assign_pointer(prev_prom->ifa_next, next_sec);
469
470 last_sec = rtnl_dereference(*last_prim);
471 rcu_assign_pointer(promote->ifa_next, last_sec);
472 rcu_assign_pointer(*last_prim, promote);
473 }
474
475 promote->ifa_flags &= ~IFA_F_SECONDARY;
476 rtmsg_ifa(RTM_NEWADDR, promote, nlh, portid);
477 blocking_notifier_call_chain(&inetaddr_chain,
478 NETDEV_UP, promote);
479 for (ifa = next_sec; ifa;
480 ifa = rtnl_dereference(ifa->ifa_next)) {
481 if (ifa1->ifa_mask != ifa->ifa_mask ||
482 !inet_ifa_match(ifa1->ifa_address, ifa))
483 continue;
484 fib_add_ifaddr(ifa);
485 }
486
487 }
488 if (destroy)
489 inet_free_ifa(ifa1);
490}
491
492static void inet_del_ifa(struct in_device *in_dev,
493 struct in_ifaddr __rcu **ifap,
494 int destroy)
495{
496 __inet_del_ifa(in_dev, ifap, destroy, NULL, 0);
497}
498
499static int __inet_insert_ifa(struct in_ifaddr *ifa, struct nlmsghdr *nlh,
500 u32 portid, struct netlink_ext_ack *extack)
501{
502 struct in_ifaddr __rcu **last_primary, **ifap;
503 struct in_device *in_dev = ifa->ifa_dev;
504 struct net *net = dev_net(in_dev->dev);
505 struct in_validator_info ivi;
506 struct in_ifaddr *ifa1;
507 int ret;
508
509 ASSERT_RTNL();
510
511 ifa->ifa_flags &= ~IFA_F_SECONDARY;
512 last_primary = &in_dev->ifa_list;
513
514 /* Don't set IPv6 only flags to IPv4 addresses */
515 ifa->ifa_flags &= ~IPV6ONLY_FLAGS;
516
517 ifap = &in_dev->ifa_list;
518 ifa1 = rtnl_dereference(*ifap);
519
520 while (ifa1) {
521 if (!(ifa1->ifa_flags & IFA_F_SECONDARY) &&
522 ifa->ifa_scope <= ifa1->ifa_scope)
523 last_primary = &ifa1->ifa_next;
524 if (ifa1->ifa_mask == ifa->ifa_mask &&
525 inet_ifa_match(ifa1->ifa_address, ifa)) {
526 if (ifa1->ifa_local == ifa->ifa_local) {
527 inet_free_ifa(ifa);
528 return -EEXIST;
529 }
530 if (ifa1->ifa_scope != ifa->ifa_scope) {
531 NL_SET_ERR_MSG(extack, "ipv4: Invalid scope value");
532 inet_free_ifa(ifa);
533 return -EINVAL;
534 }
535 ifa->ifa_flags |= IFA_F_SECONDARY;
536 }
537
538 ifap = &ifa1->ifa_next;
539 ifa1 = rtnl_dereference(*ifap);
540 }
541
542 /* Allow any devices that wish to register ifaddr validtors to weigh
543 * in now, before changes are committed. The rntl lock is serializing
544 * access here, so the state should not change between a validator call
545 * and a final notify on commit. This isn't invoked on promotion under
546 * the assumption that validators are checking the address itself, and
547 * not the flags.
548 */
549 ivi.ivi_addr = ifa->ifa_address;
550 ivi.ivi_dev = ifa->ifa_dev;
551 ivi.extack = extack;
552 ret = blocking_notifier_call_chain(&inetaddr_validator_chain,
553 NETDEV_UP, &ivi);
554 ret = notifier_to_errno(ret);
555 if (ret) {
556 inet_free_ifa(ifa);
557 return ret;
558 }
559
560 if (!(ifa->ifa_flags & IFA_F_SECONDARY))
561 ifap = last_primary;
562
563 rcu_assign_pointer(ifa->ifa_next, *ifap);
564 rcu_assign_pointer(*ifap, ifa);
565
566 inet_hash_insert(dev_net(in_dev->dev), ifa);
567
568 cancel_delayed_work(&net->ipv4.addr_chk_work);
569 queue_delayed_work(system_power_efficient_wq, &net->ipv4.addr_chk_work, 0);
570
571 /* Send message first, then call notifier.
572 Notifier will trigger FIB update, so that
573 listeners of netlink will know about new ifaddr */
574 rtmsg_ifa(RTM_NEWADDR, ifa, nlh, portid);
575 blocking_notifier_call_chain(&inetaddr_chain, NETDEV_UP, ifa);
576
577 return 0;
578}
579
580static int inet_insert_ifa(struct in_ifaddr *ifa)
581{
582 if (!ifa->ifa_local) {
583 inet_free_ifa(ifa);
584 return 0;
585 }
586
587 return __inet_insert_ifa(ifa, NULL, 0, NULL);
588}
589
590static int inet_set_ifa(struct net_device *dev, struct in_ifaddr *ifa)
591{
592 struct in_device *in_dev = __in_dev_get_rtnl_net(dev);
593
594 ipv4_devconf_setall(in_dev);
595 neigh_parms_data_state_setall(in_dev->arp_parms);
596
597 if (ipv4_is_loopback(ifa->ifa_local))
598 ifa->ifa_scope = RT_SCOPE_HOST;
599 return inet_insert_ifa(ifa);
600}
601
602/* Caller must hold RCU or RTNL :
603 * We dont take a reference on found in_device
604 */
605struct in_device *inetdev_by_index(struct net *net, int ifindex)
606{
607 struct net_device *dev;
608 struct in_device *in_dev = NULL;
609
610 rcu_read_lock();
611 dev = dev_get_by_index_rcu(net, ifindex);
612 if (dev)
613 in_dev = rcu_dereference_rtnl(dev->ip_ptr);
614 rcu_read_unlock();
615 return in_dev;
616}
617EXPORT_SYMBOL(inetdev_by_index);
618
619/* Called only from RTNL semaphored context. No locks. */
620
621struct in_ifaddr *inet_ifa_byprefix(struct in_device *in_dev, __be32 prefix,
622 __be32 mask)
623{
624 struct in_ifaddr *ifa;
625
626 ASSERT_RTNL();
627
628 in_dev_for_each_ifa_rtnl(ifa, in_dev) {
629 if (ifa->ifa_mask == mask && inet_ifa_match(prefix, ifa))
630 return ifa;
631 }
632 return NULL;
633}
634
635static int ip_mc_autojoin_config(struct net *net, bool join,
636 const struct in_ifaddr *ifa)
637{
638#if defined(CONFIG_IP_MULTICAST)
639 struct ip_mreqn mreq = {
640 .imr_multiaddr.s_addr = ifa->ifa_address,
641 .imr_ifindex = ifa->ifa_dev->dev->ifindex,
642 };
643 struct sock *sk = net->ipv4.mc_autojoin_sk;
644 int ret;
645
646 ASSERT_RTNL_NET(net);
647
648 lock_sock(sk);
649 if (join)
650 ret = ip_mc_join_group(sk, &mreq);
651 else
652 ret = ip_mc_leave_group(sk, &mreq);
653 release_sock(sk);
654
655 return ret;
656#else
657 return -EOPNOTSUPP;
658#endif
659}
660
661static int inet_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
662 struct netlink_ext_ack *extack)
663{
664 struct net *net = sock_net(skb->sk);
665 struct in_ifaddr __rcu **ifap;
666 struct nlattr *tb[IFA_MAX+1];
667 struct in_device *in_dev;
668 struct ifaddrmsg *ifm;
669 struct in_ifaddr *ifa;
670 int err;
671
672 err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
673 ifa_ipv4_policy, extack);
674 if (err < 0)
675 goto out;
676
677 ifm = nlmsg_data(nlh);
678
679 rtnl_net_lock(net);
680
681 in_dev = inetdev_by_index(net, ifm->ifa_index);
682 if (!in_dev) {
683 NL_SET_ERR_MSG(extack, "ipv4: Device not found");
684 err = -ENODEV;
685 goto unlock;
686 }
687
688 for (ifap = &in_dev->ifa_list;
689 (ifa = rtnl_net_dereference(net, *ifap)) != NULL;
690 ifap = &ifa->ifa_next) {
691 if (tb[IFA_LOCAL] &&
692 ifa->ifa_local != nla_get_in_addr(tb[IFA_LOCAL]))
693 continue;
694
695 if (tb[IFA_LABEL] && nla_strcmp(tb[IFA_LABEL], ifa->ifa_label))
696 continue;
697
698 if (tb[IFA_ADDRESS] &&
699 (ifm->ifa_prefixlen != ifa->ifa_prefixlen ||
700 !inet_ifa_match(nla_get_in_addr(tb[IFA_ADDRESS]), ifa)))
701 continue;
702
703 if (ipv4_is_multicast(ifa->ifa_address))
704 ip_mc_autojoin_config(net, false, ifa);
705
706 __inet_del_ifa(in_dev, ifap, 1, nlh, NETLINK_CB(skb).portid);
707 goto unlock;
708 }
709
710 NL_SET_ERR_MSG(extack, "ipv4: Address not found");
711 err = -EADDRNOTAVAIL;
712unlock:
713 rtnl_net_unlock(net);
714out:
715 return err;
716}
717
718static void check_lifetime(struct work_struct *work)
719{
720 unsigned long now, next, next_sec, next_sched;
721 struct in_ifaddr *ifa;
722 struct hlist_node *n;
723 struct net *net;
724 int i;
725
726 net = container_of(to_delayed_work(work), struct net, ipv4.addr_chk_work);
727 now = jiffies;
728 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
729
730 for (i = 0; i < IN4_ADDR_HSIZE; i++) {
731 struct hlist_head *head = &net->ipv4.inet_addr_lst[i];
732 bool change_needed = false;
733
734 rcu_read_lock();
735 hlist_for_each_entry_rcu(ifa, head, addr_lst) {
736 unsigned long age, tstamp;
737 u32 preferred_lft;
738 u32 valid_lft;
739 u32 flags;
740
741 flags = READ_ONCE(ifa->ifa_flags);
742 if (flags & IFA_F_PERMANENT)
743 continue;
744
745 preferred_lft = READ_ONCE(ifa->ifa_preferred_lft);
746 valid_lft = READ_ONCE(ifa->ifa_valid_lft);
747 tstamp = READ_ONCE(ifa->ifa_tstamp);
748 /* We try to batch several events at once. */
749 age = (now - tstamp +
750 ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
751
752 if (valid_lft != INFINITY_LIFE_TIME &&
753 age >= valid_lft) {
754 change_needed = true;
755 } else if (preferred_lft ==
756 INFINITY_LIFE_TIME) {
757 continue;
758 } else if (age >= preferred_lft) {
759 if (time_before(tstamp + valid_lft * HZ, next))
760 next = tstamp + valid_lft * HZ;
761
762 if (!(flags & IFA_F_DEPRECATED))
763 change_needed = true;
764 } else if (time_before(tstamp + preferred_lft * HZ,
765 next)) {
766 next = tstamp + preferred_lft * HZ;
767 }
768 }
769 rcu_read_unlock();
770 if (!change_needed)
771 continue;
772
773 rtnl_net_lock(net);
774 hlist_for_each_entry_safe(ifa, n, head, addr_lst) {
775 unsigned long age;
776
777 if (ifa->ifa_flags & IFA_F_PERMANENT)
778 continue;
779
780 /* We try to batch several events at once. */
781 age = (now - ifa->ifa_tstamp +
782 ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
783
784 if (ifa->ifa_valid_lft != INFINITY_LIFE_TIME &&
785 age >= ifa->ifa_valid_lft) {
786 struct in_ifaddr __rcu **ifap;
787 struct in_ifaddr *tmp;
788
789 ifap = &ifa->ifa_dev->ifa_list;
790 tmp = rtnl_net_dereference(net, *ifap);
791 while (tmp) {
792 if (tmp == ifa) {
793 inet_del_ifa(ifa->ifa_dev,
794 ifap, 1);
795 break;
796 }
797 ifap = &tmp->ifa_next;
798 tmp = rtnl_net_dereference(net, *ifap);
799 }
800 } else if (ifa->ifa_preferred_lft !=
801 INFINITY_LIFE_TIME &&
802 age >= ifa->ifa_preferred_lft &&
803 !(ifa->ifa_flags & IFA_F_DEPRECATED)) {
804 ifa->ifa_flags |= IFA_F_DEPRECATED;
805 rtmsg_ifa(RTM_NEWADDR, ifa, NULL, 0);
806 }
807 }
808 rtnl_net_unlock(net);
809 }
810
811 next_sec = round_jiffies_up(next);
812 next_sched = next;
813
814 /* If rounded timeout is accurate enough, accept it. */
815 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
816 next_sched = next_sec;
817
818 now = jiffies;
819 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
820 if (time_before(next_sched, now + ADDRCONF_TIMER_FUZZ_MAX))
821 next_sched = now + ADDRCONF_TIMER_FUZZ_MAX;
822
823 queue_delayed_work(system_power_efficient_wq, &net->ipv4.addr_chk_work,
824 next_sched - now);
825}
826
827static void set_ifa_lifetime(struct in_ifaddr *ifa, __u32 valid_lft,
828 __u32 prefered_lft)
829{
830 unsigned long timeout;
831 u32 flags;
832
833 flags = ifa->ifa_flags & ~(IFA_F_PERMANENT | IFA_F_DEPRECATED);
834
835 timeout = addrconf_timeout_fixup(valid_lft, HZ);
836 if (addrconf_finite_timeout(timeout))
837 WRITE_ONCE(ifa->ifa_valid_lft, timeout);
838 else
839 flags |= IFA_F_PERMANENT;
840
841 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
842 if (addrconf_finite_timeout(timeout)) {
843 if (timeout == 0)
844 flags |= IFA_F_DEPRECATED;
845 WRITE_ONCE(ifa->ifa_preferred_lft, timeout);
846 }
847 WRITE_ONCE(ifa->ifa_flags, flags);
848 WRITE_ONCE(ifa->ifa_tstamp, jiffies);
849 if (!ifa->ifa_cstamp)
850 WRITE_ONCE(ifa->ifa_cstamp, ifa->ifa_tstamp);
851}
852
853static int inet_validate_rtm(struct nlmsghdr *nlh, struct nlattr **tb,
854 struct netlink_ext_ack *extack,
855 __u32 *valid_lft, __u32 *prefered_lft)
856{
857 struct ifaddrmsg *ifm = nlmsg_data(nlh);
858 int err;
859
860 err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
861 ifa_ipv4_policy, extack);
862 if (err < 0)
863 return err;
864
865 if (ifm->ifa_prefixlen > 32) {
866 NL_SET_ERR_MSG(extack, "ipv4: Invalid prefix length");
867 return -EINVAL;
868 }
869
870 if (!tb[IFA_LOCAL]) {
871 NL_SET_ERR_MSG(extack, "ipv4: Local address is not supplied");
872 return -EINVAL;
873 }
874
875 if (tb[IFA_CACHEINFO]) {
876 struct ifa_cacheinfo *ci;
877
878 ci = nla_data(tb[IFA_CACHEINFO]);
879 if (!ci->ifa_valid || ci->ifa_prefered > ci->ifa_valid) {
880 NL_SET_ERR_MSG(extack, "ipv4: address lifetime invalid");
881 return -EINVAL;
882 }
883
884 *valid_lft = ci->ifa_valid;
885 *prefered_lft = ci->ifa_prefered;
886 }
887
888 return 0;
889}
890
891static struct in_ifaddr *inet_rtm_to_ifa(struct net *net, struct nlmsghdr *nlh,
892 struct nlattr **tb,
893 struct netlink_ext_ack *extack)
894{
895 struct ifaddrmsg *ifm = nlmsg_data(nlh);
896 struct in_device *in_dev;
897 struct net_device *dev;
898 struct in_ifaddr *ifa;
899 int err;
900
901 dev = __dev_get_by_index(net, ifm->ifa_index);
902 err = -ENODEV;
903 if (!dev) {
904 NL_SET_ERR_MSG(extack, "ipv4: Device not found");
905 goto errout;
906 }
907
908 in_dev = __in_dev_get_rtnl_net(dev);
909 err = -ENOBUFS;
910 if (!in_dev)
911 goto errout;
912
913 ifa = inet_alloc_ifa(in_dev);
914 if (!ifa)
915 /*
916 * A potential indev allocation can be left alive, it stays
917 * assigned to its device and is destroy with it.
918 */
919 goto errout;
920
921 ipv4_devconf_setall(in_dev);
922 neigh_parms_data_state_setall(in_dev->arp_parms);
923
924 if (!tb[IFA_ADDRESS])
925 tb[IFA_ADDRESS] = tb[IFA_LOCAL];
926
927 ifa->ifa_prefixlen = ifm->ifa_prefixlen;
928 ifa->ifa_mask = inet_make_mask(ifm->ifa_prefixlen);
929 ifa->ifa_flags = nla_get_u32_default(tb[IFA_FLAGS], ifm->ifa_flags);
930 ifa->ifa_scope = ifm->ifa_scope;
931 ifa->ifa_local = nla_get_in_addr(tb[IFA_LOCAL]);
932 ifa->ifa_address = nla_get_in_addr(tb[IFA_ADDRESS]);
933
934 if (tb[IFA_BROADCAST])
935 ifa->ifa_broadcast = nla_get_in_addr(tb[IFA_BROADCAST]);
936
937 if (tb[IFA_LABEL])
938 nla_strscpy(ifa->ifa_label, tb[IFA_LABEL], IFNAMSIZ);
939 else
940 memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
941
942 if (tb[IFA_RT_PRIORITY])
943 ifa->ifa_rt_priority = nla_get_u32(tb[IFA_RT_PRIORITY]);
944
945 if (tb[IFA_PROTO])
946 ifa->ifa_proto = nla_get_u8(tb[IFA_PROTO]);
947
948 return ifa;
949
950errout:
951 return ERR_PTR(err);
952}
953
954static struct in_ifaddr *find_matching_ifa(struct net *net, struct in_ifaddr *ifa)
955{
956 struct in_device *in_dev = ifa->ifa_dev;
957 struct in_ifaddr *ifa1;
958
959 in_dev_for_each_ifa_rtnl_net(net, ifa1, in_dev) {
960 if (ifa1->ifa_mask == ifa->ifa_mask &&
961 inet_ifa_match(ifa1->ifa_address, ifa) &&
962 ifa1->ifa_local == ifa->ifa_local)
963 return ifa1;
964 }
965
966 return NULL;
967}
968
969static int inet_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
970 struct netlink_ext_ack *extack)
971{
972 __u32 prefered_lft = INFINITY_LIFE_TIME;
973 __u32 valid_lft = INFINITY_LIFE_TIME;
974 struct net *net = sock_net(skb->sk);
975 struct in_ifaddr *ifa_existing;
976 struct nlattr *tb[IFA_MAX + 1];
977 struct in_ifaddr *ifa;
978 int ret;
979
980 ret = inet_validate_rtm(nlh, tb, extack, &valid_lft, &prefered_lft);
981 if (ret < 0)
982 return ret;
983
984 if (!nla_get_in_addr(tb[IFA_LOCAL]))
985 return 0;
986
987 rtnl_net_lock(net);
988
989 ifa = inet_rtm_to_ifa(net, nlh, tb, extack);
990 if (IS_ERR(ifa)) {
991 ret = PTR_ERR(ifa);
992 goto unlock;
993 }
994
995 ifa_existing = find_matching_ifa(net, ifa);
996 if (!ifa_existing) {
997 /* It would be best to check for !NLM_F_CREATE here but
998 * userspace already relies on not having to provide this.
999 */
1000 set_ifa_lifetime(ifa, valid_lft, prefered_lft);
1001 if (ifa->ifa_flags & IFA_F_MCAUTOJOIN) {
1002 ret = ip_mc_autojoin_config(net, true, ifa);
1003 if (ret < 0) {
1004 NL_SET_ERR_MSG(extack, "ipv4: Multicast auto join failed");
1005 inet_free_ifa(ifa);
1006 goto unlock;
1007 }
1008 }
1009
1010 ret = __inet_insert_ifa(ifa, nlh, NETLINK_CB(skb).portid, extack);
1011 } else {
1012 u32 new_metric = ifa->ifa_rt_priority;
1013 u8 new_proto = ifa->ifa_proto;
1014
1015 inet_free_ifa(ifa);
1016
1017 if (nlh->nlmsg_flags & NLM_F_EXCL ||
1018 !(nlh->nlmsg_flags & NLM_F_REPLACE)) {
1019 NL_SET_ERR_MSG(extack, "ipv4: Address already assigned");
1020 ret = -EEXIST;
1021 goto unlock;
1022 }
1023 ifa = ifa_existing;
1024
1025 if (ifa->ifa_rt_priority != new_metric) {
1026 fib_modify_prefix_metric(ifa, new_metric);
1027 ifa->ifa_rt_priority = new_metric;
1028 }
1029
1030 ifa->ifa_proto = new_proto;
1031
1032 set_ifa_lifetime(ifa, valid_lft, prefered_lft);
1033 cancel_delayed_work(&net->ipv4.addr_chk_work);
1034 queue_delayed_work(system_power_efficient_wq,
1035 &net->ipv4.addr_chk_work, 0);
1036 rtmsg_ifa(RTM_NEWADDR, ifa, nlh, NETLINK_CB(skb).portid);
1037 }
1038
1039unlock:
1040 rtnl_net_unlock(net);
1041
1042 return ret;
1043}
1044
1045/*
1046 * Determine a default network mask, based on the IP address.
1047 */
1048
1049static int inet_abc_len(__be32 addr)
1050{
1051 int rc = -1; /* Something else, probably a multicast. */
1052
1053 if (ipv4_is_zeronet(addr) || ipv4_is_lbcast(addr))
1054 rc = 0;
1055 else {
1056 __u32 haddr = ntohl(addr);
1057 if (IN_CLASSA(haddr))
1058 rc = 8;
1059 else if (IN_CLASSB(haddr))
1060 rc = 16;
1061 else if (IN_CLASSC(haddr))
1062 rc = 24;
1063 else if (IN_CLASSE(haddr))
1064 rc = 32;
1065 }
1066
1067 return rc;
1068}
1069
1070
1071int devinet_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr)
1072{
1073 struct sockaddr_in sin_orig;
1074 struct sockaddr_in *sin = (struct sockaddr_in *)&ifr->ifr_addr;
1075 struct in_ifaddr __rcu **ifap = NULL;
1076 struct in_device *in_dev;
1077 struct in_ifaddr *ifa = NULL;
1078 struct net_device *dev;
1079 char *colon;
1080 int ret = -EFAULT;
1081 int tryaddrmatch = 0;
1082
1083 ifr->ifr_name[IFNAMSIZ - 1] = 0;
1084
1085 /* save original address for comparison */
1086 memcpy(&sin_orig, sin, sizeof(*sin));
1087
1088 colon = strchr(ifr->ifr_name, ':');
1089 if (colon)
1090 *colon = 0;
1091
1092 dev_load(net, ifr->ifr_name);
1093
1094 switch (cmd) {
1095 case SIOCGIFADDR: /* Get interface address */
1096 case SIOCGIFBRDADDR: /* Get the broadcast address */
1097 case SIOCGIFDSTADDR: /* Get the destination address */
1098 case SIOCGIFNETMASK: /* Get the netmask for the interface */
1099 /* Note that these ioctls will not sleep,
1100 so that we do not impose a lock.
1101 One day we will be forced to put shlock here (I mean SMP)
1102 */
1103 tryaddrmatch = (sin_orig.sin_family == AF_INET);
1104 memset(sin, 0, sizeof(*sin));
1105 sin->sin_family = AF_INET;
1106 break;
1107
1108 case SIOCSIFFLAGS:
1109 ret = -EPERM;
1110 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1111 goto out;
1112 break;
1113 case SIOCSIFADDR: /* Set interface address (and family) */
1114 case SIOCSIFBRDADDR: /* Set the broadcast address */
1115 case SIOCSIFDSTADDR: /* Set the destination address */
1116 case SIOCSIFNETMASK: /* Set the netmask for the interface */
1117 ret = -EPERM;
1118 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1119 goto out;
1120 ret = -EINVAL;
1121 if (sin->sin_family != AF_INET)
1122 goto out;
1123 break;
1124 default:
1125 ret = -EINVAL;
1126 goto out;
1127 }
1128
1129 rtnl_net_lock(net);
1130
1131 ret = -ENODEV;
1132 dev = __dev_get_by_name(net, ifr->ifr_name);
1133 if (!dev)
1134 goto done;
1135
1136 if (colon)
1137 *colon = ':';
1138
1139 in_dev = __in_dev_get_rtnl_net(dev);
1140 if (in_dev) {
1141 if (tryaddrmatch) {
1142 /* Matthias Andree */
1143 /* compare label and address (4.4BSD style) */
1144 /* note: we only do this for a limited set of ioctls
1145 and only if the original address family was AF_INET.
1146 This is checked above. */
1147
1148 for (ifap = &in_dev->ifa_list;
1149 (ifa = rtnl_net_dereference(net, *ifap)) != NULL;
1150 ifap = &ifa->ifa_next) {
1151 if (!strcmp(ifr->ifr_name, ifa->ifa_label) &&
1152 sin_orig.sin_addr.s_addr ==
1153 ifa->ifa_local) {
1154 break; /* found */
1155 }
1156 }
1157 }
1158 /* we didn't get a match, maybe the application is
1159 4.3BSD-style and passed in junk so we fall back to
1160 comparing just the label */
1161 if (!ifa) {
1162 for (ifap = &in_dev->ifa_list;
1163 (ifa = rtnl_net_dereference(net, *ifap)) != NULL;
1164 ifap = &ifa->ifa_next)
1165 if (!strcmp(ifr->ifr_name, ifa->ifa_label))
1166 break;
1167 }
1168 }
1169
1170 ret = -EADDRNOTAVAIL;
1171 if (!ifa && cmd != SIOCSIFADDR && cmd != SIOCSIFFLAGS)
1172 goto done;
1173
1174 switch (cmd) {
1175 case SIOCGIFADDR: /* Get interface address */
1176 ret = 0;
1177 sin->sin_addr.s_addr = ifa->ifa_local;
1178 break;
1179
1180 case SIOCGIFBRDADDR: /* Get the broadcast address */
1181 ret = 0;
1182 sin->sin_addr.s_addr = ifa->ifa_broadcast;
1183 break;
1184
1185 case SIOCGIFDSTADDR: /* Get the destination address */
1186 ret = 0;
1187 sin->sin_addr.s_addr = ifa->ifa_address;
1188 break;
1189
1190 case SIOCGIFNETMASK: /* Get the netmask for the interface */
1191 ret = 0;
1192 sin->sin_addr.s_addr = ifa->ifa_mask;
1193 break;
1194
1195 case SIOCSIFFLAGS:
1196 if (colon) {
1197 ret = -EADDRNOTAVAIL;
1198 if (!ifa)
1199 break;
1200 ret = 0;
1201 if (!(ifr->ifr_flags & IFF_UP))
1202 inet_del_ifa(in_dev, ifap, 1);
1203 break;
1204 }
1205
1206 /* NETDEV_UP/DOWN/CHANGE could touch a peer dev */
1207 ASSERT_RTNL();
1208 ret = dev_change_flags(dev, ifr->ifr_flags, NULL);
1209 break;
1210
1211 case SIOCSIFADDR: /* Set interface address (and family) */
1212 ret = -EINVAL;
1213 if (inet_abc_len(sin->sin_addr.s_addr) < 0)
1214 break;
1215
1216 if (!ifa) {
1217 ret = -ENOBUFS;
1218 if (!in_dev)
1219 break;
1220 ifa = inet_alloc_ifa(in_dev);
1221 if (!ifa)
1222 break;
1223
1224 if (colon)
1225 memcpy(ifa->ifa_label, ifr->ifr_name, IFNAMSIZ);
1226 else
1227 memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
1228 } else {
1229 ret = 0;
1230 if (ifa->ifa_local == sin->sin_addr.s_addr)
1231 break;
1232 inet_del_ifa(in_dev, ifap, 0);
1233 ifa->ifa_broadcast = 0;
1234 ifa->ifa_scope = 0;
1235 }
1236
1237 ifa->ifa_address = ifa->ifa_local = sin->sin_addr.s_addr;
1238
1239 if (!(dev->flags & IFF_POINTOPOINT)) {
1240 ifa->ifa_prefixlen = inet_abc_len(ifa->ifa_address);
1241 ifa->ifa_mask = inet_make_mask(ifa->ifa_prefixlen);
1242 if ((dev->flags & IFF_BROADCAST) &&
1243 ifa->ifa_prefixlen < 31)
1244 ifa->ifa_broadcast = ifa->ifa_address |
1245 ~ifa->ifa_mask;
1246 } else {
1247 ifa->ifa_prefixlen = 32;
1248 ifa->ifa_mask = inet_make_mask(32);
1249 }
1250 set_ifa_lifetime(ifa, INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
1251 ret = inet_set_ifa(dev, ifa);
1252 break;
1253
1254 case SIOCSIFBRDADDR: /* Set the broadcast address */
1255 ret = 0;
1256 if (ifa->ifa_broadcast != sin->sin_addr.s_addr) {
1257 inet_del_ifa(in_dev, ifap, 0);
1258 ifa->ifa_broadcast = sin->sin_addr.s_addr;
1259 inet_insert_ifa(ifa);
1260 }
1261 break;
1262
1263 case SIOCSIFDSTADDR: /* Set the destination address */
1264 ret = 0;
1265 if (ifa->ifa_address == sin->sin_addr.s_addr)
1266 break;
1267 ret = -EINVAL;
1268 if (inet_abc_len(sin->sin_addr.s_addr) < 0)
1269 break;
1270 ret = 0;
1271 inet_del_ifa(in_dev, ifap, 0);
1272 ifa->ifa_address = sin->sin_addr.s_addr;
1273 inet_insert_ifa(ifa);
1274 break;
1275
1276 case SIOCSIFNETMASK: /* Set the netmask for the interface */
1277
1278 /*
1279 * The mask we set must be legal.
1280 */
1281 ret = -EINVAL;
1282 if (bad_mask(sin->sin_addr.s_addr, 0))
1283 break;
1284 ret = 0;
1285 if (ifa->ifa_mask != sin->sin_addr.s_addr) {
1286 __be32 old_mask = ifa->ifa_mask;
1287 inet_del_ifa(in_dev, ifap, 0);
1288 ifa->ifa_mask = sin->sin_addr.s_addr;
1289 ifa->ifa_prefixlen = inet_mask_len(ifa->ifa_mask);
1290
1291 /* See if current broadcast address matches
1292 * with current netmask, then recalculate
1293 * the broadcast address. Otherwise it's a
1294 * funny address, so don't touch it since
1295 * the user seems to know what (s)he's doing...
1296 */
1297 if ((dev->flags & IFF_BROADCAST) &&
1298 (ifa->ifa_prefixlen < 31) &&
1299 (ifa->ifa_broadcast ==
1300 (ifa->ifa_local|~old_mask))) {
1301 ifa->ifa_broadcast = (ifa->ifa_local |
1302 ~sin->sin_addr.s_addr);
1303 }
1304 inet_insert_ifa(ifa);
1305 }
1306 break;
1307 }
1308done:
1309 rtnl_net_unlock(net);
1310out:
1311 return ret;
1312}
1313
1314int inet_gifconf(struct net_device *dev, char __user *buf, int len, int size)
1315{
1316 struct in_device *in_dev = __in_dev_get_rtnl_net(dev);
1317 const struct in_ifaddr *ifa;
1318 struct ifreq ifr;
1319 int done = 0;
1320
1321 if (WARN_ON(size > sizeof(struct ifreq)))
1322 goto out;
1323
1324 if (!in_dev)
1325 goto out;
1326
1327 in_dev_for_each_ifa_rtnl_net(dev_net(dev), ifa, in_dev) {
1328 if (!buf) {
1329 done += size;
1330 continue;
1331 }
1332 if (len < size)
1333 break;
1334 memset(&ifr, 0, sizeof(struct ifreq));
1335 strcpy(ifr.ifr_name, ifa->ifa_label);
1336
1337 (*(struct sockaddr_in *)&ifr.ifr_addr).sin_family = AF_INET;
1338 (*(struct sockaddr_in *)&ifr.ifr_addr).sin_addr.s_addr =
1339 ifa->ifa_local;
1340
1341 if (copy_to_user(buf + done, &ifr, size)) {
1342 done = -EFAULT;
1343 break;
1344 }
1345 len -= size;
1346 done += size;
1347 }
1348out:
1349 return done;
1350}
1351
1352static __be32 in_dev_select_addr(const struct in_device *in_dev,
1353 int scope)
1354{
1355 const struct in_ifaddr *ifa;
1356
1357 in_dev_for_each_ifa_rcu(ifa, in_dev) {
1358 if (READ_ONCE(ifa->ifa_flags) & IFA_F_SECONDARY)
1359 continue;
1360 if (ifa->ifa_scope != RT_SCOPE_LINK &&
1361 ifa->ifa_scope <= scope)
1362 return ifa->ifa_local;
1363 }
1364
1365 return 0;
1366}
1367
1368__be32 inet_select_addr(const struct net_device *dev, __be32 dst, int scope)
1369{
1370 const struct in_ifaddr *ifa;
1371 __be32 addr = 0;
1372 unsigned char localnet_scope = RT_SCOPE_HOST;
1373 struct in_device *in_dev;
1374 struct net *net;
1375 int master_idx;
1376
1377 rcu_read_lock();
1378 net = dev_net_rcu(dev);
1379 in_dev = __in_dev_get_rcu(dev);
1380 if (!in_dev)
1381 goto no_in_dev;
1382
1383 if (unlikely(IN_DEV_ROUTE_LOCALNET(in_dev)))
1384 localnet_scope = RT_SCOPE_LINK;
1385
1386 in_dev_for_each_ifa_rcu(ifa, in_dev) {
1387 if (READ_ONCE(ifa->ifa_flags) & IFA_F_SECONDARY)
1388 continue;
1389 if (min(ifa->ifa_scope, localnet_scope) > scope)
1390 continue;
1391 if (!dst || inet_ifa_match(dst, ifa)) {
1392 addr = ifa->ifa_local;
1393 break;
1394 }
1395 if (!addr)
1396 addr = ifa->ifa_local;
1397 }
1398
1399 if (addr)
1400 goto out_unlock;
1401no_in_dev:
1402 master_idx = l3mdev_master_ifindex_rcu(dev);
1403
1404 /* For VRFs, the VRF device takes the place of the loopback device,
1405 * with addresses on it being preferred. Note in such cases the
1406 * loopback device will be among the devices that fail the master_idx
1407 * equality check in the loop below.
1408 */
1409 if (master_idx &&
1410 (dev = dev_get_by_index_rcu(net, master_idx)) &&
1411 (in_dev = __in_dev_get_rcu(dev))) {
1412 addr = in_dev_select_addr(in_dev, scope);
1413 if (addr)
1414 goto out_unlock;
1415 }
1416
1417 /* Not loopback addresses on loopback should be preferred
1418 in this case. It is important that lo is the first interface
1419 in dev_base list.
1420 */
1421 for_each_netdev_rcu(net, dev) {
1422 if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1423 continue;
1424
1425 in_dev = __in_dev_get_rcu(dev);
1426 if (!in_dev)
1427 continue;
1428
1429 addr = in_dev_select_addr(in_dev, scope);
1430 if (addr)
1431 goto out_unlock;
1432 }
1433out_unlock:
1434 rcu_read_unlock();
1435 return addr;
1436}
1437EXPORT_SYMBOL(inet_select_addr);
1438
1439static __be32 confirm_addr_indev(struct in_device *in_dev, __be32 dst,
1440 __be32 local, int scope)
1441{
1442 unsigned char localnet_scope = RT_SCOPE_HOST;
1443 const struct in_ifaddr *ifa;
1444 __be32 addr = 0;
1445 int same = 0;
1446
1447 if (unlikely(IN_DEV_ROUTE_LOCALNET(in_dev)))
1448 localnet_scope = RT_SCOPE_LINK;
1449
1450 in_dev_for_each_ifa_rcu(ifa, in_dev) {
1451 unsigned char min_scope = min(ifa->ifa_scope, localnet_scope);
1452
1453 if (!addr &&
1454 (local == ifa->ifa_local || !local) &&
1455 min_scope <= scope) {
1456 addr = ifa->ifa_local;
1457 if (same)
1458 break;
1459 }
1460 if (!same) {
1461 same = (!local || inet_ifa_match(local, ifa)) &&
1462 (!dst || inet_ifa_match(dst, ifa));
1463 if (same && addr) {
1464 if (local || !dst)
1465 break;
1466 /* Is the selected addr into dst subnet? */
1467 if (inet_ifa_match(addr, ifa))
1468 break;
1469 /* No, then can we use new local src? */
1470 if (min_scope <= scope) {
1471 addr = ifa->ifa_local;
1472 break;
1473 }
1474 /* search for large dst subnet for addr */
1475 same = 0;
1476 }
1477 }
1478 }
1479
1480 return same ? addr : 0;
1481}
1482
1483/*
1484 * Confirm that local IP address exists using wildcards:
1485 * - net: netns to check, cannot be NULL
1486 * - in_dev: only on this interface, NULL=any interface
1487 * - dst: only in the same subnet as dst, 0=any dst
1488 * - local: address, 0=autoselect the local address
1489 * - scope: maximum allowed scope value for the local address
1490 */
1491__be32 inet_confirm_addr(struct net *net, struct in_device *in_dev,
1492 __be32 dst, __be32 local, int scope)
1493{
1494 __be32 addr = 0;
1495 struct net_device *dev;
1496
1497 if (in_dev)
1498 return confirm_addr_indev(in_dev, dst, local, scope);
1499
1500 rcu_read_lock();
1501 for_each_netdev_rcu(net, dev) {
1502 in_dev = __in_dev_get_rcu(dev);
1503 if (in_dev) {
1504 addr = confirm_addr_indev(in_dev, dst, local, scope);
1505 if (addr)
1506 break;
1507 }
1508 }
1509 rcu_read_unlock();
1510
1511 return addr;
1512}
1513EXPORT_SYMBOL(inet_confirm_addr);
1514
1515/*
1516 * Device notifier
1517 */
1518
1519int register_inetaddr_notifier(struct notifier_block *nb)
1520{
1521 return blocking_notifier_chain_register(&inetaddr_chain, nb);
1522}
1523EXPORT_SYMBOL(register_inetaddr_notifier);
1524
1525int unregister_inetaddr_notifier(struct notifier_block *nb)
1526{
1527 return blocking_notifier_chain_unregister(&inetaddr_chain, nb);
1528}
1529EXPORT_SYMBOL(unregister_inetaddr_notifier);
1530
1531int register_inetaddr_validator_notifier(struct notifier_block *nb)
1532{
1533 return blocking_notifier_chain_register(&inetaddr_validator_chain, nb);
1534}
1535EXPORT_SYMBOL(register_inetaddr_validator_notifier);
1536
1537int unregister_inetaddr_validator_notifier(struct notifier_block *nb)
1538{
1539 return blocking_notifier_chain_unregister(&inetaddr_validator_chain,
1540 nb);
1541}
1542EXPORT_SYMBOL(unregister_inetaddr_validator_notifier);
1543
1544/* Rename ifa_labels for a device name change. Make some effort to preserve
1545 * existing alias numbering and to create unique labels if possible.
1546*/
1547static void inetdev_changename(struct net_device *dev, struct in_device *in_dev)
1548{
1549 struct in_ifaddr *ifa;
1550 int named = 0;
1551
1552 in_dev_for_each_ifa_rtnl(ifa, in_dev) {
1553 char old[IFNAMSIZ], *dot;
1554
1555 memcpy(old, ifa->ifa_label, IFNAMSIZ);
1556 memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
1557 if (named++ == 0)
1558 goto skip;
1559 dot = strchr(old, ':');
1560 if (!dot) {
1561 sprintf(old, ":%d", named);
1562 dot = old;
1563 }
1564 if (strlen(dot) + strlen(dev->name) < IFNAMSIZ)
1565 strcat(ifa->ifa_label, dot);
1566 else
1567 strcpy(ifa->ifa_label + (IFNAMSIZ - strlen(dot) - 1), dot);
1568skip:
1569 rtmsg_ifa(RTM_NEWADDR, ifa, NULL, 0);
1570 }
1571}
1572
1573static void inetdev_send_gratuitous_arp(struct net_device *dev,
1574 struct in_device *in_dev)
1575
1576{
1577 const struct in_ifaddr *ifa;
1578
1579 in_dev_for_each_ifa_rtnl(ifa, in_dev) {
1580 arp_send(ARPOP_REQUEST, ETH_P_ARP,
1581 ifa->ifa_local, dev,
1582 ifa->ifa_local, NULL,
1583 dev->dev_addr, NULL);
1584 }
1585}
1586
1587/* Called only under RTNL semaphore */
1588
1589static int inetdev_event(struct notifier_block *this, unsigned long event,
1590 void *ptr)
1591{
1592 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1593 struct in_device *in_dev = __in_dev_get_rtnl(dev);
1594
1595 ASSERT_RTNL();
1596
1597 if (!in_dev) {
1598 if (event == NETDEV_REGISTER) {
1599 in_dev = inetdev_init(dev);
1600 if (IS_ERR(in_dev))
1601 return notifier_from_errno(PTR_ERR(in_dev));
1602 if (dev->flags & IFF_LOOPBACK) {
1603 IN_DEV_CONF_SET(in_dev, NOXFRM, 1);
1604 IN_DEV_CONF_SET(in_dev, NOPOLICY, 1);
1605 }
1606 } else if (event == NETDEV_CHANGEMTU) {
1607 /* Re-enabling IP */
1608 if (inetdev_valid_mtu(dev->mtu))
1609 in_dev = inetdev_init(dev);
1610 }
1611 goto out;
1612 }
1613
1614 switch (event) {
1615 case NETDEV_REGISTER:
1616 pr_debug("%s: bug\n", __func__);
1617 RCU_INIT_POINTER(dev->ip_ptr, NULL);
1618 break;
1619 case NETDEV_UP:
1620 if (!inetdev_valid_mtu(dev->mtu))
1621 break;
1622 if (dev->flags & IFF_LOOPBACK) {
1623 struct in_ifaddr *ifa = inet_alloc_ifa(in_dev);
1624
1625 if (ifa) {
1626 ifa->ifa_local =
1627 ifa->ifa_address = htonl(INADDR_LOOPBACK);
1628 ifa->ifa_prefixlen = 8;
1629 ifa->ifa_mask = inet_make_mask(8);
1630 ifa->ifa_scope = RT_SCOPE_HOST;
1631 memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
1632 set_ifa_lifetime(ifa, INFINITY_LIFE_TIME,
1633 INFINITY_LIFE_TIME);
1634 ipv4_devconf_setall(in_dev);
1635 neigh_parms_data_state_setall(in_dev->arp_parms);
1636 inet_insert_ifa(ifa);
1637 }
1638 }
1639 ip_mc_up(in_dev);
1640 fallthrough;
1641 case NETDEV_CHANGEADDR:
1642 if (!IN_DEV_ARP_NOTIFY(in_dev))
1643 break;
1644 fallthrough;
1645 case NETDEV_NOTIFY_PEERS:
1646 /* Send gratuitous ARP to notify of link change */
1647 inetdev_send_gratuitous_arp(dev, in_dev);
1648 break;
1649 case NETDEV_DOWN:
1650 ip_mc_down(in_dev);
1651 break;
1652 case NETDEV_PRE_TYPE_CHANGE:
1653 ip_mc_unmap(in_dev);
1654 break;
1655 case NETDEV_POST_TYPE_CHANGE:
1656 ip_mc_remap(in_dev);
1657 break;
1658 case NETDEV_CHANGEMTU:
1659 if (inetdev_valid_mtu(dev->mtu))
1660 break;
1661 /* disable IP when MTU is not enough */
1662 fallthrough;
1663 case NETDEV_UNREGISTER:
1664 inetdev_destroy(in_dev);
1665 break;
1666 case NETDEV_CHANGENAME:
1667 /* Do not notify about label change, this event is
1668 * not interesting to applications using netlink.
1669 */
1670 inetdev_changename(dev, in_dev);
1671
1672 devinet_sysctl_unregister(in_dev);
1673 devinet_sysctl_register(in_dev);
1674 break;
1675 }
1676out:
1677 return NOTIFY_DONE;
1678}
1679
1680static struct notifier_block ip_netdev_notifier = {
1681 .notifier_call = inetdev_event,
1682};
1683
1684static size_t inet_nlmsg_size(void)
1685{
1686 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
1687 + nla_total_size(4) /* IFA_ADDRESS */
1688 + nla_total_size(4) /* IFA_LOCAL */
1689 + nla_total_size(4) /* IFA_BROADCAST */
1690 + nla_total_size(IFNAMSIZ) /* IFA_LABEL */
1691 + nla_total_size(4) /* IFA_FLAGS */
1692 + nla_total_size(1) /* IFA_PROTO */
1693 + nla_total_size(4) /* IFA_RT_PRIORITY */
1694 + nla_total_size(sizeof(struct ifa_cacheinfo)); /* IFA_CACHEINFO */
1695}
1696
1697static inline u32 cstamp_delta(unsigned long cstamp)
1698{
1699 return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
1700}
1701
1702static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
1703 unsigned long tstamp, u32 preferred, u32 valid)
1704{
1705 struct ifa_cacheinfo ci;
1706
1707 ci.cstamp = cstamp_delta(cstamp);
1708 ci.tstamp = cstamp_delta(tstamp);
1709 ci.ifa_prefered = preferred;
1710 ci.ifa_valid = valid;
1711
1712 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
1713}
1714
1715static int inet_fill_ifaddr(struct sk_buff *skb, const struct in_ifaddr *ifa,
1716 struct inet_fill_args *args)
1717{
1718 struct ifaddrmsg *ifm;
1719 struct nlmsghdr *nlh;
1720 unsigned long tstamp;
1721 u32 preferred, valid;
1722 u32 flags;
1723
1724 nlh = nlmsg_put(skb, args->portid, args->seq, args->event, sizeof(*ifm),
1725 args->flags);
1726 if (!nlh)
1727 return -EMSGSIZE;
1728
1729 ifm = nlmsg_data(nlh);
1730 ifm->ifa_family = AF_INET;
1731 ifm->ifa_prefixlen = ifa->ifa_prefixlen;
1732
1733 flags = READ_ONCE(ifa->ifa_flags);
1734 /* Warning : ifm->ifa_flags is an __u8, it holds only 8 bits.
1735 * The 32bit value is given in IFA_FLAGS attribute.
1736 */
1737 ifm->ifa_flags = (__u8)flags;
1738
1739 ifm->ifa_scope = ifa->ifa_scope;
1740 ifm->ifa_index = ifa->ifa_dev->dev->ifindex;
1741
1742 if (args->netnsid >= 0 &&
1743 nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
1744 goto nla_put_failure;
1745
1746 tstamp = READ_ONCE(ifa->ifa_tstamp);
1747 if (!(flags & IFA_F_PERMANENT)) {
1748 preferred = READ_ONCE(ifa->ifa_preferred_lft);
1749 valid = READ_ONCE(ifa->ifa_valid_lft);
1750 if (preferred != INFINITY_LIFE_TIME) {
1751 long tval = (jiffies - tstamp) / HZ;
1752
1753 if (preferred > tval)
1754 preferred -= tval;
1755 else
1756 preferred = 0;
1757 if (valid != INFINITY_LIFE_TIME) {
1758 if (valid > tval)
1759 valid -= tval;
1760 else
1761 valid = 0;
1762 }
1763 }
1764 } else {
1765 preferred = INFINITY_LIFE_TIME;
1766 valid = INFINITY_LIFE_TIME;
1767 }
1768 if ((ifa->ifa_address &&
1769 nla_put_in_addr(skb, IFA_ADDRESS, ifa->ifa_address)) ||
1770 (ifa->ifa_local &&
1771 nla_put_in_addr(skb, IFA_LOCAL, ifa->ifa_local)) ||
1772 (ifa->ifa_broadcast &&
1773 nla_put_in_addr(skb, IFA_BROADCAST, ifa->ifa_broadcast)) ||
1774 (ifa->ifa_label[0] &&
1775 nla_put_string(skb, IFA_LABEL, ifa->ifa_label)) ||
1776 (ifa->ifa_proto &&
1777 nla_put_u8(skb, IFA_PROTO, ifa->ifa_proto)) ||
1778 nla_put_u32(skb, IFA_FLAGS, flags) ||
1779 (ifa->ifa_rt_priority &&
1780 nla_put_u32(skb, IFA_RT_PRIORITY, ifa->ifa_rt_priority)) ||
1781 put_cacheinfo(skb, READ_ONCE(ifa->ifa_cstamp), tstamp,
1782 preferred, valid))
1783 goto nla_put_failure;
1784
1785 nlmsg_end(skb, nlh);
1786 return 0;
1787
1788nla_put_failure:
1789 nlmsg_cancel(skb, nlh);
1790 return -EMSGSIZE;
1791}
1792
1793static int inet_valid_dump_ifaddr_req(const struct nlmsghdr *nlh,
1794 struct inet_fill_args *fillargs,
1795 struct net **tgt_net, struct sock *sk,
1796 struct netlink_callback *cb)
1797{
1798 struct netlink_ext_ack *extack = cb->extack;
1799 struct nlattr *tb[IFA_MAX+1];
1800 struct ifaddrmsg *ifm;
1801 int err, i;
1802
1803 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
1804 NL_SET_ERR_MSG(extack, "ipv4: Invalid header for address dump request");
1805 return -EINVAL;
1806 }
1807
1808 ifm = nlmsg_data(nlh);
1809 if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
1810 NL_SET_ERR_MSG(extack, "ipv4: Invalid values in header for address dump request");
1811 return -EINVAL;
1812 }
1813
1814 fillargs->ifindex = ifm->ifa_index;
1815 if (fillargs->ifindex) {
1816 cb->answer_flags |= NLM_F_DUMP_FILTERED;
1817 fillargs->flags |= NLM_F_DUMP_FILTERED;
1818 }
1819
1820 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
1821 ifa_ipv4_policy, extack);
1822 if (err < 0)
1823 return err;
1824
1825 for (i = 0; i <= IFA_MAX; ++i) {
1826 if (!tb[i])
1827 continue;
1828
1829 if (i == IFA_TARGET_NETNSID) {
1830 struct net *net;
1831
1832 fillargs->netnsid = nla_get_s32(tb[i]);
1833
1834 net = rtnl_get_net_ns_capable(sk, fillargs->netnsid);
1835 if (IS_ERR(net)) {
1836 fillargs->netnsid = -1;
1837 NL_SET_ERR_MSG(extack, "ipv4: Invalid target network namespace id");
1838 return PTR_ERR(net);
1839 }
1840 *tgt_net = net;
1841 } else {
1842 NL_SET_ERR_MSG(extack, "ipv4: Unsupported attribute in dump request");
1843 return -EINVAL;
1844 }
1845 }
1846
1847 return 0;
1848}
1849
1850static int in_dev_dump_addr(struct in_device *in_dev, struct sk_buff *skb,
1851 struct netlink_callback *cb, int *s_ip_idx,
1852 struct inet_fill_args *fillargs)
1853{
1854 struct in_ifaddr *ifa;
1855 int ip_idx = 0;
1856 int err;
1857
1858 in_dev_for_each_ifa_rcu(ifa, in_dev) {
1859 if (ip_idx < *s_ip_idx) {
1860 ip_idx++;
1861 continue;
1862 }
1863 err = inet_fill_ifaddr(skb, ifa, fillargs);
1864 if (err < 0)
1865 goto done;
1866
1867 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1868 ip_idx++;
1869 }
1870 err = 0;
1871 ip_idx = 0;
1872done:
1873 *s_ip_idx = ip_idx;
1874
1875 return err;
1876}
1877
1878/* Combine dev_addr_genid and dev_base_seq to detect changes.
1879 */
1880static u32 inet_base_seq(const struct net *net)
1881{
1882 u32 res = atomic_read(&net->ipv4.dev_addr_genid) +
1883 READ_ONCE(net->dev_base_seq);
1884
1885 /* Must not return 0 (see nl_dump_check_consistent()).
1886 * Chose a value far away from 0.
1887 */
1888 if (!res)
1889 res = 0x80000000;
1890 return res;
1891}
1892
1893static int inet_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
1894{
1895 const struct nlmsghdr *nlh = cb->nlh;
1896 struct inet_fill_args fillargs = {
1897 .portid = NETLINK_CB(cb->skb).portid,
1898 .seq = nlh->nlmsg_seq,
1899 .event = RTM_NEWADDR,
1900 .flags = NLM_F_MULTI,
1901 .netnsid = -1,
1902 };
1903 struct net *net = sock_net(skb->sk);
1904 struct net *tgt_net = net;
1905 struct {
1906 unsigned long ifindex;
1907 int ip_idx;
1908 } *ctx = (void *)cb->ctx;
1909 struct in_device *in_dev;
1910 struct net_device *dev;
1911 int err = 0;
1912
1913 rcu_read_lock();
1914 if (cb->strict_check) {
1915 err = inet_valid_dump_ifaddr_req(nlh, &fillargs, &tgt_net,
1916 skb->sk, cb);
1917 if (err < 0)
1918 goto done;
1919
1920 if (fillargs.ifindex) {
1921 dev = dev_get_by_index_rcu(tgt_net, fillargs.ifindex);
1922 if (!dev) {
1923 err = -ENODEV;
1924 goto done;
1925 }
1926 in_dev = __in_dev_get_rcu(dev);
1927 if (!in_dev)
1928 goto done;
1929 err = in_dev_dump_addr(in_dev, skb, cb, &ctx->ip_idx,
1930 &fillargs);
1931 goto done;
1932 }
1933 }
1934
1935 cb->seq = inet_base_seq(tgt_net);
1936
1937 for_each_netdev_dump(tgt_net, dev, ctx->ifindex) {
1938 in_dev = __in_dev_get_rcu(dev);
1939 if (!in_dev)
1940 continue;
1941 err = in_dev_dump_addr(in_dev, skb, cb, &ctx->ip_idx,
1942 &fillargs);
1943 if (err < 0)
1944 goto done;
1945 }
1946done:
1947 if (fillargs.netnsid >= 0)
1948 put_net(tgt_net);
1949 rcu_read_unlock();
1950 return err;
1951}
1952
1953static void rtmsg_ifa(int event, struct in_ifaddr *ifa, struct nlmsghdr *nlh,
1954 u32 portid)
1955{
1956 struct inet_fill_args fillargs = {
1957 .portid = portid,
1958 .seq = nlh ? nlh->nlmsg_seq : 0,
1959 .event = event,
1960 .flags = 0,
1961 .netnsid = -1,
1962 };
1963 struct sk_buff *skb;
1964 int err = -ENOBUFS;
1965 struct net *net;
1966
1967 net = dev_net(ifa->ifa_dev->dev);
1968 skb = nlmsg_new(inet_nlmsg_size(), GFP_KERNEL);
1969 if (!skb)
1970 goto errout;
1971
1972 err = inet_fill_ifaddr(skb, ifa, &fillargs);
1973 if (err < 0) {
1974 /* -EMSGSIZE implies BUG in inet_nlmsg_size() */
1975 WARN_ON(err == -EMSGSIZE);
1976 kfree_skb(skb);
1977 goto errout;
1978 }
1979 rtnl_notify(skb, net, portid, RTNLGRP_IPV4_IFADDR, nlh, GFP_KERNEL);
1980 return;
1981errout:
1982 rtnl_set_sk_err(net, RTNLGRP_IPV4_IFADDR, err);
1983}
1984
1985static size_t inet_get_link_af_size(const struct net_device *dev,
1986 u32 ext_filter_mask)
1987{
1988 struct in_device *in_dev = rcu_dereference_rtnl(dev->ip_ptr);
1989
1990 if (!in_dev)
1991 return 0;
1992
1993 return nla_total_size(IPV4_DEVCONF_MAX * 4); /* IFLA_INET_CONF */
1994}
1995
1996static int inet_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
1997 u32 ext_filter_mask)
1998{
1999 struct in_device *in_dev = rcu_dereference_rtnl(dev->ip_ptr);
2000 struct nlattr *nla;
2001 int i;
2002
2003 if (!in_dev)
2004 return -ENODATA;
2005
2006 nla = nla_reserve(skb, IFLA_INET_CONF, IPV4_DEVCONF_MAX * 4);
2007 if (!nla)
2008 return -EMSGSIZE;
2009
2010 for (i = 0; i < IPV4_DEVCONF_MAX; i++)
2011 ((u32 *) nla_data(nla))[i] = READ_ONCE(in_dev->cnf.data[i]);
2012
2013 return 0;
2014}
2015
2016static const struct nla_policy inet_af_policy[IFLA_INET_MAX+1] = {
2017 [IFLA_INET_CONF] = { .type = NLA_NESTED },
2018};
2019
2020static int inet_validate_link_af(const struct net_device *dev,
2021 const struct nlattr *nla,
2022 struct netlink_ext_ack *extack)
2023{
2024 struct nlattr *a, *tb[IFLA_INET_MAX+1];
2025 int err, rem;
2026
2027 if (dev && !__in_dev_get_rtnl(dev))
2028 return -EAFNOSUPPORT;
2029
2030 err = nla_parse_nested_deprecated(tb, IFLA_INET_MAX, nla,
2031 inet_af_policy, extack);
2032 if (err < 0)
2033 return err;
2034
2035 if (tb[IFLA_INET_CONF]) {
2036 nla_for_each_nested(a, tb[IFLA_INET_CONF], rem) {
2037 int cfgid = nla_type(a);
2038
2039 if (nla_len(a) < 4)
2040 return -EINVAL;
2041
2042 if (cfgid <= 0 || cfgid > IPV4_DEVCONF_MAX)
2043 return -EINVAL;
2044 }
2045 }
2046
2047 return 0;
2048}
2049
2050static int inet_set_link_af(struct net_device *dev, const struct nlattr *nla,
2051 struct netlink_ext_ack *extack)
2052{
2053 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2054 struct nlattr *a, *tb[IFLA_INET_MAX+1];
2055 int rem;
2056
2057 if (!in_dev)
2058 return -EAFNOSUPPORT;
2059
2060 if (nla_parse_nested_deprecated(tb, IFLA_INET_MAX, nla, NULL, NULL) < 0)
2061 return -EINVAL;
2062
2063 if (tb[IFLA_INET_CONF]) {
2064 nla_for_each_nested(a, tb[IFLA_INET_CONF], rem)
2065 ipv4_devconf_set(in_dev, nla_type(a), nla_get_u32(a));
2066 }
2067
2068 return 0;
2069}
2070
2071static int inet_netconf_msgsize_devconf(int type)
2072{
2073 int size = NLMSG_ALIGN(sizeof(struct netconfmsg))
2074 + nla_total_size(4); /* NETCONFA_IFINDEX */
2075 bool all = false;
2076
2077 if (type == NETCONFA_ALL)
2078 all = true;
2079
2080 if (all || type == NETCONFA_FORWARDING)
2081 size += nla_total_size(4);
2082 if (all || type == NETCONFA_RP_FILTER)
2083 size += nla_total_size(4);
2084 if (all || type == NETCONFA_MC_FORWARDING)
2085 size += nla_total_size(4);
2086 if (all || type == NETCONFA_BC_FORWARDING)
2087 size += nla_total_size(4);
2088 if (all || type == NETCONFA_PROXY_NEIGH)
2089 size += nla_total_size(4);
2090 if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
2091 size += nla_total_size(4);
2092
2093 return size;
2094}
2095
2096static int inet_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
2097 const struct ipv4_devconf *devconf,
2098 u32 portid, u32 seq, int event,
2099 unsigned int flags, int type)
2100{
2101 struct nlmsghdr *nlh;
2102 struct netconfmsg *ncm;
2103 bool all = false;
2104
2105 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
2106 flags);
2107 if (!nlh)
2108 return -EMSGSIZE;
2109
2110 if (type == NETCONFA_ALL)
2111 all = true;
2112
2113 ncm = nlmsg_data(nlh);
2114 ncm->ncm_family = AF_INET;
2115
2116 if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
2117 goto nla_put_failure;
2118
2119 if (!devconf)
2120 goto out;
2121
2122 if ((all || type == NETCONFA_FORWARDING) &&
2123 nla_put_s32(skb, NETCONFA_FORWARDING,
2124 IPV4_DEVCONF_RO(*devconf, FORWARDING)) < 0)
2125 goto nla_put_failure;
2126 if ((all || type == NETCONFA_RP_FILTER) &&
2127 nla_put_s32(skb, NETCONFA_RP_FILTER,
2128 IPV4_DEVCONF_RO(*devconf, RP_FILTER)) < 0)
2129 goto nla_put_failure;
2130 if ((all || type == NETCONFA_MC_FORWARDING) &&
2131 nla_put_s32(skb, NETCONFA_MC_FORWARDING,
2132 IPV4_DEVCONF_RO(*devconf, MC_FORWARDING)) < 0)
2133 goto nla_put_failure;
2134 if ((all || type == NETCONFA_BC_FORWARDING) &&
2135 nla_put_s32(skb, NETCONFA_BC_FORWARDING,
2136 IPV4_DEVCONF_RO(*devconf, BC_FORWARDING)) < 0)
2137 goto nla_put_failure;
2138 if ((all || type == NETCONFA_PROXY_NEIGH) &&
2139 nla_put_s32(skb, NETCONFA_PROXY_NEIGH,
2140 IPV4_DEVCONF_RO(*devconf, PROXY_ARP)) < 0)
2141 goto nla_put_failure;
2142 if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
2143 nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
2144 IPV4_DEVCONF_RO(*devconf,
2145 IGNORE_ROUTES_WITH_LINKDOWN)) < 0)
2146 goto nla_put_failure;
2147
2148out:
2149 nlmsg_end(skb, nlh);
2150 return 0;
2151
2152nla_put_failure:
2153 nlmsg_cancel(skb, nlh);
2154 return -EMSGSIZE;
2155}
2156
2157void inet_netconf_notify_devconf(struct net *net, int event, int type,
2158 int ifindex, struct ipv4_devconf *devconf)
2159{
2160 struct sk_buff *skb;
2161 int err = -ENOBUFS;
2162
2163 skb = nlmsg_new(inet_netconf_msgsize_devconf(type), GFP_KERNEL);
2164 if (!skb)
2165 goto errout;
2166
2167 err = inet_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
2168 event, 0, type);
2169 if (err < 0) {
2170 /* -EMSGSIZE implies BUG in inet_netconf_msgsize_devconf() */
2171 WARN_ON(err == -EMSGSIZE);
2172 kfree_skb(skb);
2173 goto errout;
2174 }
2175 rtnl_notify(skb, net, 0, RTNLGRP_IPV4_NETCONF, NULL, GFP_KERNEL);
2176 return;
2177errout:
2178 rtnl_set_sk_err(net, RTNLGRP_IPV4_NETCONF, err);
2179}
2180
2181static const struct nla_policy devconf_ipv4_policy[NETCONFA_MAX+1] = {
2182 [NETCONFA_IFINDEX] = { .len = sizeof(int) },
2183 [NETCONFA_FORWARDING] = { .len = sizeof(int) },
2184 [NETCONFA_RP_FILTER] = { .len = sizeof(int) },
2185 [NETCONFA_PROXY_NEIGH] = { .len = sizeof(int) },
2186 [NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN] = { .len = sizeof(int) },
2187};
2188
2189static int inet_netconf_valid_get_req(struct sk_buff *skb,
2190 const struct nlmsghdr *nlh,
2191 struct nlattr **tb,
2192 struct netlink_ext_ack *extack)
2193{
2194 int i, err;
2195
2196 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(struct netconfmsg))) {
2197 NL_SET_ERR_MSG(extack, "ipv4: Invalid header for netconf get request");
2198 return -EINVAL;
2199 }
2200
2201 if (!netlink_strict_get_check(skb))
2202 return nlmsg_parse_deprecated(nlh, sizeof(struct netconfmsg),
2203 tb, NETCONFA_MAX,
2204 devconf_ipv4_policy, extack);
2205
2206 err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct netconfmsg),
2207 tb, NETCONFA_MAX,
2208 devconf_ipv4_policy, extack);
2209 if (err)
2210 return err;
2211
2212 for (i = 0; i <= NETCONFA_MAX; i++) {
2213 if (!tb[i])
2214 continue;
2215
2216 switch (i) {
2217 case NETCONFA_IFINDEX:
2218 break;
2219 default:
2220 NL_SET_ERR_MSG(extack, "ipv4: Unsupported attribute in netconf get request");
2221 return -EINVAL;
2222 }
2223 }
2224
2225 return 0;
2226}
2227
2228static int inet_netconf_get_devconf(struct sk_buff *in_skb,
2229 struct nlmsghdr *nlh,
2230 struct netlink_ext_ack *extack)
2231{
2232 struct net *net = sock_net(in_skb->sk);
2233 struct nlattr *tb[NETCONFA_MAX + 1];
2234 const struct ipv4_devconf *devconf;
2235 struct in_device *in_dev = NULL;
2236 struct net_device *dev = NULL;
2237 struct sk_buff *skb;
2238 int ifindex;
2239 int err;
2240
2241 err = inet_netconf_valid_get_req(in_skb, nlh, tb, extack);
2242 if (err)
2243 return err;
2244
2245 if (!tb[NETCONFA_IFINDEX])
2246 return -EINVAL;
2247
2248 ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
2249 switch (ifindex) {
2250 case NETCONFA_IFINDEX_ALL:
2251 devconf = net->ipv4.devconf_all;
2252 break;
2253 case NETCONFA_IFINDEX_DEFAULT:
2254 devconf = net->ipv4.devconf_dflt;
2255 break;
2256 default:
2257 err = -ENODEV;
2258 dev = dev_get_by_index(net, ifindex);
2259 if (dev)
2260 in_dev = in_dev_get(dev);
2261 if (!in_dev)
2262 goto errout;
2263 devconf = &in_dev->cnf;
2264 break;
2265 }
2266
2267 err = -ENOBUFS;
2268 skb = nlmsg_new(inet_netconf_msgsize_devconf(NETCONFA_ALL), GFP_KERNEL);
2269 if (!skb)
2270 goto errout;
2271
2272 err = inet_netconf_fill_devconf(skb, ifindex, devconf,
2273 NETLINK_CB(in_skb).portid,
2274 nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
2275 NETCONFA_ALL);
2276 if (err < 0) {
2277 /* -EMSGSIZE implies BUG in inet_netconf_msgsize_devconf() */
2278 WARN_ON(err == -EMSGSIZE);
2279 kfree_skb(skb);
2280 goto errout;
2281 }
2282 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
2283errout:
2284 if (in_dev)
2285 in_dev_put(in_dev);
2286 dev_put(dev);
2287 return err;
2288}
2289
2290static int inet_netconf_dump_devconf(struct sk_buff *skb,
2291 struct netlink_callback *cb)
2292{
2293 const struct nlmsghdr *nlh = cb->nlh;
2294 struct net *net = sock_net(skb->sk);
2295 struct {
2296 unsigned long ifindex;
2297 unsigned int all_default;
2298 } *ctx = (void *)cb->ctx;
2299 const struct in_device *in_dev;
2300 struct net_device *dev;
2301 int err = 0;
2302
2303 if (cb->strict_check) {
2304 struct netlink_ext_ack *extack = cb->extack;
2305 struct netconfmsg *ncm;
2306
2307 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ncm))) {
2308 NL_SET_ERR_MSG(extack, "ipv4: Invalid header for netconf dump request");
2309 return -EINVAL;
2310 }
2311
2312 if (nlmsg_attrlen(nlh, sizeof(*ncm))) {
2313 NL_SET_ERR_MSG(extack, "ipv4: Invalid data after header in netconf dump request");
2314 return -EINVAL;
2315 }
2316 }
2317
2318 rcu_read_lock();
2319 for_each_netdev_dump(net, dev, ctx->ifindex) {
2320 in_dev = __in_dev_get_rcu(dev);
2321 if (!in_dev)
2322 continue;
2323 err = inet_netconf_fill_devconf(skb, dev->ifindex,
2324 &in_dev->cnf,
2325 NETLINK_CB(cb->skb).portid,
2326 nlh->nlmsg_seq,
2327 RTM_NEWNETCONF, NLM_F_MULTI,
2328 NETCONFA_ALL);
2329 if (err < 0)
2330 goto done;
2331 }
2332 if (ctx->all_default == 0) {
2333 err = inet_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
2334 net->ipv4.devconf_all,
2335 NETLINK_CB(cb->skb).portid,
2336 nlh->nlmsg_seq,
2337 RTM_NEWNETCONF, NLM_F_MULTI,
2338 NETCONFA_ALL);
2339 if (err < 0)
2340 goto done;
2341 ctx->all_default++;
2342 }
2343 if (ctx->all_default == 1) {
2344 err = inet_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
2345 net->ipv4.devconf_dflt,
2346 NETLINK_CB(cb->skb).portid,
2347 nlh->nlmsg_seq,
2348 RTM_NEWNETCONF, NLM_F_MULTI,
2349 NETCONFA_ALL);
2350 if (err < 0)
2351 goto done;
2352 ctx->all_default++;
2353 }
2354done:
2355 rcu_read_unlock();
2356 return err;
2357}
2358
2359#ifdef CONFIG_SYSCTL
2360
2361static void devinet_copy_dflt_conf(struct net *net, int i)
2362{
2363 struct net_device *dev;
2364
2365 rcu_read_lock();
2366 for_each_netdev_rcu(net, dev) {
2367 struct in_device *in_dev;
2368
2369 in_dev = __in_dev_get_rcu(dev);
2370 if (in_dev && !test_bit(i, in_dev->cnf.state))
2371 in_dev->cnf.data[i] = net->ipv4.devconf_dflt->data[i];
2372 }
2373 rcu_read_unlock();
2374}
2375
2376/* called with RTNL locked */
2377static void inet_forward_change(struct net *net)
2378{
2379 struct net_device *dev;
2380 int on = IPV4_DEVCONF_ALL(net, FORWARDING);
2381
2382 IPV4_DEVCONF_ALL(net, ACCEPT_REDIRECTS) = !on;
2383 IPV4_DEVCONF_DFLT(net, FORWARDING) = on;
2384 inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
2385 NETCONFA_FORWARDING,
2386 NETCONFA_IFINDEX_ALL,
2387 net->ipv4.devconf_all);
2388 inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
2389 NETCONFA_FORWARDING,
2390 NETCONFA_IFINDEX_DEFAULT,
2391 net->ipv4.devconf_dflt);
2392
2393 for_each_netdev(net, dev) {
2394 struct in_device *in_dev;
2395
2396 if (on)
2397 dev_disable_lro(dev);
2398
2399 in_dev = __in_dev_get_rtnl_net(dev);
2400 if (in_dev) {
2401 IN_DEV_CONF_SET(in_dev, FORWARDING, on);
2402 inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
2403 NETCONFA_FORWARDING,
2404 dev->ifindex, &in_dev->cnf);
2405 }
2406 }
2407}
2408
2409static int devinet_conf_ifindex(struct net *net, struct ipv4_devconf *cnf)
2410{
2411 if (cnf == net->ipv4.devconf_dflt)
2412 return NETCONFA_IFINDEX_DEFAULT;
2413 else if (cnf == net->ipv4.devconf_all)
2414 return NETCONFA_IFINDEX_ALL;
2415 else {
2416 struct in_device *idev
2417 = container_of(cnf, struct in_device, cnf);
2418 return idev->dev->ifindex;
2419 }
2420}
2421
2422static int devinet_conf_proc(const struct ctl_table *ctl, int write,
2423 void *buffer, size_t *lenp, loff_t *ppos)
2424{
2425 int old_value = *(int *)ctl->data;
2426 int ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
2427 int new_value = *(int *)ctl->data;
2428
2429 if (write) {
2430 struct ipv4_devconf *cnf = ctl->extra1;
2431 struct net *net = ctl->extra2;
2432 int i = (int *)ctl->data - cnf->data;
2433 int ifindex;
2434
2435 set_bit(i, cnf->state);
2436
2437 if (cnf == net->ipv4.devconf_dflt)
2438 devinet_copy_dflt_conf(net, i);
2439 if (i == IPV4_DEVCONF_ACCEPT_LOCAL - 1 ||
2440 i == IPV4_DEVCONF_ROUTE_LOCALNET - 1)
2441 if ((new_value == 0) && (old_value != 0))
2442 rt_cache_flush(net);
2443
2444 if (i == IPV4_DEVCONF_BC_FORWARDING - 1 &&
2445 new_value != old_value)
2446 rt_cache_flush(net);
2447
2448 if (i == IPV4_DEVCONF_RP_FILTER - 1 &&
2449 new_value != old_value) {
2450 ifindex = devinet_conf_ifindex(net, cnf);
2451 inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
2452 NETCONFA_RP_FILTER,
2453 ifindex, cnf);
2454 }
2455 if (i == IPV4_DEVCONF_PROXY_ARP - 1 &&
2456 new_value != old_value) {
2457 ifindex = devinet_conf_ifindex(net, cnf);
2458 inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
2459 NETCONFA_PROXY_NEIGH,
2460 ifindex, cnf);
2461 }
2462 if (i == IPV4_DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN - 1 &&
2463 new_value != old_value) {
2464 ifindex = devinet_conf_ifindex(net, cnf);
2465 inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
2466 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
2467 ifindex, cnf);
2468 }
2469 }
2470
2471 return ret;
2472}
2473
2474static int devinet_sysctl_forward(const struct ctl_table *ctl, int write,
2475 void *buffer, size_t *lenp, loff_t *ppos)
2476{
2477 int *valp = ctl->data;
2478 int val = *valp;
2479 loff_t pos = *ppos;
2480 struct net *net = ctl->extra2;
2481 int ret;
2482
2483 if (write && !ns_capable(net->user_ns, CAP_NET_ADMIN))
2484 return -EPERM;
2485
2486 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
2487
2488 if (write && *valp != val) {
2489 if (valp != &IPV4_DEVCONF_DFLT(net, FORWARDING)) {
2490 if (!rtnl_net_trylock(net)) {
2491 /* Restore the original values before restarting */
2492 *valp = val;
2493 *ppos = pos;
2494 return restart_syscall();
2495 }
2496 if (valp == &IPV4_DEVCONF_ALL(net, FORWARDING)) {
2497 inet_forward_change(net);
2498 } else {
2499 struct ipv4_devconf *cnf = ctl->extra1;
2500 struct in_device *idev =
2501 container_of(cnf, struct in_device, cnf);
2502 if (*valp)
2503 dev_disable_lro(idev->dev);
2504 inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
2505 NETCONFA_FORWARDING,
2506 idev->dev->ifindex,
2507 cnf);
2508 }
2509 rtnl_net_unlock(net);
2510 rt_cache_flush(net);
2511 } else
2512 inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
2513 NETCONFA_FORWARDING,
2514 NETCONFA_IFINDEX_DEFAULT,
2515 net->ipv4.devconf_dflt);
2516 }
2517
2518 return ret;
2519}
2520
2521static int ipv4_doint_and_flush(const struct ctl_table *ctl, int write,
2522 void *buffer, size_t *lenp, loff_t *ppos)
2523{
2524 int *valp = ctl->data;
2525 int val = *valp;
2526 int ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
2527 struct net *net = ctl->extra2;
2528
2529 if (write && *valp != val)
2530 rt_cache_flush(net);
2531
2532 return ret;
2533}
2534
2535#define DEVINET_SYSCTL_ENTRY(attr, name, mval, proc) \
2536 { \
2537 .procname = name, \
2538 .data = ipv4_devconf.data + \
2539 IPV4_DEVCONF_ ## attr - 1, \
2540 .maxlen = sizeof(int), \
2541 .mode = mval, \
2542 .proc_handler = proc, \
2543 .extra1 = &ipv4_devconf, \
2544 }
2545
2546#define DEVINET_SYSCTL_RW_ENTRY(attr, name) \
2547 DEVINET_SYSCTL_ENTRY(attr, name, 0644, devinet_conf_proc)
2548
2549#define DEVINET_SYSCTL_RO_ENTRY(attr, name) \
2550 DEVINET_SYSCTL_ENTRY(attr, name, 0444, devinet_conf_proc)
2551
2552#define DEVINET_SYSCTL_COMPLEX_ENTRY(attr, name, proc) \
2553 DEVINET_SYSCTL_ENTRY(attr, name, 0644, proc)
2554
2555#define DEVINET_SYSCTL_FLUSHING_ENTRY(attr, name) \
2556 DEVINET_SYSCTL_COMPLEX_ENTRY(attr, name, ipv4_doint_and_flush)
2557
2558static struct devinet_sysctl_table {
2559 struct ctl_table_header *sysctl_header;
2560 struct ctl_table devinet_vars[IPV4_DEVCONF_MAX];
2561} devinet_sysctl = {
2562 .devinet_vars = {
2563 DEVINET_SYSCTL_COMPLEX_ENTRY(FORWARDING, "forwarding",
2564 devinet_sysctl_forward),
2565 DEVINET_SYSCTL_RO_ENTRY(MC_FORWARDING, "mc_forwarding"),
2566 DEVINET_SYSCTL_RW_ENTRY(BC_FORWARDING, "bc_forwarding"),
2567
2568 DEVINET_SYSCTL_RW_ENTRY(ACCEPT_REDIRECTS, "accept_redirects"),
2569 DEVINET_SYSCTL_RW_ENTRY(SECURE_REDIRECTS, "secure_redirects"),
2570 DEVINET_SYSCTL_RW_ENTRY(SHARED_MEDIA, "shared_media"),
2571 DEVINET_SYSCTL_RW_ENTRY(RP_FILTER, "rp_filter"),
2572 DEVINET_SYSCTL_RW_ENTRY(SEND_REDIRECTS, "send_redirects"),
2573 DEVINET_SYSCTL_RW_ENTRY(ACCEPT_SOURCE_ROUTE,
2574 "accept_source_route"),
2575 DEVINET_SYSCTL_RW_ENTRY(ACCEPT_LOCAL, "accept_local"),
2576 DEVINET_SYSCTL_RW_ENTRY(SRC_VMARK, "src_valid_mark"),
2577 DEVINET_SYSCTL_RW_ENTRY(PROXY_ARP, "proxy_arp"),
2578 DEVINET_SYSCTL_RW_ENTRY(MEDIUM_ID, "medium_id"),
2579 DEVINET_SYSCTL_RW_ENTRY(BOOTP_RELAY, "bootp_relay"),
2580 DEVINET_SYSCTL_RW_ENTRY(LOG_MARTIANS, "log_martians"),
2581 DEVINET_SYSCTL_RW_ENTRY(TAG, "tag"),
2582 DEVINET_SYSCTL_RW_ENTRY(ARPFILTER, "arp_filter"),
2583 DEVINET_SYSCTL_RW_ENTRY(ARP_ANNOUNCE, "arp_announce"),
2584 DEVINET_SYSCTL_RW_ENTRY(ARP_IGNORE, "arp_ignore"),
2585 DEVINET_SYSCTL_RW_ENTRY(ARP_ACCEPT, "arp_accept"),
2586 DEVINET_SYSCTL_RW_ENTRY(ARP_NOTIFY, "arp_notify"),
2587 DEVINET_SYSCTL_RW_ENTRY(ARP_EVICT_NOCARRIER,
2588 "arp_evict_nocarrier"),
2589 DEVINET_SYSCTL_RW_ENTRY(PROXY_ARP_PVLAN, "proxy_arp_pvlan"),
2590 DEVINET_SYSCTL_RW_ENTRY(FORCE_IGMP_VERSION,
2591 "force_igmp_version"),
2592 DEVINET_SYSCTL_RW_ENTRY(IGMPV2_UNSOLICITED_REPORT_INTERVAL,
2593 "igmpv2_unsolicited_report_interval"),
2594 DEVINET_SYSCTL_RW_ENTRY(IGMPV3_UNSOLICITED_REPORT_INTERVAL,
2595 "igmpv3_unsolicited_report_interval"),
2596 DEVINET_SYSCTL_RW_ENTRY(IGNORE_ROUTES_WITH_LINKDOWN,
2597 "ignore_routes_with_linkdown"),
2598 DEVINET_SYSCTL_RW_ENTRY(DROP_GRATUITOUS_ARP,
2599 "drop_gratuitous_arp"),
2600
2601 DEVINET_SYSCTL_FLUSHING_ENTRY(NOXFRM, "disable_xfrm"),
2602 DEVINET_SYSCTL_FLUSHING_ENTRY(NOPOLICY, "disable_policy"),
2603 DEVINET_SYSCTL_FLUSHING_ENTRY(PROMOTE_SECONDARIES,
2604 "promote_secondaries"),
2605 DEVINET_SYSCTL_FLUSHING_ENTRY(ROUTE_LOCALNET,
2606 "route_localnet"),
2607 DEVINET_SYSCTL_FLUSHING_ENTRY(DROP_UNICAST_IN_L2_MULTICAST,
2608 "drop_unicast_in_l2_multicast"),
2609 },
2610};
2611
2612static int __devinet_sysctl_register(struct net *net, char *dev_name,
2613 int ifindex, struct ipv4_devconf *p)
2614{
2615 int i;
2616 struct devinet_sysctl_table *t;
2617 char path[sizeof("net/ipv4/conf/") + IFNAMSIZ];
2618
2619 t = kmemdup(&devinet_sysctl, sizeof(*t), GFP_KERNEL_ACCOUNT);
2620 if (!t)
2621 goto out;
2622
2623 for (i = 0; i < ARRAY_SIZE(t->devinet_vars); i++) {
2624 t->devinet_vars[i].data += (char *)p - (char *)&ipv4_devconf;
2625 t->devinet_vars[i].extra1 = p;
2626 t->devinet_vars[i].extra2 = net;
2627 }
2628
2629 snprintf(path, sizeof(path), "net/ipv4/conf/%s", dev_name);
2630
2631 t->sysctl_header = register_net_sysctl(net, path, t->devinet_vars);
2632 if (!t->sysctl_header)
2633 goto free;
2634
2635 p->sysctl = t;
2636
2637 inet_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
2638 ifindex, p);
2639 return 0;
2640
2641free:
2642 kfree(t);
2643out:
2644 return -ENOMEM;
2645}
2646
2647static void __devinet_sysctl_unregister(struct net *net,
2648 struct ipv4_devconf *cnf, int ifindex)
2649{
2650 struct devinet_sysctl_table *t = cnf->sysctl;
2651
2652 if (t) {
2653 cnf->sysctl = NULL;
2654 unregister_net_sysctl_table(t->sysctl_header);
2655 kfree(t);
2656 }
2657
2658 inet_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
2659}
2660
2661static int devinet_sysctl_register(struct in_device *idev)
2662{
2663 int err;
2664
2665 if (!sysctl_dev_name_is_allowed(idev->dev->name))
2666 return -EINVAL;
2667
2668 err = neigh_sysctl_register(idev->dev, idev->arp_parms, NULL);
2669 if (err)
2670 return err;
2671 err = __devinet_sysctl_register(dev_net(idev->dev), idev->dev->name,
2672 idev->dev->ifindex, &idev->cnf);
2673 if (err)
2674 neigh_sysctl_unregister(idev->arp_parms);
2675 return err;
2676}
2677
2678static void devinet_sysctl_unregister(struct in_device *idev)
2679{
2680 struct net *net = dev_net(idev->dev);
2681
2682 __devinet_sysctl_unregister(net, &idev->cnf, idev->dev->ifindex);
2683 neigh_sysctl_unregister(idev->arp_parms);
2684}
2685
2686static struct ctl_table ctl_forward_entry[] = {
2687 {
2688 .procname = "ip_forward",
2689 .data = &ipv4_devconf.data[
2690 IPV4_DEVCONF_FORWARDING - 1],
2691 .maxlen = sizeof(int),
2692 .mode = 0644,
2693 .proc_handler = devinet_sysctl_forward,
2694 .extra1 = &ipv4_devconf,
2695 .extra2 = &init_net,
2696 },
2697};
2698#endif
2699
2700static __net_init int devinet_init_net(struct net *net)
2701{
2702#ifdef CONFIG_SYSCTL
2703 struct ctl_table_header *forw_hdr;
2704 struct ctl_table *tbl;
2705#endif
2706 struct ipv4_devconf *all, *dflt;
2707 int err;
2708 int i;
2709
2710 err = -ENOMEM;
2711 net->ipv4.inet_addr_lst = kmalloc_array(IN4_ADDR_HSIZE,
2712 sizeof(struct hlist_head),
2713 GFP_KERNEL);
2714 if (!net->ipv4.inet_addr_lst)
2715 goto err_alloc_hash;
2716
2717 all = kmemdup(&ipv4_devconf, sizeof(ipv4_devconf), GFP_KERNEL);
2718 if (!all)
2719 goto err_alloc_all;
2720
2721 dflt = kmemdup(&ipv4_devconf_dflt, sizeof(ipv4_devconf_dflt), GFP_KERNEL);
2722 if (!dflt)
2723 goto err_alloc_dflt;
2724
2725#ifdef CONFIG_SYSCTL
2726 tbl = kmemdup(ctl_forward_entry, sizeof(ctl_forward_entry), GFP_KERNEL);
2727 if (!tbl)
2728 goto err_alloc_ctl;
2729
2730 tbl[0].data = &all->data[IPV4_DEVCONF_FORWARDING - 1];
2731 tbl[0].extra1 = all;
2732 tbl[0].extra2 = net;
2733#endif
2734
2735 if (!net_eq(net, &init_net)) {
2736 switch (net_inherit_devconf()) {
2737 case 3:
2738 /* copy from the current netns */
2739 memcpy(all, current->nsproxy->net_ns->ipv4.devconf_all,
2740 sizeof(ipv4_devconf));
2741 memcpy(dflt,
2742 current->nsproxy->net_ns->ipv4.devconf_dflt,
2743 sizeof(ipv4_devconf_dflt));
2744 break;
2745 case 0:
2746 case 1:
2747 /* copy from init_net */
2748 memcpy(all, init_net.ipv4.devconf_all,
2749 sizeof(ipv4_devconf));
2750 memcpy(dflt, init_net.ipv4.devconf_dflt,
2751 sizeof(ipv4_devconf_dflt));
2752 break;
2753 case 2:
2754 /* use compiled values */
2755 break;
2756 }
2757 }
2758
2759#ifdef CONFIG_SYSCTL
2760 err = __devinet_sysctl_register(net, "all", NETCONFA_IFINDEX_ALL, all);
2761 if (err < 0)
2762 goto err_reg_all;
2763
2764 err = __devinet_sysctl_register(net, "default",
2765 NETCONFA_IFINDEX_DEFAULT, dflt);
2766 if (err < 0)
2767 goto err_reg_dflt;
2768
2769 err = -ENOMEM;
2770 forw_hdr = register_net_sysctl_sz(net, "net/ipv4", tbl,
2771 ARRAY_SIZE(ctl_forward_entry));
2772 if (!forw_hdr)
2773 goto err_reg_ctl;
2774 net->ipv4.forw_hdr = forw_hdr;
2775#endif
2776
2777 for (i = 0; i < IN4_ADDR_HSIZE; i++)
2778 INIT_HLIST_HEAD(&net->ipv4.inet_addr_lst[i]);
2779
2780 INIT_DEFERRABLE_WORK(&net->ipv4.addr_chk_work, check_lifetime);
2781
2782 net->ipv4.devconf_all = all;
2783 net->ipv4.devconf_dflt = dflt;
2784 return 0;
2785
2786#ifdef CONFIG_SYSCTL
2787err_reg_ctl:
2788 __devinet_sysctl_unregister(net, dflt, NETCONFA_IFINDEX_DEFAULT);
2789err_reg_dflt:
2790 __devinet_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
2791err_reg_all:
2792 kfree(tbl);
2793err_alloc_ctl:
2794#endif
2795 kfree(dflt);
2796err_alloc_dflt:
2797 kfree(all);
2798err_alloc_all:
2799 kfree(net->ipv4.inet_addr_lst);
2800err_alloc_hash:
2801 return err;
2802}
2803
2804static __net_exit void devinet_exit_net(struct net *net)
2805{
2806#ifdef CONFIG_SYSCTL
2807 const struct ctl_table *tbl;
2808#endif
2809
2810 cancel_delayed_work_sync(&net->ipv4.addr_chk_work);
2811
2812#ifdef CONFIG_SYSCTL
2813 tbl = net->ipv4.forw_hdr->ctl_table_arg;
2814 unregister_net_sysctl_table(net->ipv4.forw_hdr);
2815 __devinet_sysctl_unregister(net, net->ipv4.devconf_dflt,
2816 NETCONFA_IFINDEX_DEFAULT);
2817 __devinet_sysctl_unregister(net, net->ipv4.devconf_all,
2818 NETCONFA_IFINDEX_ALL);
2819 kfree(tbl);
2820#endif
2821 kfree(net->ipv4.devconf_dflt);
2822 kfree(net->ipv4.devconf_all);
2823 kfree(net->ipv4.inet_addr_lst);
2824}
2825
2826static __net_initdata struct pernet_operations devinet_ops = {
2827 .init = devinet_init_net,
2828 .exit = devinet_exit_net,
2829};
2830
2831static struct rtnl_af_ops inet_af_ops __read_mostly = {
2832 .family = AF_INET,
2833 .fill_link_af = inet_fill_link_af,
2834 .get_link_af_size = inet_get_link_af_size,
2835 .validate_link_af = inet_validate_link_af,
2836 .set_link_af = inet_set_link_af,
2837};
2838
2839static const struct rtnl_msg_handler devinet_rtnl_msg_handlers[] __initconst = {
2840 {.protocol = PF_INET, .msgtype = RTM_NEWADDR, .doit = inet_rtm_newaddr,
2841 .flags = RTNL_FLAG_DOIT_PERNET},
2842 {.protocol = PF_INET, .msgtype = RTM_DELADDR, .doit = inet_rtm_deladdr,
2843 .flags = RTNL_FLAG_DOIT_PERNET},
2844 {.protocol = PF_INET, .msgtype = RTM_GETADDR, .dumpit = inet_dump_ifaddr,
2845 .flags = RTNL_FLAG_DUMP_UNLOCKED | RTNL_FLAG_DUMP_SPLIT_NLM_DONE},
2846 {.protocol = PF_INET, .msgtype = RTM_GETNETCONF,
2847 .doit = inet_netconf_get_devconf, .dumpit = inet_netconf_dump_devconf,
2848 .flags = RTNL_FLAG_DOIT_UNLOCKED | RTNL_FLAG_DUMP_UNLOCKED},
2849};
2850
2851void __init devinet_init(void)
2852{
2853 register_pernet_subsys(&devinet_ops);
2854 register_netdevice_notifier(&ip_netdev_notifier);
2855
2856 if (rtnl_af_register(&inet_af_ops))
2857 panic("Unable to register inet_af_ops\n");
2858
2859 rtnl_register_many(devinet_rtnl_msg_handlers);
2860}
1// SPDX-License-Identifier: GPL-2.0-or-later
2/*
3 * NET3 IP device support routines.
4 *
5 * Derived from the IP parts of dev.c 1.0.19
6 * Authors: Ross Biro
7 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
8 * Mark Evans, <evansmp@uhura.aston.ac.uk>
9 *
10 * Additional Authors:
11 * Alan Cox, <gw4pts@gw4pts.ampr.org>
12 * Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
13 *
14 * Changes:
15 * Alexey Kuznetsov: pa_* fields are replaced with ifaddr
16 * lists.
17 * Cyrus Durgin: updated for kmod
18 * Matthias Andree: in devinet_ioctl, compare label and
19 * address (4.4BSD alias style support),
20 * fall back to comparing just the label
21 * if no match found.
22 */
23
24
25#include <linux/uaccess.h>
26#include <linux/bitops.h>
27#include <linux/capability.h>
28#include <linux/module.h>
29#include <linux/types.h>
30#include <linux/kernel.h>
31#include <linux/sched/signal.h>
32#include <linux/string.h>
33#include <linux/mm.h>
34#include <linux/socket.h>
35#include <linux/sockios.h>
36#include <linux/in.h>
37#include <linux/errno.h>
38#include <linux/interrupt.h>
39#include <linux/if_addr.h>
40#include <linux/if_ether.h>
41#include <linux/inet.h>
42#include <linux/netdevice.h>
43#include <linux/etherdevice.h>
44#include <linux/skbuff.h>
45#include <linux/init.h>
46#include <linux/notifier.h>
47#include <linux/inetdevice.h>
48#include <linux/igmp.h>
49#include <linux/slab.h>
50#include <linux/hash.h>
51#ifdef CONFIG_SYSCTL
52#include <linux/sysctl.h>
53#endif
54#include <linux/kmod.h>
55#include <linux/netconf.h>
56
57#include <net/arp.h>
58#include <net/ip.h>
59#include <net/route.h>
60#include <net/ip_fib.h>
61#include <net/rtnetlink.h>
62#include <net/net_namespace.h>
63#include <net/addrconf.h>
64
65#define IPV6ONLY_FLAGS \
66 (IFA_F_NODAD | IFA_F_OPTIMISTIC | IFA_F_DADFAILED | \
67 IFA_F_HOMEADDRESS | IFA_F_TENTATIVE | \
68 IFA_F_MANAGETEMPADDR | IFA_F_STABLE_PRIVACY)
69
70static struct ipv4_devconf ipv4_devconf = {
71 .data = {
72 [IPV4_DEVCONF_ACCEPT_REDIRECTS - 1] = 1,
73 [IPV4_DEVCONF_SEND_REDIRECTS - 1] = 1,
74 [IPV4_DEVCONF_SECURE_REDIRECTS - 1] = 1,
75 [IPV4_DEVCONF_SHARED_MEDIA - 1] = 1,
76 [IPV4_DEVCONF_IGMPV2_UNSOLICITED_REPORT_INTERVAL - 1] = 10000 /*ms*/,
77 [IPV4_DEVCONF_IGMPV3_UNSOLICITED_REPORT_INTERVAL - 1] = 1000 /*ms*/,
78 },
79};
80
81static struct ipv4_devconf ipv4_devconf_dflt = {
82 .data = {
83 [IPV4_DEVCONF_ACCEPT_REDIRECTS - 1] = 1,
84 [IPV4_DEVCONF_SEND_REDIRECTS - 1] = 1,
85 [IPV4_DEVCONF_SECURE_REDIRECTS - 1] = 1,
86 [IPV4_DEVCONF_SHARED_MEDIA - 1] = 1,
87 [IPV4_DEVCONF_ACCEPT_SOURCE_ROUTE - 1] = 1,
88 [IPV4_DEVCONF_IGMPV2_UNSOLICITED_REPORT_INTERVAL - 1] = 10000 /*ms*/,
89 [IPV4_DEVCONF_IGMPV3_UNSOLICITED_REPORT_INTERVAL - 1] = 1000 /*ms*/,
90 },
91};
92
93#define IPV4_DEVCONF_DFLT(net, attr) \
94 IPV4_DEVCONF((*net->ipv4.devconf_dflt), attr)
95
96static const struct nla_policy ifa_ipv4_policy[IFA_MAX+1] = {
97 [IFA_LOCAL] = { .type = NLA_U32 },
98 [IFA_ADDRESS] = { .type = NLA_U32 },
99 [IFA_BROADCAST] = { .type = NLA_U32 },
100 [IFA_LABEL] = { .type = NLA_STRING, .len = IFNAMSIZ - 1 },
101 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
102 [IFA_FLAGS] = { .type = NLA_U32 },
103 [IFA_RT_PRIORITY] = { .type = NLA_U32 },
104 [IFA_TARGET_NETNSID] = { .type = NLA_S32 },
105};
106
107struct inet_fill_args {
108 u32 portid;
109 u32 seq;
110 int event;
111 unsigned int flags;
112 int netnsid;
113 int ifindex;
114};
115
116#define IN4_ADDR_HSIZE_SHIFT 8
117#define IN4_ADDR_HSIZE (1U << IN4_ADDR_HSIZE_SHIFT)
118
119static struct hlist_head inet_addr_lst[IN4_ADDR_HSIZE];
120
121static u32 inet_addr_hash(const struct net *net, __be32 addr)
122{
123 u32 val = (__force u32) addr ^ net_hash_mix(net);
124
125 return hash_32(val, IN4_ADDR_HSIZE_SHIFT);
126}
127
128static void inet_hash_insert(struct net *net, struct in_ifaddr *ifa)
129{
130 u32 hash = inet_addr_hash(net, ifa->ifa_local);
131
132 ASSERT_RTNL();
133 hlist_add_head_rcu(&ifa->hash, &inet_addr_lst[hash]);
134}
135
136static void inet_hash_remove(struct in_ifaddr *ifa)
137{
138 ASSERT_RTNL();
139 hlist_del_init_rcu(&ifa->hash);
140}
141
142/**
143 * __ip_dev_find - find the first device with a given source address.
144 * @net: the net namespace
145 * @addr: the source address
146 * @devref: if true, take a reference on the found device
147 *
148 * If a caller uses devref=false, it should be protected by RCU, or RTNL
149 */
150struct net_device *__ip_dev_find(struct net *net, __be32 addr, bool devref)
151{
152 struct net_device *result = NULL;
153 struct in_ifaddr *ifa;
154
155 rcu_read_lock();
156 ifa = inet_lookup_ifaddr_rcu(net, addr);
157 if (!ifa) {
158 struct flowi4 fl4 = { .daddr = addr };
159 struct fib_result res = { 0 };
160 struct fib_table *local;
161
162 /* Fallback to FIB local table so that communication
163 * over loopback subnets work.
164 */
165 local = fib_get_table(net, RT_TABLE_LOCAL);
166 if (local &&
167 !fib_table_lookup(local, &fl4, &res, FIB_LOOKUP_NOREF) &&
168 res.type == RTN_LOCAL)
169 result = FIB_RES_DEV(res);
170 } else {
171 result = ifa->ifa_dev->dev;
172 }
173 if (result && devref)
174 dev_hold(result);
175 rcu_read_unlock();
176 return result;
177}
178EXPORT_SYMBOL(__ip_dev_find);
179
180/* called under RCU lock */
181struct in_ifaddr *inet_lookup_ifaddr_rcu(struct net *net, __be32 addr)
182{
183 u32 hash = inet_addr_hash(net, addr);
184 struct in_ifaddr *ifa;
185
186 hlist_for_each_entry_rcu(ifa, &inet_addr_lst[hash], hash)
187 if (ifa->ifa_local == addr &&
188 net_eq(dev_net(ifa->ifa_dev->dev), net))
189 return ifa;
190
191 return NULL;
192}
193
194static void rtmsg_ifa(int event, struct in_ifaddr *, struct nlmsghdr *, u32);
195
196static BLOCKING_NOTIFIER_HEAD(inetaddr_chain);
197static BLOCKING_NOTIFIER_HEAD(inetaddr_validator_chain);
198static void inet_del_ifa(struct in_device *in_dev,
199 struct in_ifaddr __rcu **ifap,
200 int destroy);
201#ifdef CONFIG_SYSCTL
202static int devinet_sysctl_register(struct in_device *idev);
203static void devinet_sysctl_unregister(struct in_device *idev);
204#else
205static int devinet_sysctl_register(struct in_device *idev)
206{
207 return 0;
208}
209static void devinet_sysctl_unregister(struct in_device *idev)
210{
211}
212#endif
213
214/* Locks all the inet devices. */
215
216static struct in_ifaddr *inet_alloc_ifa(void)
217{
218 return kzalloc(sizeof(struct in_ifaddr), GFP_KERNEL);
219}
220
221static void inet_rcu_free_ifa(struct rcu_head *head)
222{
223 struct in_ifaddr *ifa = container_of(head, struct in_ifaddr, rcu_head);
224 if (ifa->ifa_dev)
225 in_dev_put(ifa->ifa_dev);
226 kfree(ifa);
227}
228
229static void inet_free_ifa(struct in_ifaddr *ifa)
230{
231 call_rcu(&ifa->rcu_head, inet_rcu_free_ifa);
232}
233
234void in_dev_finish_destroy(struct in_device *idev)
235{
236 struct net_device *dev = idev->dev;
237
238 WARN_ON(idev->ifa_list);
239 WARN_ON(idev->mc_list);
240 kfree(rcu_dereference_protected(idev->mc_hash, 1));
241#ifdef NET_REFCNT_DEBUG
242 pr_debug("%s: %p=%s\n", __func__, idev, dev ? dev->name : "NIL");
243#endif
244 dev_put(dev);
245 if (!idev->dead)
246 pr_err("Freeing alive in_device %p\n", idev);
247 else
248 kfree(idev);
249}
250EXPORT_SYMBOL(in_dev_finish_destroy);
251
252static struct in_device *inetdev_init(struct net_device *dev)
253{
254 struct in_device *in_dev;
255 int err = -ENOMEM;
256
257 ASSERT_RTNL();
258
259 in_dev = kzalloc(sizeof(*in_dev), GFP_KERNEL);
260 if (!in_dev)
261 goto out;
262 memcpy(&in_dev->cnf, dev_net(dev)->ipv4.devconf_dflt,
263 sizeof(in_dev->cnf));
264 in_dev->cnf.sysctl = NULL;
265 in_dev->dev = dev;
266 in_dev->arp_parms = neigh_parms_alloc(dev, &arp_tbl);
267 if (!in_dev->arp_parms)
268 goto out_kfree;
269 if (IPV4_DEVCONF(in_dev->cnf, FORWARDING))
270 dev_disable_lro(dev);
271 /* Reference in_dev->dev */
272 dev_hold(dev);
273 /* Account for reference dev->ip_ptr (below) */
274 refcount_set(&in_dev->refcnt, 1);
275
276 err = devinet_sysctl_register(in_dev);
277 if (err) {
278 in_dev->dead = 1;
279 neigh_parms_release(&arp_tbl, in_dev->arp_parms);
280 in_dev_put(in_dev);
281 in_dev = NULL;
282 goto out;
283 }
284 ip_mc_init_dev(in_dev);
285 if (dev->flags & IFF_UP)
286 ip_mc_up(in_dev);
287
288 /* we can receive as soon as ip_ptr is set -- do this last */
289 rcu_assign_pointer(dev->ip_ptr, in_dev);
290out:
291 return in_dev ?: ERR_PTR(err);
292out_kfree:
293 kfree(in_dev);
294 in_dev = NULL;
295 goto out;
296}
297
298static void in_dev_rcu_put(struct rcu_head *head)
299{
300 struct in_device *idev = container_of(head, struct in_device, rcu_head);
301 in_dev_put(idev);
302}
303
304static void inetdev_destroy(struct in_device *in_dev)
305{
306 struct net_device *dev;
307 struct in_ifaddr *ifa;
308
309 ASSERT_RTNL();
310
311 dev = in_dev->dev;
312
313 in_dev->dead = 1;
314
315 ip_mc_destroy_dev(in_dev);
316
317 while ((ifa = rtnl_dereference(in_dev->ifa_list)) != NULL) {
318 inet_del_ifa(in_dev, &in_dev->ifa_list, 0);
319 inet_free_ifa(ifa);
320 }
321
322 RCU_INIT_POINTER(dev->ip_ptr, NULL);
323
324 devinet_sysctl_unregister(in_dev);
325 neigh_parms_release(&arp_tbl, in_dev->arp_parms);
326 arp_ifdown(dev);
327
328 call_rcu(&in_dev->rcu_head, in_dev_rcu_put);
329}
330
331int inet_addr_onlink(struct in_device *in_dev, __be32 a, __be32 b)
332{
333 const struct in_ifaddr *ifa;
334
335 rcu_read_lock();
336 in_dev_for_each_ifa_rcu(ifa, in_dev) {
337 if (inet_ifa_match(a, ifa)) {
338 if (!b || inet_ifa_match(b, ifa)) {
339 rcu_read_unlock();
340 return 1;
341 }
342 }
343 }
344 rcu_read_unlock();
345 return 0;
346}
347
348static void __inet_del_ifa(struct in_device *in_dev,
349 struct in_ifaddr __rcu **ifap,
350 int destroy, struct nlmsghdr *nlh, u32 portid)
351{
352 struct in_ifaddr *promote = NULL;
353 struct in_ifaddr *ifa, *ifa1;
354 struct in_ifaddr *last_prim;
355 struct in_ifaddr *prev_prom = NULL;
356 int do_promote = IN_DEV_PROMOTE_SECONDARIES(in_dev);
357
358 ASSERT_RTNL();
359
360 ifa1 = rtnl_dereference(*ifap);
361 last_prim = rtnl_dereference(in_dev->ifa_list);
362 if (in_dev->dead)
363 goto no_promotions;
364
365 /* 1. Deleting primary ifaddr forces deletion all secondaries
366 * unless alias promotion is set
367 **/
368
369 if (!(ifa1->ifa_flags & IFA_F_SECONDARY)) {
370 struct in_ifaddr __rcu **ifap1 = &ifa1->ifa_next;
371
372 while ((ifa = rtnl_dereference(*ifap1)) != NULL) {
373 if (!(ifa->ifa_flags & IFA_F_SECONDARY) &&
374 ifa1->ifa_scope <= ifa->ifa_scope)
375 last_prim = ifa;
376
377 if (!(ifa->ifa_flags & IFA_F_SECONDARY) ||
378 ifa1->ifa_mask != ifa->ifa_mask ||
379 !inet_ifa_match(ifa1->ifa_address, ifa)) {
380 ifap1 = &ifa->ifa_next;
381 prev_prom = ifa;
382 continue;
383 }
384
385 if (!do_promote) {
386 inet_hash_remove(ifa);
387 *ifap1 = ifa->ifa_next;
388
389 rtmsg_ifa(RTM_DELADDR, ifa, nlh, portid);
390 blocking_notifier_call_chain(&inetaddr_chain,
391 NETDEV_DOWN, ifa);
392 inet_free_ifa(ifa);
393 } else {
394 promote = ifa;
395 break;
396 }
397 }
398 }
399
400 /* On promotion all secondaries from subnet are changing
401 * the primary IP, we must remove all their routes silently
402 * and later to add them back with new prefsrc. Do this
403 * while all addresses are on the device list.
404 */
405 for (ifa = promote; ifa; ifa = rtnl_dereference(ifa->ifa_next)) {
406 if (ifa1->ifa_mask == ifa->ifa_mask &&
407 inet_ifa_match(ifa1->ifa_address, ifa))
408 fib_del_ifaddr(ifa, ifa1);
409 }
410
411no_promotions:
412 /* 2. Unlink it */
413
414 *ifap = ifa1->ifa_next;
415 inet_hash_remove(ifa1);
416
417 /* 3. Announce address deletion */
418
419 /* Send message first, then call notifier.
420 At first sight, FIB update triggered by notifier
421 will refer to already deleted ifaddr, that could confuse
422 netlink listeners. It is not true: look, gated sees
423 that route deleted and if it still thinks that ifaddr
424 is valid, it will try to restore deleted routes... Grr.
425 So that, this order is correct.
426 */
427 rtmsg_ifa(RTM_DELADDR, ifa1, nlh, portid);
428 blocking_notifier_call_chain(&inetaddr_chain, NETDEV_DOWN, ifa1);
429
430 if (promote) {
431 struct in_ifaddr *next_sec;
432
433 next_sec = rtnl_dereference(promote->ifa_next);
434 if (prev_prom) {
435 struct in_ifaddr *last_sec;
436
437 rcu_assign_pointer(prev_prom->ifa_next, next_sec);
438
439 last_sec = rtnl_dereference(last_prim->ifa_next);
440 rcu_assign_pointer(promote->ifa_next, last_sec);
441 rcu_assign_pointer(last_prim->ifa_next, promote);
442 }
443
444 promote->ifa_flags &= ~IFA_F_SECONDARY;
445 rtmsg_ifa(RTM_NEWADDR, promote, nlh, portid);
446 blocking_notifier_call_chain(&inetaddr_chain,
447 NETDEV_UP, promote);
448 for (ifa = next_sec; ifa;
449 ifa = rtnl_dereference(ifa->ifa_next)) {
450 if (ifa1->ifa_mask != ifa->ifa_mask ||
451 !inet_ifa_match(ifa1->ifa_address, ifa))
452 continue;
453 fib_add_ifaddr(ifa);
454 }
455
456 }
457 if (destroy)
458 inet_free_ifa(ifa1);
459}
460
461static void inet_del_ifa(struct in_device *in_dev,
462 struct in_ifaddr __rcu **ifap,
463 int destroy)
464{
465 __inet_del_ifa(in_dev, ifap, destroy, NULL, 0);
466}
467
468static void check_lifetime(struct work_struct *work);
469
470static DECLARE_DELAYED_WORK(check_lifetime_work, check_lifetime);
471
472static int __inet_insert_ifa(struct in_ifaddr *ifa, struct nlmsghdr *nlh,
473 u32 portid, struct netlink_ext_ack *extack)
474{
475 struct in_ifaddr __rcu **last_primary, **ifap;
476 struct in_device *in_dev = ifa->ifa_dev;
477 struct in_validator_info ivi;
478 struct in_ifaddr *ifa1;
479 int ret;
480
481 ASSERT_RTNL();
482
483 if (!ifa->ifa_local) {
484 inet_free_ifa(ifa);
485 return 0;
486 }
487
488 ifa->ifa_flags &= ~IFA_F_SECONDARY;
489 last_primary = &in_dev->ifa_list;
490
491 /* Don't set IPv6 only flags to IPv4 addresses */
492 ifa->ifa_flags &= ~IPV6ONLY_FLAGS;
493
494 ifap = &in_dev->ifa_list;
495 ifa1 = rtnl_dereference(*ifap);
496
497 while (ifa1) {
498 if (!(ifa1->ifa_flags & IFA_F_SECONDARY) &&
499 ifa->ifa_scope <= ifa1->ifa_scope)
500 last_primary = &ifa1->ifa_next;
501 if (ifa1->ifa_mask == ifa->ifa_mask &&
502 inet_ifa_match(ifa1->ifa_address, ifa)) {
503 if (ifa1->ifa_local == ifa->ifa_local) {
504 inet_free_ifa(ifa);
505 return -EEXIST;
506 }
507 if (ifa1->ifa_scope != ifa->ifa_scope) {
508 inet_free_ifa(ifa);
509 return -EINVAL;
510 }
511 ifa->ifa_flags |= IFA_F_SECONDARY;
512 }
513
514 ifap = &ifa1->ifa_next;
515 ifa1 = rtnl_dereference(*ifap);
516 }
517
518 /* Allow any devices that wish to register ifaddr validtors to weigh
519 * in now, before changes are committed. The rntl lock is serializing
520 * access here, so the state should not change between a validator call
521 * and a final notify on commit. This isn't invoked on promotion under
522 * the assumption that validators are checking the address itself, and
523 * not the flags.
524 */
525 ivi.ivi_addr = ifa->ifa_address;
526 ivi.ivi_dev = ifa->ifa_dev;
527 ivi.extack = extack;
528 ret = blocking_notifier_call_chain(&inetaddr_validator_chain,
529 NETDEV_UP, &ivi);
530 ret = notifier_to_errno(ret);
531 if (ret) {
532 inet_free_ifa(ifa);
533 return ret;
534 }
535
536 if (!(ifa->ifa_flags & IFA_F_SECONDARY)) {
537 prandom_seed((__force u32) ifa->ifa_local);
538 ifap = last_primary;
539 }
540
541 rcu_assign_pointer(ifa->ifa_next, *ifap);
542 rcu_assign_pointer(*ifap, ifa);
543
544 inet_hash_insert(dev_net(in_dev->dev), ifa);
545
546 cancel_delayed_work(&check_lifetime_work);
547 queue_delayed_work(system_power_efficient_wq, &check_lifetime_work, 0);
548
549 /* Send message first, then call notifier.
550 Notifier will trigger FIB update, so that
551 listeners of netlink will know about new ifaddr */
552 rtmsg_ifa(RTM_NEWADDR, ifa, nlh, portid);
553 blocking_notifier_call_chain(&inetaddr_chain, NETDEV_UP, ifa);
554
555 return 0;
556}
557
558static int inet_insert_ifa(struct in_ifaddr *ifa)
559{
560 return __inet_insert_ifa(ifa, NULL, 0, NULL);
561}
562
563static int inet_set_ifa(struct net_device *dev, struct in_ifaddr *ifa)
564{
565 struct in_device *in_dev = __in_dev_get_rtnl(dev);
566
567 ASSERT_RTNL();
568
569 if (!in_dev) {
570 inet_free_ifa(ifa);
571 return -ENOBUFS;
572 }
573 ipv4_devconf_setall(in_dev);
574 neigh_parms_data_state_setall(in_dev->arp_parms);
575 if (ifa->ifa_dev != in_dev) {
576 WARN_ON(ifa->ifa_dev);
577 in_dev_hold(in_dev);
578 ifa->ifa_dev = in_dev;
579 }
580 if (ipv4_is_loopback(ifa->ifa_local))
581 ifa->ifa_scope = RT_SCOPE_HOST;
582 return inet_insert_ifa(ifa);
583}
584
585/* Caller must hold RCU or RTNL :
586 * We dont take a reference on found in_device
587 */
588struct in_device *inetdev_by_index(struct net *net, int ifindex)
589{
590 struct net_device *dev;
591 struct in_device *in_dev = NULL;
592
593 rcu_read_lock();
594 dev = dev_get_by_index_rcu(net, ifindex);
595 if (dev)
596 in_dev = rcu_dereference_rtnl(dev->ip_ptr);
597 rcu_read_unlock();
598 return in_dev;
599}
600EXPORT_SYMBOL(inetdev_by_index);
601
602/* Called only from RTNL semaphored context. No locks. */
603
604struct in_ifaddr *inet_ifa_byprefix(struct in_device *in_dev, __be32 prefix,
605 __be32 mask)
606{
607 struct in_ifaddr *ifa;
608
609 ASSERT_RTNL();
610
611 in_dev_for_each_ifa_rtnl(ifa, in_dev) {
612 if (ifa->ifa_mask == mask && inet_ifa_match(prefix, ifa))
613 return ifa;
614 }
615 return NULL;
616}
617
618static int ip_mc_autojoin_config(struct net *net, bool join,
619 const struct in_ifaddr *ifa)
620{
621#if defined(CONFIG_IP_MULTICAST)
622 struct ip_mreqn mreq = {
623 .imr_multiaddr.s_addr = ifa->ifa_address,
624 .imr_ifindex = ifa->ifa_dev->dev->ifindex,
625 };
626 struct sock *sk = net->ipv4.mc_autojoin_sk;
627 int ret;
628
629 ASSERT_RTNL();
630
631 lock_sock(sk);
632 if (join)
633 ret = ip_mc_join_group(sk, &mreq);
634 else
635 ret = ip_mc_leave_group(sk, &mreq);
636 release_sock(sk);
637
638 return ret;
639#else
640 return -EOPNOTSUPP;
641#endif
642}
643
644static int inet_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
645 struct netlink_ext_ack *extack)
646{
647 struct net *net = sock_net(skb->sk);
648 struct in_ifaddr __rcu **ifap;
649 struct nlattr *tb[IFA_MAX+1];
650 struct in_device *in_dev;
651 struct ifaddrmsg *ifm;
652 struct in_ifaddr *ifa;
653
654 int err = -EINVAL;
655
656 ASSERT_RTNL();
657
658 err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
659 ifa_ipv4_policy, extack);
660 if (err < 0)
661 goto errout;
662
663 ifm = nlmsg_data(nlh);
664 in_dev = inetdev_by_index(net, ifm->ifa_index);
665 if (!in_dev) {
666 err = -ENODEV;
667 goto errout;
668 }
669
670 for (ifap = &in_dev->ifa_list; (ifa = rtnl_dereference(*ifap)) != NULL;
671 ifap = &ifa->ifa_next) {
672 if (tb[IFA_LOCAL] &&
673 ifa->ifa_local != nla_get_in_addr(tb[IFA_LOCAL]))
674 continue;
675
676 if (tb[IFA_LABEL] && nla_strcmp(tb[IFA_LABEL], ifa->ifa_label))
677 continue;
678
679 if (tb[IFA_ADDRESS] &&
680 (ifm->ifa_prefixlen != ifa->ifa_prefixlen ||
681 !inet_ifa_match(nla_get_in_addr(tb[IFA_ADDRESS]), ifa)))
682 continue;
683
684 if (ipv4_is_multicast(ifa->ifa_address))
685 ip_mc_autojoin_config(net, false, ifa);
686 __inet_del_ifa(in_dev, ifap, 1, nlh, NETLINK_CB(skb).portid);
687 return 0;
688 }
689
690 err = -EADDRNOTAVAIL;
691errout:
692 return err;
693}
694
695#define INFINITY_LIFE_TIME 0xFFFFFFFF
696
697static void check_lifetime(struct work_struct *work)
698{
699 unsigned long now, next, next_sec, next_sched;
700 struct in_ifaddr *ifa;
701 struct hlist_node *n;
702 int i;
703
704 now = jiffies;
705 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
706
707 for (i = 0; i < IN4_ADDR_HSIZE; i++) {
708 bool change_needed = false;
709
710 rcu_read_lock();
711 hlist_for_each_entry_rcu(ifa, &inet_addr_lst[i], hash) {
712 unsigned long age;
713
714 if (ifa->ifa_flags & IFA_F_PERMANENT)
715 continue;
716
717 /* We try to batch several events at once. */
718 age = (now - ifa->ifa_tstamp +
719 ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
720
721 if (ifa->ifa_valid_lft != INFINITY_LIFE_TIME &&
722 age >= ifa->ifa_valid_lft) {
723 change_needed = true;
724 } else if (ifa->ifa_preferred_lft ==
725 INFINITY_LIFE_TIME) {
726 continue;
727 } else if (age >= ifa->ifa_preferred_lft) {
728 if (time_before(ifa->ifa_tstamp +
729 ifa->ifa_valid_lft * HZ, next))
730 next = ifa->ifa_tstamp +
731 ifa->ifa_valid_lft * HZ;
732
733 if (!(ifa->ifa_flags & IFA_F_DEPRECATED))
734 change_needed = true;
735 } else if (time_before(ifa->ifa_tstamp +
736 ifa->ifa_preferred_lft * HZ,
737 next)) {
738 next = ifa->ifa_tstamp +
739 ifa->ifa_preferred_lft * HZ;
740 }
741 }
742 rcu_read_unlock();
743 if (!change_needed)
744 continue;
745 rtnl_lock();
746 hlist_for_each_entry_safe(ifa, n, &inet_addr_lst[i], hash) {
747 unsigned long age;
748
749 if (ifa->ifa_flags & IFA_F_PERMANENT)
750 continue;
751
752 /* We try to batch several events at once. */
753 age = (now - ifa->ifa_tstamp +
754 ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
755
756 if (ifa->ifa_valid_lft != INFINITY_LIFE_TIME &&
757 age >= ifa->ifa_valid_lft) {
758 struct in_ifaddr __rcu **ifap;
759 struct in_ifaddr *tmp;
760
761 ifap = &ifa->ifa_dev->ifa_list;
762 tmp = rtnl_dereference(*ifap);
763 while (tmp) {
764 if (tmp == ifa) {
765 inet_del_ifa(ifa->ifa_dev,
766 ifap, 1);
767 break;
768 }
769 ifap = &tmp->ifa_next;
770 tmp = rtnl_dereference(*ifap);
771 }
772 } else if (ifa->ifa_preferred_lft !=
773 INFINITY_LIFE_TIME &&
774 age >= ifa->ifa_preferred_lft &&
775 !(ifa->ifa_flags & IFA_F_DEPRECATED)) {
776 ifa->ifa_flags |= IFA_F_DEPRECATED;
777 rtmsg_ifa(RTM_NEWADDR, ifa, NULL, 0);
778 }
779 }
780 rtnl_unlock();
781 }
782
783 next_sec = round_jiffies_up(next);
784 next_sched = next;
785
786 /* If rounded timeout is accurate enough, accept it. */
787 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
788 next_sched = next_sec;
789
790 now = jiffies;
791 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
792 if (time_before(next_sched, now + ADDRCONF_TIMER_FUZZ_MAX))
793 next_sched = now + ADDRCONF_TIMER_FUZZ_MAX;
794
795 queue_delayed_work(system_power_efficient_wq, &check_lifetime_work,
796 next_sched - now);
797}
798
799static void set_ifa_lifetime(struct in_ifaddr *ifa, __u32 valid_lft,
800 __u32 prefered_lft)
801{
802 unsigned long timeout;
803
804 ifa->ifa_flags &= ~(IFA_F_PERMANENT | IFA_F_DEPRECATED);
805
806 timeout = addrconf_timeout_fixup(valid_lft, HZ);
807 if (addrconf_finite_timeout(timeout))
808 ifa->ifa_valid_lft = timeout;
809 else
810 ifa->ifa_flags |= IFA_F_PERMANENT;
811
812 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
813 if (addrconf_finite_timeout(timeout)) {
814 if (timeout == 0)
815 ifa->ifa_flags |= IFA_F_DEPRECATED;
816 ifa->ifa_preferred_lft = timeout;
817 }
818 ifa->ifa_tstamp = jiffies;
819 if (!ifa->ifa_cstamp)
820 ifa->ifa_cstamp = ifa->ifa_tstamp;
821}
822
823static struct in_ifaddr *rtm_to_ifaddr(struct net *net, struct nlmsghdr *nlh,
824 __u32 *pvalid_lft, __u32 *pprefered_lft,
825 struct netlink_ext_ack *extack)
826{
827 struct nlattr *tb[IFA_MAX+1];
828 struct in_ifaddr *ifa;
829 struct ifaddrmsg *ifm;
830 struct net_device *dev;
831 struct in_device *in_dev;
832 int err;
833
834 err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
835 ifa_ipv4_policy, extack);
836 if (err < 0)
837 goto errout;
838
839 ifm = nlmsg_data(nlh);
840 err = -EINVAL;
841 if (ifm->ifa_prefixlen > 32 || !tb[IFA_LOCAL])
842 goto errout;
843
844 dev = __dev_get_by_index(net, ifm->ifa_index);
845 err = -ENODEV;
846 if (!dev)
847 goto errout;
848
849 in_dev = __in_dev_get_rtnl(dev);
850 err = -ENOBUFS;
851 if (!in_dev)
852 goto errout;
853
854 ifa = inet_alloc_ifa();
855 if (!ifa)
856 /*
857 * A potential indev allocation can be left alive, it stays
858 * assigned to its device and is destroy with it.
859 */
860 goto errout;
861
862 ipv4_devconf_setall(in_dev);
863 neigh_parms_data_state_setall(in_dev->arp_parms);
864 in_dev_hold(in_dev);
865
866 if (!tb[IFA_ADDRESS])
867 tb[IFA_ADDRESS] = tb[IFA_LOCAL];
868
869 INIT_HLIST_NODE(&ifa->hash);
870 ifa->ifa_prefixlen = ifm->ifa_prefixlen;
871 ifa->ifa_mask = inet_make_mask(ifm->ifa_prefixlen);
872 ifa->ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) :
873 ifm->ifa_flags;
874 ifa->ifa_scope = ifm->ifa_scope;
875 ifa->ifa_dev = in_dev;
876
877 ifa->ifa_local = nla_get_in_addr(tb[IFA_LOCAL]);
878 ifa->ifa_address = nla_get_in_addr(tb[IFA_ADDRESS]);
879
880 if (tb[IFA_BROADCAST])
881 ifa->ifa_broadcast = nla_get_in_addr(tb[IFA_BROADCAST]);
882
883 if (tb[IFA_LABEL])
884 nla_strlcpy(ifa->ifa_label, tb[IFA_LABEL], IFNAMSIZ);
885 else
886 memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
887
888 if (tb[IFA_RT_PRIORITY])
889 ifa->ifa_rt_priority = nla_get_u32(tb[IFA_RT_PRIORITY]);
890
891 if (tb[IFA_CACHEINFO]) {
892 struct ifa_cacheinfo *ci;
893
894 ci = nla_data(tb[IFA_CACHEINFO]);
895 if (!ci->ifa_valid || ci->ifa_prefered > ci->ifa_valid) {
896 err = -EINVAL;
897 goto errout_free;
898 }
899 *pvalid_lft = ci->ifa_valid;
900 *pprefered_lft = ci->ifa_prefered;
901 }
902
903 return ifa;
904
905errout_free:
906 inet_free_ifa(ifa);
907errout:
908 return ERR_PTR(err);
909}
910
911static struct in_ifaddr *find_matching_ifa(struct in_ifaddr *ifa)
912{
913 struct in_device *in_dev = ifa->ifa_dev;
914 struct in_ifaddr *ifa1;
915
916 if (!ifa->ifa_local)
917 return NULL;
918
919 in_dev_for_each_ifa_rtnl(ifa1, in_dev) {
920 if (ifa1->ifa_mask == ifa->ifa_mask &&
921 inet_ifa_match(ifa1->ifa_address, ifa) &&
922 ifa1->ifa_local == ifa->ifa_local)
923 return ifa1;
924 }
925 return NULL;
926}
927
928static int inet_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
929 struct netlink_ext_ack *extack)
930{
931 struct net *net = sock_net(skb->sk);
932 struct in_ifaddr *ifa;
933 struct in_ifaddr *ifa_existing;
934 __u32 valid_lft = INFINITY_LIFE_TIME;
935 __u32 prefered_lft = INFINITY_LIFE_TIME;
936
937 ASSERT_RTNL();
938
939 ifa = rtm_to_ifaddr(net, nlh, &valid_lft, &prefered_lft, extack);
940 if (IS_ERR(ifa))
941 return PTR_ERR(ifa);
942
943 ifa_existing = find_matching_ifa(ifa);
944 if (!ifa_existing) {
945 /* It would be best to check for !NLM_F_CREATE here but
946 * userspace already relies on not having to provide this.
947 */
948 set_ifa_lifetime(ifa, valid_lft, prefered_lft);
949 if (ifa->ifa_flags & IFA_F_MCAUTOJOIN) {
950 int ret = ip_mc_autojoin_config(net, true, ifa);
951
952 if (ret < 0) {
953 inet_free_ifa(ifa);
954 return ret;
955 }
956 }
957 return __inet_insert_ifa(ifa, nlh, NETLINK_CB(skb).portid,
958 extack);
959 } else {
960 u32 new_metric = ifa->ifa_rt_priority;
961
962 inet_free_ifa(ifa);
963
964 if (nlh->nlmsg_flags & NLM_F_EXCL ||
965 !(nlh->nlmsg_flags & NLM_F_REPLACE))
966 return -EEXIST;
967 ifa = ifa_existing;
968
969 if (ifa->ifa_rt_priority != new_metric) {
970 fib_modify_prefix_metric(ifa, new_metric);
971 ifa->ifa_rt_priority = new_metric;
972 }
973
974 set_ifa_lifetime(ifa, valid_lft, prefered_lft);
975 cancel_delayed_work(&check_lifetime_work);
976 queue_delayed_work(system_power_efficient_wq,
977 &check_lifetime_work, 0);
978 rtmsg_ifa(RTM_NEWADDR, ifa, nlh, NETLINK_CB(skb).portid);
979 }
980 return 0;
981}
982
983/*
984 * Determine a default network mask, based on the IP address.
985 */
986
987static int inet_abc_len(__be32 addr)
988{
989 int rc = -1; /* Something else, probably a multicast. */
990
991 if (ipv4_is_zeronet(addr) || ipv4_is_lbcast(addr))
992 rc = 0;
993 else {
994 __u32 haddr = ntohl(addr);
995 if (IN_CLASSA(haddr))
996 rc = 8;
997 else if (IN_CLASSB(haddr))
998 rc = 16;
999 else if (IN_CLASSC(haddr))
1000 rc = 24;
1001 else if (IN_CLASSE(haddr))
1002 rc = 32;
1003 }
1004
1005 return rc;
1006}
1007
1008
1009int devinet_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr)
1010{
1011 struct sockaddr_in sin_orig;
1012 struct sockaddr_in *sin = (struct sockaddr_in *)&ifr->ifr_addr;
1013 struct in_ifaddr __rcu **ifap = NULL;
1014 struct in_device *in_dev;
1015 struct in_ifaddr *ifa = NULL;
1016 struct net_device *dev;
1017 char *colon;
1018 int ret = -EFAULT;
1019 int tryaddrmatch = 0;
1020
1021 ifr->ifr_name[IFNAMSIZ - 1] = 0;
1022
1023 /* save original address for comparison */
1024 memcpy(&sin_orig, sin, sizeof(*sin));
1025
1026 colon = strchr(ifr->ifr_name, ':');
1027 if (colon)
1028 *colon = 0;
1029
1030 dev_load(net, ifr->ifr_name);
1031
1032 switch (cmd) {
1033 case SIOCGIFADDR: /* Get interface address */
1034 case SIOCGIFBRDADDR: /* Get the broadcast address */
1035 case SIOCGIFDSTADDR: /* Get the destination address */
1036 case SIOCGIFNETMASK: /* Get the netmask for the interface */
1037 /* Note that these ioctls will not sleep,
1038 so that we do not impose a lock.
1039 One day we will be forced to put shlock here (I mean SMP)
1040 */
1041 tryaddrmatch = (sin_orig.sin_family == AF_INET);
1042 memset(sin, 0, sizeof(*sin));
1043 sin->sin_family = AF_INET;
1044 break;
1045
1046 case SIOCSIFFLAGS:
1047 ret = -EPERM;
1048 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1049 goto out;
1050 break;
1051 case SIOCSIFADDR: /* Set interface address (and family) */
1052 case SIOCSIFBRDADDR: /* Set the broadcast address */
1053 case SIOCSIFDSTADDR: /* Set the destination address */
1054 case SIOCSIFNETMASK: /* Set the netmask for the interface */
1055 ret = -EPERM;
1056 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1057 goto out;
1058 ret = -EINVAL;
1059 if (sin->sin_family != AF_INET)
1060 goto out;
1061 break;
1062 default:
1063 ret = -EINVAL;
1064 goto out;
1065 }
1066
1067 rtnl_lock();
1068
1069 ret = -ENODEV;
1070 dev = __dev_get_by_name(net, ifr->ifr_name);
1071 if (!dev)
1072 goto done;
1073
1074 if (colon)
1075 *colon = ':';
1076
1077 in_dev = __in_dev_get_rtnl(dev);
1078 if (in_dev) {
1079 if (tryaddrmatch) {
1080 /* Matthias Andree */
1081 /* compare label and address (4.4BSD style) */
1082 /* note: we only do this for a limited set of ioctls
1083 and only if the original address family was AF_INET.
1084 This is checked above. */
1085
1086 for (ifap = &in_dev->ifa_list;
1087 (ifa = rtnl_dereference(*ifap)) != NULL;
1088 ifap = &ifa->ifa_next) {
1089 if (!strcmp(ifr->ifr_name, ifa->ifa_label) &&
1090 sin_orig.sin_addr.s_addr ==
1091 ifa->ifa_local) {
1092 break; /* found */
1093 }
1094 }
1095 }
1096 /* we didn't get a match, maybe the application is
1097 4.3BSD-style and passed in junk so we fall back to
1098 comparing just the label */
1099 if (!ifa) {
1100 for (ifap = &in_dev->ifa_list;
1101 (ifa = rtnl_dereference(*ifap)) != NULL;
1102 ifap = &ifa->ifa_next)
1103 if (!strcmp(ifr->ifr_name, ifa->ifa_label))
1104 break;
1105 }
1106 }
1107
1108 ret = -EADDRNOTAVAIL;
1109 if (!ifa && cmd != SIOCSIFADDR && cmd != SIOCSIFFLAGS)
1110 goto done;
1111
1112 switch (cmd) {
1113 case SIOCGIFADDR: /* Get interface address */
1114 ret = 0;
1115 sin->sin_addr.s_addr = ifa->ifa_local;
1116 break;
1117
1118 case SIOCGIFBRDADDR: /* Get the broadcast address */
1119 ret = 0;
1120 sin->sin_addr.s_addr = ifa->ifa_broadcast;
1121 break;
1122
1123 case SIOCGIFDSTADDR: /* Get the destination address */
1124 ret = 0;
1125 sin->sin_addr.s_addr = ifa->ifa_address;
1126 break;
1127
1128 case SIOCGIFNETMASK: /* Get the netmask for the interface */
1129 ret = 0;
1130 sin->sin_addr.s_addr = ifa->ifa_mask;
1131 break;
1132
1133 case SIOCSIFFLAGS:
1134 if (colon) {
1135 ret = -EADDRNOTAVAIL;
1136 if (!ifa)
1137 break;
1138 ret = 0;
1139 if (!(ifr->ifr_flags & IFF_UP))
1140 inet_del_ifa(in_dev, ifap, 1);
1141 break;
1142 }
1143 ret = dev_change_flags(dev, ifr->ifr_flags, NULL);
1144 break;
1145
1146 case SIOCSIFADDR: /* Set interface address (and family) */
1147 ret = -EINVAL;
1148 if (inet_abc_len(sin->sin_addr.s_addr) < 0)
1149 break;
1150
1151 if (!ifa) {
1152 ret = -ENOBUFS;
1153 ifa = inet_alloc_ifa();
1154 if (!ifa)
1155 break;
1156 INIT_HLIST_NODE(&ifa->hash);
1157 if (colon)
1158 memcpy(ifa->ifa_label, ifr->ifr_name, IFNAMSIZ);
1159 else
1160 memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
1161 } else {
1162 ret = 0;
1163 if (ifa->ifa_local == sin->sin_addr.s_addr)
1164 break;
1165 inet_del_ifa(in_dev, ifap, 0);
1166 ifa->ifa_broadcast = 0;
1167 ifa->ifa_scope = 0;
1168 }
1169
1170 ifa->ifa_address = ifa->ifa_local = sin->sin_addr.s_addr;
1171
1172 if (!(dev->flags & IFF_POINTOPOINT)) {
1173 ifa->ifa_prefixlen = inet_abc_len(ifa->ifa_address);
1174 ifa->ifa_mask = inet_make_mask(ifa->ifa_prefixlen);
1175 if ((dev->flags & IFF_BROADCAST) &&
1176 ifa->ifa_prefixlen < 31)
1177 ifa->ifa_broadcast = ifa->ifa_address |
1178 ~ifa->ifa_mask;
1179 } else {
1180 ifa->ifa_prefixlen = 32;
1181 ifa->ifa_mask = inet_make_mask(32);
1182 }
1183 set_ifa_lifetime(ifa, INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
1184 ret = inet_set_ifa(dev, ifa);
1185 break;
1186
1187 case SIOCSIFBRDADDR: /* Set the broadcast address */
1188 ret = 0;
1189 if (ifa->ifa_broadcast != sin->sin_addr.s_addr) {
1190 inet_del_ifa(in_dev, ifap, 0);
1191 ifa->ifa_broadcast = sin->sin_addr.s_addr;
1192 inet_insert_ifa(ifa);
1193 }
1194 break;
1195
1196 case SIOCSIFDSTADDR: /* Set the destination address */
1197 ret = 0;
1198 if (ifa->ifa_address == sin->sin_addr.s_addr)
1199 break;
1200 ret = -EINVAL;
1201 if (inet_abc_len(sin->sin_addr.s_addr) < 0)
1202 break;
1203 ret = 0;
1204 inet_del_ifa(in_dev, ifap, 0);
1205 ifa->ifa_address = sin->sin_addr.s_addr;
1206 inet_insert_ifa(ifa);
1207 break;
1208
1209 case SIOCSIFNETMASK: /* Set the netmask for the interface */
1210
1211 /*
1212 * The mask we set must be legal.
1213 */
1214 ret = -EINVAL;
1215 if (bad_mask(sin->sin_addr.s_addr, 0))
1216 break;
1217 ret = 0;
1218 if (ifa->ifa_mask != sin->sin_addr.s_addr) {
1219 __be32 old_mask = ifa->ifa_mask;
1220 inet_del_ifa(in_dev, ifap, 0);
1221 ifa->ifa_mask = sin->sin_addr.s_addr;
1222 ifa->ifa_prefixlen = inet_mask_len(ifa->ifa_mask);
1223
1224 /* See if current broadcast address matches
1225 * with current netmask, then recalculate
1226 * the broadcast address. Otherwise it's a
1227 * funny address, so don't touch it since
1228 * the user seems to know what (s)he's doing...
1229 */
1230 if ((dev->flags & IFF_BROADCAST) &&
1231 (ifa->ifa_prefixlen < 31) &&
1232 (ifa->ifa_broadcast ==
1233 (ifa->ifa_local|~old_mask))) {
1234 ifa->ifa_broadcast = (ifa->ifa_local |
1235 ~sin->sin_addr.s_addr);
1236 }
1237 inet_insert_ifa(ifa);
1238 }
1239 break;
1240 }
1241done:
1242 rtnl_unlock();
1243out:
1244 return ret;
1245}
1246
1247static int inet_gifconf(struct net_device *dev, char __user *buf, int len, int size)
1248{
1249 struct in_device *in_dev = __in_dev_get_rtnl(dev);
1250 const struct in_ifaddr *ifa;
1251 struct ifreq ifr;
1252 int done = 0;
1253
1254 if (WARN_ON(size > sizeof(struct ifreq)))
1255 goto out;
1256
1257 if (!in_dev)
1258 goto out;
1259
1260 in_dev_for_each_ifa_rtnl(ifa, in_dev) {
1261 if (!buf) {
1262 done += size;
1263 continue;
1264 }
1265 if (len < size)
1266 break;
1267 memset(&ifr, 0, sizeof(struct ifreq));
1268 strcpy(ifr.ifr_name, ifa->ifa_label);
1269
1270 (*(struct sockaddr_in *)&ifr.ifr_addr).sin_family = AF_INET;
1271 (*(struct sockaddr_in *)&ifr.ifr_addr).sin_addr.s_addr =
1272 ifa->ifa_local;
1273
1274 if (copy_to_user(buf + done, &ifr, size)) {
1275 done = -EFAULT;
1276 break;
1277 }
1278 len -= size;
1279 done += size;
1280 }
1281out:
1282 return done;
1283}
1284
1285static __be32 in_dev_select_addr(const struct in_device *in_dev,
1286 int scope)
1287{
1288 const struct in_ifaddr *ifa;
1289
1290 in_dev_for_each_ifa_rcu(ifa, in_dev) {
1291 if (ifa->ifa_flags & IFA_F_SECONDARY)
1292 continue;
1293 if (ifa->ifa_scope != RT_SCOPE_LINK &&
1294 ifa->ifa_scope <= scope)
1295 return ifa->ifa_local;
1296 }
1297
1298 return 0;
1299}
1300
1301__be32 inet_select_addr(const struct net_device *dev, __be32 dst, int scope)
1302{
1303 const struct in_ifaddr *ifa;
1304 __be32 addr = 0;
1305 unsigned char localnet_scope = RT_SCOPE_HOST;
1306 struct in_device *in_dev;
1307 struct net *net = dev_net(dev);
1308 int master_idx;
1309
1310 rcu_read_lock();
1311 in_dev = __in_dev_get_rcu(dev);
1312 if (!in_dev)
1313 goto no_in_dev;
1314
1315 if (unlikely(IN_DEV_ROUTE_LOCALNET(in_dev)))
1316 localnet_scope = RT_SCOPE_LINK;
1317
1318 in_dev_for_each_ifa_rcu(ifa, in_dev) {
1319 if (ifa->ifa_flags & IFA_F_SECONDARY)
1320 continue;
1321 if (min(ifa->ifa_scope, localnet_scope) > scope)
1322 continue;
1323 if (!dst || inet_ifa_match(dst, ifa)) {
1324 addr = ifa->ifa_local;
1325 break;
1326 }
1327 if (!addr)
1328 addr = ifa->ifa_local;
1329 }
1330
1331 if (addr)
1332 goto out_unlock;
1333no_in_dev:
1334 master_idx = l3mdev_master_ifindex_rcu(dev);
1335
1336 /* For VRFs, the VRF device takes the place of the loopback device,
1337 * with addresses on it being preferred. Note in such cases the
1338 * loopback device will be among the devices that fail the master_idx
1339 * equality check in the loop below.
1340 */
1341 if (master_idx &&
1342 (dev = dev_get_by_index_rcu(net, master_idx)) &&
1343 (in_dev = __in_dev_get_rcu(dev))) {
1344 addr = in_dev_select_addr(in_dev, scope);
1345 if (addr)
1346 goto out_unlock;
1347 }
1348
1349 /* Not loopback addresses on loopback should be preferred
1350 in this case. It is important that lo is the first interface
1351 in dev_base list.
1352 */
1353 for_each_netdev_rcu(net, dev) {
1354 if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1355 continue;
1356
1357 in_dev = __in_dev_get_rcu(dev);
1358 if (!in_dev)
1359 continue;
1360
1361 addr = in_dev_select_addr(in_dev, scope);
1362 if (addr)
1363 goto out_unlock;
1364 }
1365out_unlock:
1366 rcu_read_unlock();
1367 return addr;
1368}
1369EXPORT_SYMBOL(inet_select_addr);
1370
1371static __be32 confirm_addr_indev(struct in_device *in_dev, __be32 dst,
1372 __be32 local, int scope)
1373{
1374 unsigned char localnet_scope = RT_SCOPE_HOST;
1375 const struct in_ifaddr *ifa;
1376 __be32 addr = 0;
1377 int same = 0;
1378
1379 if (unlikely(IN_DEV_ROUTE_LOCALNET(in_dev)))
1380 localnet_scope = RT_SCOPE_LINK;
1381
1382 in_dev_for_each_ifa_rcu(ifa, in_dev) {
1383 unsigned char min_scope = min(ifa->ifa_scope, localnet_scope);
1384
1385 if (!addr &&
1386 (local == ifa->ifa_local || !local) &&
1387 min_scope <= scope) {
1388 addr = ifa->ifa_local;
1389 if (same)
1390 break;
1391 }
1392 if (!same) {
1393 same = (!local || inet_ifa_match(local, ifa)) &&
1394 (!dst || inet_ifa_match(dst, ifa));
1395 if (same && addr) {
1396 if (local || !dst)
1397 break;
1398 /* Is the selected addr into dst subnet? */
1399 if (inet_ifa_match(addr, ifa))
1400 break;
1401 /* No, then can we use new local src? */
1402 if (min_scope <= scope) {
1403 addr = ifa->ifa_local;
1404 break;
1405 }
1406 /* search for large dst subnet for addr */
1407 same = 0;
1408 }
1409 }
1410 }
1411
1412 return same ? addr : 0;
1413}
1414
1415/*
1416 * Confirm that local IP address exists using wildcards:
1417 * - net: netns to check, cannot be NULL
1418 * - in_dev: only on this interface, NULL=any interface
1419 * - dst: only in the same subnet as dst, 0=any dst
1420 * - local: address, 0=autoselect the local address
1421 * - scope: maximum allowed scope value for the local address
1422 */
1423__be32 inet_confirm_addr(struct net *net, struct in_device *in_dev,
1424 __be32 dst, __be32 local, int scope)
1425{
1426 __be32 addr = 0;
1427 struct net_device *dev;
1428
1429 if (in_dev)
1430 return confirm_addr_indev(in_dev, dst, local, scope);
1431
1432 rcu_read_lock();
1433 for_each_netdev_rcu(net, dev) {
1434 in_dev = __in_dev_get_rcu(dev);
1435 if (in_dev) {
1436 addr = confirm_addr_indev(in_dev, dst, local, scope);
1437 if (addr)
1438 break;
1439 }
1440 }
1441 rcu_read_unlock();
1442
1443 return addr;
1444}
1445EXPORT_SYMBOL(inet_confirm_addr);
1446
1447/*
1448 * Device notifier
1449 */
1450
1451int register_inetaddr_notifier(struct notifier_block *nb)
1452{
1453 return blocking_notifier_chain_register(&inetaddr_chain, nb);
1454}
1455EXPORT_SYMBOL(register_inetaddr_notifier);
1456
1457int unregister_inetaddr_notifier(struct notifier_block *nb)
1458{
1459 return blocking_notifier_chain_unregister(&inetaddr_chain, nb);
1460}
1461EXPORT_SYMBOL(unregister_inetaddr_notifier);
1462
1463int register_inetaddr_validator_notifier(struct notifier_block *nb)
1464{
1465 return blocking_notifier_chain_register(&inetaddr_validator_chain, nb);
1466}
1467EXPORT_SYMBOL(register_inetaddr_validator_notifier);
1468
1469int unregister_inetaddr_validator_notifier(struct notifier_block *nb)
1470{
1471 return blocking_notifier_chain_unregister(&inetaddr_validator_chain,
1472 nb);
1473}
1474EXPORT_SYMBOL(unregister_inetaddr_validator_notifier);
1475
1476/* Rename ifa_labels for a device name change. Make some effort to preserve
1477 * existing alias numbering and to create unique labels if possible.
1478*/
1479static void inetdev_changename(struct net_device *dev, struct in_device *in_dev)
1480{
1481 struct in_ifaddr *ifa;
1482 int named = 0;
1483
1484 in_dev_for_each_ifa_rtnl(ifa, in_dev) {
1485 char old[IFNAMSIZ], *dot;
1486
1487 memcpy(old, ifa->ifa_label, IFNAMSIZ);
1488 memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
1489 if (named++ == 0)
1490 goto skip;
1491 dot = strchr(old, ':');
1492 if (!dot) {
1493 sprintf(old, ":%d", named);
1494 dot = old;
1495 }
1496 if (strlen(dot) + strlen(dev->name) < IFNAMSIZ)
1497 strcat(ifa->ifa_label, dot);
1498 else
1499 strcpy(ifa->ifa_label + (IFNAMSIZ - strlen(dot) - 1), dot);
1500skip:
1501 rtmsg_ifa(RTM_NEWADDR, ifa, NULL, 0);
1502 }
1503}
1504
1505static void inetdev_send_gratuitous_arp(struct net_device *dev,
1506 struct in_device *in_dev)
1507
1508{
1509 const struct in_ifaddr *ifa;
1510
1511 in_dev_for_each_ifa_rtnl(ifa, in_dev) {
1512 arp_send(ARPOP_REQUEST, ETH_P_ARP,
1513 ifa->ifa_local, dev,
1514 ifa->ifa_local, NULL,
1515 dev->dev_addr, NULL);
1516 }
1517}
1518
1519/* Called only under RTNL semaphore */
1520
1521static int inetdev_event(struct notifier_block *this, unsigned long event,
1522 void *ptr)
1523{
1524 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1525 struct in_device *in_dev = __in_dev_get_rtnl(dev);
1526
1527 ASSERT_RTNL();
1528
1529 if (!in_dev) {
1530 if (event == NETDEV_REGISTER) {
1531 in_dev = inetdev_init(dev);
1532 if (IS_ERR(in_dev))
1533 return notifier_from_errno(PTR_ERR(in_dev));
1534 if (dev->flags & IFF_LOOPBACK) {
1535 IN_DEV_CONF_SET(in_dev, NOXFRM, 1);
1536 IN_DEV_CONF_SET(in_dev, NOPOLICY, 1);
1537 }
1538 } else if (event == NETDEV_CHANGEMTU) {
1539 /* Re-enabling IP */
1540 if (inetdev_valid_mtu(dev->mtu))
1541 in_dev = inetdev_init(dev);
1542 }
1543 goto out;
1544 }
1545
1546 switch (event) {
1547 case NETDEV_REGISTER:
1548 pr_debug("%s: bug\n", __func__);
1549 RCU_INIT_POINTER(dev->ip_ptr, NULL);
1550 break;
1551 case NETDEV_UP:
1552 if (!inetdev_valid_mtu(dev->mtu))
1553 break;
1554 if (dev->flags & IFF_LOOPBACK) {
1555 struct in_ifaddr *ifa = inet_alloc_ifa();
1556
1557 if (ifa) {
1558 INIT_HLIST_NODE(&ifa->hash);
1559 ifa->ifa_local =
1560 ifa->ifa_address = htonl(INADDR_LOOPBACK);
1561 ifa->ifa_prefixlen = 8;
1562 ifa->ifa_mask = inet_make_mask(8);
1563 in_dev_hold(in_dev);
1564 ifa->ifa_dev = in_dev;
1565 ifa->ifa_scope = RT_SCOPE_HOST;
1566 memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
1567 set_ifa_lifetime(ifa, INFINITY_LIFE_TIME,
1568 INFINITY_LIFE_TIME);
1569 ipv4_devconf_setall(in_dev);
1570 neigh_parms_data_state_setall(in_dev->arp_parms);
1571 inet_insert_ifa(ifa);
1572 }
1573 }
1574 ip_mc_up(in_dev);
1575 fallthrough;
1576 case NETDEV_CHANGEADDR:
1577 if (!IN_DEV_ARP_NOTIFY(in_dev))
1578 break;
1579 fallthrough;
1580 case NETDEV_NOTIFY_PEERS:
1581 /* Send gratuitous ARP to notify of link change */
1582 inetdev_send_gratuitous_arp(dev, in_dev);
1583 break;
1584 case NETDEV_DOWN:
1585 ip_mc_down(in_dev);
1586 break;
1587 case NETDEV_PRE_TYPE_CHANGE:
1588 ip_mc_unmap(in_dev);
1589 break;
1590 case NETDEV_POST_TYPE_CHANGE:
1591 ip_mc_remap(in_dev);
1592 break;
1593 case NETDEV_CHANGEMTU:
1594 if (inetdev_valid_mtu(dev->mtu))
1595 break;
1596 /* disable IP when MTU is not enough */
1597 fallthrough;
1598 case NETDEV_UNREGISTER:
1599 inetdev_destroy(in_dev);
1600 break;
1601 case NETDEV_CHANGENAME:
1602 /* Do not notify about label change, this event is
1603 * not interesting to applications using netlink.
1604 */
1605 inetdev_changename(dev, in_dev);
1606
1607 devinet_sysctl_unregister(in_dev);
1608 devinet_sysctl_register(in_dev);
1609 break;
1610 }
1611out:
1612 return NOTIFY_DONE;
1613}
1614
1615static struct notifier_block ip_netdev_notifier = {
1616 .notifier_call = inetdev_event,
1617};
1618
1619static size_t inet_nlmsg_size(void)
1620{
1621 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
1622 + nla_total_size(4) /* IFA_ADDRESS */
1623 + nla_total_size(4) /* IFA_LOCAL */
1624 + nla_total_size(4) /* IFA_BROADCAST */
1625 + nla_total_size(IFNAMSIZ) /* IFA_LABEL */
1626 + nla_total_size(4) /* IFA_FLAGS */
1627 + nla_total_size(4) /* IFA_RT_PRIORITY */
1628 + nla_total_size(sizeof(struct ifa_cacheinfo)); /* IFA_CACHEINFO */
1629}
1630
1631static inline u32 cstamp_delta(unsigned long cstamp)
1632{
1633 return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
1634}
1635
1636static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
1637 unsigned long tstamp, u32 preferred, u32 valid)
1638{
1639 struct ifa_cacheinfo ci;
1640
1641 ci.cstamp = cstamp_delta(cstamp);
1642 ci.tstamp = cstamp_delta(tstamp);
1643 ci.ifa_prefered = preferred;
1644 ci.ifa_valid = valid;
1645
1646 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
1647}
1648
1649static int inet_fill_ifaddr(struct sk_buff *skb, struct in_ifaddr *ifa,
1650 struct inet_fill_args *args)
1651{
1652 struct ifaddrmsg *ifm;
1653 struct nlmsghdr *nlh;
1654 u32 preferred, valid;
1655
1656 nlh = nlmsg_put(skb, args->portid, args->seq, args->event, sizeof(*ifm),
1657 args->flags);
1658 if (!nlh)
1659 return -EMSGSIZE;
1660
1661 ifm = nlmsg_data(nlh);
1662 ifm->ifa_family = AF_INET;
1663 ifm->ifa_prefixlen = ifa->ifa_prefixlen;
1664 ifm->ifa_flags = ifa->ifa_flags;
1665 ifm->ifa_scope = ifa->ifa_scope;
1666 ifm->ifa_index = ifa->ifa_dev->dev->ifindex;
1667
1668 if (args->netnsid >= 0 &&
1669 nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
1670 goto nla_put_failure;
1671
1672 if (!(ifm->ifa_flags & IFA_F_PERMANENT)) {
1673 preferred = ifa->ifa_preferred_lft;
1674 valid = ifa->ifa_valid_lft;
1675 if (preferred != INFINITY_LIFE_TIME) {
1676 long tval = (jiffies - ifa->ifa_tstamp) / HZ;
1677
1678 if (preferred > tval)
1679 preferred -= tval;
1680 else
1681 preferred = 0;
1682 if (valid != INFINITY_LIFE_TIME) {
1683 if (valid > tval)
1684 valid -= tval;
1685 else
1686 valid = 0;
1687 }
1688 }
1689 } else {
1690 preferred = INFINITY_LIFE_TIME;
1691 valid = INFINITY_LIFE_TIME;
1692 }
1693 if ((ifa->ifa_address &&
1694 nla_put_in_addr(skb, IFA_ADDRESS, ifa->ifa_address)) ||
1695 (ifa->ifa_local &&
1696 nla_put_in_addr(skb, IFA_LOCAL, ifa->ifa_local)) ||
1697 (ifa->ifa_broadcast &&
1698 nla_put_in_addr(skb, IFA_BROADCAST, ifa->ifa_broadcast)) ||
1699 (ifa->ifa_label[0] &&
1700 nla_put_string(skb, IFA_LABEL, ifa->ifa_label)) ||
1701 nla_put_u32(skb, IFA_FLAGS, ifa->ifa_flags) ||
1702 (ifa->ifa_rt_priority &&
1703 nla_put_u32(skb, IFA_RT_PRIORITY, ifa->ifa_rt_priority)) ||
1704 put_cacheinfo(skb, ifa->ifa_cstamp, ifa->ifa_tstamp,
1705 preferred, valid))
1706 goto nla_put_failure;
1707
1708 nlmsg_end(skb, nlh);
1709 return 0;
1710
1711nla_put_failure:
1712 nlmsg_cancel(skb, nlh);
1713 return -EMSGSIZE;
1714}
1715
1716static int inet_valid_dump_ifaddr_req(const struct nlmsghdr *nlh,
1717 struct inet_fill_args *fillargs,
1718 struct net **tgt_net, struct sock *sk,
1719 struct netlink_callback *cb)
1720{
1721 struct netlink_ext_ack *extack = cb->extack;
1722 struct nlattr *tb[IFA_MAX+1];
1723 struct ifaddrmsg *ifm;
1724 int err, i;
1725
1726 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
1727 NL_SET_ERR_MSG(extack, "ipv4: Invalid header for address dump request");
1728 return -EINVAL;
1729 }
1730
1731 ifm = nlmsg_data(nlh);
1732 if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
1733 NL_SET_ERR_MSG(extack, "ipv4: Invalid values in header for address dump request");
1734 return -EINVAL;
1735 }
1736
1737 fillargs->ifindex = ifm->ifa_index;
1738 if (fillargs->ifindex) {
1739 cb->answer_flags |= NLM_F_DUMP_FILTERED;
1740 fillargs->flags |= NLM_F_DUMP_FILTERED;
1741 }
1742
1743 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
1744 ifa_ipv4_policy, extack);
1745 if (err < 0)
1746 return err;
1747
1748 for (i = 0; i <= IFA_MAX; ++i) {
1749 if (!tb[i])
1750 continue;
1751
1752 if (i == IFA_TARGET_NETNSID) {
1753 struct net *net;
1754
1755 fillargs->netnsid = nla_get_s32(tb[i]);
1756
1757 net = rtnl_get_net_ns_capable(sk, fillargs->netnsid);
1758 if (IS_ERR(net)) {
1759 fillargs->netnsid = -1;
1760 NL_SET_ERR_MSG(extack, "ipv4: Invalid target network namespace id");
1761 return PTR_ERR(net);
1762 }
1763 *tgt_net = net;
1764 } else {
1765 NL_SET_ERR_MSG(extack, "ipv4: Unsupported attribute in dump request");
1766 return -EINVAL;
1767 }
1768 }
1769
1770 return 0;
1771}
1772
1773static int in_dev_dump_addr(struct in_device *in_dev, struct sk_buff *skb,
1774 struct netlink_callback *cb, int s_ip_idx,
1775 struct inet_fill_args *fillargs)
1776{
1777 struct in_ifaddr *ifa;
1778 int ip_idx = 0;
1779 int err;
1780
1781 in_dev_for_each_ifa_rtnl(ifa, in_dev) {
1782 if (ip_idx < s_ip_idx) {
1783 ip_idx++;
1784 continue;
1785 }
1786 err = inet_fill_ifaddr(skb, ifa, fillargs);
1787 if (err < 0)
1788 goto done;
1789
1790 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1791 ip_idx++;
1792 }
1793 err = 0;
1794
1795done:
1796 cb->args[2] = ip_idx;
1797
1798 return err;
1799}
1800
1801static int inet_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
1802{
1803 const struct nlmsghdr *nlh = cb->nlh;
1804 struct inet_fill_args fillargs = {
1805 .portid = NETLINK_CB(cb->skb).portid,
1806 .seq = nlh->nlmsg_seq,
1807 .event = RTM_NEWADDR,
1808 .flags = NLM_F_MULTI,
1809 .netnsid = -1,
1810 };
1811 struct net *net = sock_net(skb->sk);
1812 struct net *tgt_net = net;
1813 int h, s_h;
1814 int idx, s_idx;
1815 int s_ip_idx;
1816 struct net_device *dev;
1817 struct in_device *in_dev;
1818 struct hlist_head *head;
1819 int err = 0;
1820
1821 s_h = cb->args[0];
1822 s_idx = idx = cb->args[1];
1823 s_ip_idx = cb->args[2];
1824
1825 if (cb->strict_check) {
1826 err = inet_valid_dump_ifaddr_req(nlh, &fillargs, &tgt_net,
1827 skb->sk, cb);
1828 if (err < 0)
1829 goto put_tgt_net;
1830
1831 err = 0;
1832 if (fillargs.ifindex) {
1833 dev = __dev_get_by_index(tgt_net, fillargs.ifindex);
1834 if (!dev) {
1835 err = -ENODEV;
1836 goto put_tgt_net;
1837 }
1838
1839 in_dev = __in_dev_get_rtnl(dev);
1840 if (in_dev) {
1841 err = in_dev_dump_addr(in_dev, skb, cb, s_ip_idx,
1842 &fillargs);
1843 }
1844 goto put_tgt_net;
1845 }
1846 }
1847
1848 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1849 idx = 0;
1850 head = &tgt_net->dev_index_head[h];
1851 rcu_read_lock();
1852 cb->seq = atomic_read(&tgt_net->ipv4.dev_addr_genid) ^
1853 tgt_net->dev_base_seq;
1854 hlist_for_each_entry_rcu(dev, head, index_hlist) {
1855 if (idx < s_idx)
1856 goto cont;
1857 if (h > s_h || idx > s_idx)
1858 s_ip_idx = 0;
1859 in_dev = __in_dev_get_rcu(dev);
1860 if (!in_dev)
1861 goto cont;
1862
1863 err = in_dev_dump_addr(in_dev, skb, cb, s_ip_idx,
1864 &fillargs);
1865 if (err < 0) {
1866 rcu_read_unlock();
1867 goto done;
1868 }
1869cont:
1870 idx++;
1871 }
1872 rcu_read_unlock();
1873 }
1874
1875done:
1876 cb->args[0] = h;
1877 cb->args[1] = idx;
1878put_tgt_net:
1879 if (fillargs.netnsid >= 0)
1880 put_net(tgt_net);
1881
1882 return skb->len ? : err;
1883}
1884
1885static void rtmsg_ifa(int event, struct in_ifaddr *ifa, struct nlmsghdr *nlh,
1886 u32 portid)
1887{
1888 struct inet_fill_args fillargs = {
1889 .portid = portid,
1890 .seq = nlh ? nlh->nlmsg_seq : 0,
1891 .event = event,
1892 .flags = 0,
1893 .netnsid = -1,
1894 };
1895 struct sk_buff *skb;
1896 int err = -ENOBUFS;
1897 struct net *net;
1898
1899 net = dev_net(ifa->ifa_dev->dev);
1900 skb = nlmsg_new(inet_nlmsg_size(), GFP_KERNEL);
1901 if (!skb)
1902 goto errout;
1903
1904 err = inet_fill_ifaddr(skb, ifa, &fillargs);
1905 if (err < 0) {
1906 /* -EMSGSIZE implies BUG in inet_nlmsg_size() */
1907 WARN_ON(err == -EMSGSIZE);
1908 kfree_skb(skb);
1909 goto errout;
1910 }
1911 rtnl_notify(skb, net, portid, RTNLGRP_IPV4_IFADDR, nlh, GFP_KERNEL);
1912 return;
1913errout:
1914 if (err < 0)
1915 rtnl_set_sk_err(net, RTNLGRP_IPV4_IFADDR, err);
1916}
1917
1918static size_t inet_get_link_af_size(const struct net_device *dev,
1919 u32 ext_filter_mask)
1920{
1921 struct in_device *in_dev = rcu_dereference_rtnl(dev->ip_ptr);
1922
1923 if (!in_dev)
1924 return 0;
1925
1926 return nla_total_size(IPV4_DEVCONF_MAX * 4); /* IFLA_INET_CONF */
1927}
1928
1929static int inet_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
1930 u32 ext_filter_mask)
1931{
1932 struct in_device *in_dev = rcu_dereference_rtnl(dev->ip_ptr);
1933 struct nlattr *nla;
1934 int i;
1935
1936 if (!in_dev)
1937 return -ENODATA;
1938
1939 nla = nla_reserve(skb, IFLA_INET_CONF, IPV4_DEVCONF_MAX * 4);
1940 if (!nla)
1941 return -EMSGSIZE;
1942
1943 for (i = 0; i < IPV4_DEVCONF_MAX; i++)
1944 ((u32 *) nla_data(nla))[i] = in_dev->cnf.data[i];
1945
1946 return 0;
1947}
1948
1949static const struct nla_policy inet_af_policy[IFLA_INET_MAX+1] = {
1950 [IFLA_INET_CONF] = { .type = NLA_NESTED },
1951};
1952
1953static int inet_validate_link_af(const struct net_device *dev,
1954 const struct nlattr *nla)
1955{
1956 struct nlattr *a, *tb[IFLA_INET_MAX+1];
1957 int err, rem;
1958
1959 if (dev && !__in_dev_get_rcu(dev))
1960 return -EAFNOSUPPORT;
1961
1962 err = nla_parse_nested_deprecated(tb, IFLA_INET_MAX, nla,
1963 inet_af_policy, NULL);
1964 if (err < 0)
1965 return err;
1966
1967 if (tb[IFLA_INET_CONF]) {
1968 nla_for_each_nested(a, tb[IFLA_INET_CONF], rem) {
1969 int cfgid = nla_type(a);
1970
1971 if (nla_len(a) < 4)
1972 return -EINVAL;
1973
1974 if (cfgid <= 0 || cfgid > IPV4_DEVCONF_MAX)
1975 return -EINVAL;
1976 }
1977 }
1978
1979 return 0;
1980}
1981
1982static int inet_set_link_af(struct net_device *dev, const struct nlattr *nla)
1983{
1984 struct in_device *in_dev = __in_dev_get_rcu(dev);
1985 struct nlattr *a, *tb[IFLA_INET_MAX+1];
1986 int rem;
1987
1988 if (!in_dev)
1989 return -EAFNOSUPPORT;
1990
1991 if (nla_parse_nested_deprecated(tb, IFLA_INET_MAX, nla, NULL, NULL) < 0)
1992 BUG();
1993
1994 if (tb[IFLA_INET_CONF]) {
1995 nla_for_each_nested(a, tb[IFLA_INET_CONF], rem)
1996 ipv4_devconf_set(in_dev, nla_type(a), nla_get_u32(a));
1997 }
1998
1999 return 0;
2000}
2001
2002static int inet_netconf_msgsize_devconf(int type)
2003{
2004 int size = NLMSG_ALIGN(sizeof(struct netconfmsg))
2005 + nla_total_size(4); /* NETCONFA_IFINDEX */
2006 bool all = false;
2007
2008 if (type == NETCONFA_ALL)
2009 all = true;
2010
2011 if (all || type == NETCONFA_FORWARDING)
2012 size += nla_total_size(4);
2013 if (all || type == NETCONFA_RP_FILTER)
2014 size += nla_total_size(4);
2015 if (all || type == NETCONFA_MC_FORWARDING)
2016 size += nla_total_size(4);
2017 if (all || type == NETCONFA_BC_FORWARDING)
2018 size += nla_total_size(4);
2019 if (all || type == NETCONFA_PROXY_NEIGH)
2020 size += nla_total_size(4);
2021 if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
2022 size += nla_total_size(4);
2023
2024 return size;
2025}
2026
2027static int inet_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
2028 struct ipv4_devconf *devconf, u32 portid,
2029 u32 seq, int event, unsigned int flags,
2030 int type)
2031{
2032 struct nlmsghdr *nlh;
2033 struct netconfmsg *ncm;
2034 bool all = false;
2035
2036 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
2037 flags);
2038 if (!nlh)
2039 return -EMSGSIZE;
2040
2041 if (type == NETCONFA_ALL)
2042 all = true;
2043
2044 ncm = nlmsg_data(nlh);
2045 ncm->ncm_family = AF_INET;
2046
2047 if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
2048 goto nla_put_failure;
2049
2050 if (!devconf)
2051 goto out;
2052
2053 if ((all || type == NETCONFA_FORWARDING) &&
2054 nla_put_s32(skb, NETCONFA_FORWARDING,
2055 IPV4_DEVCONF(*devconf, FORWARDING)) < 0)
2056 goto nla_put_failure;
2057 if ((all || type == NETCONFA_RP_FILTER) &&
2058 nla_put_s32(skb, NETCONFA_RP_FILTER,
2059 IPV4_DEVCONF(*devconf, RP_FILTER)) < 0)
2060 goto nla_put_failure;
2061 if ((all || type == NETCONFA_MC_FORWARDING) &&
2062 nla_put_s32(skb, NETCONFA_MC_FORWARDING,
2063 IPV4_DEVCONF(*devconf, MC_FORWARDING)) < 0)
2064 goto nla_put_failure;
2065 if ((all || type == NETCONFA_BC_FORWARDING) &&
2066 nla_put_s32(skb, NETCONFA_BC_FORWARDING,
2067 IPV4_DEVCONF(*devconf, BC_FORWARDING)) < 0)
2068 goto nla_put_failure;
2069 if ((all || type == NETCONFA_PROXY_NEIGH) &&
2070 nla_put_s32(skb, NETCONFA_PROXY_NEIGH,
2071 IPV4_DEVCONF(*devconf, PROXY_ARP)) < 0)
2072 goto nla_put_failure;
2073 if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
2074 nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
2075 IPV4_DEVCONF(*devconf, IGNORE_ROUTES_WITH_LINKDOWN)) < 0)
2076 goto nla_put_failure;
2077
2078out:
2079 nlmsg_end(skb, nlh);
2080 return 0;
2081
2082nla_put_failure:
2083 nlmsg_cancel(skb, nlh);
2084 return -EMSGSIZE;
2085}
2086
2087void inet_netconf_notify_devconf(struct net *net, int event, int type,
2088 int ifindex, struct ipv4_devconf *devconf)
2089{
2090 struct sk_buff *skb;
2091 int err = -ENOBUFS;
2092
2093 skb = nlmsg_new(inet_netconf_msgsize_devconf(type), GFP_KERNEL);
2094 if (!skb)
2095 goto errout;
2096
2097 err = inet_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
2098 event, 0, type);
2099 if (err < 0) {
2100 /* -EMSGSIZE implies BUG in inet_netconf_msgsize_devconf() */
2101 WARN_ON(err == -EMSGSIZE);
2102 kfree_skb(skb);
2103 goto errout;
2104 }
2105 rtnl_notify(skb, net, 0, RTNLGRP_IPV4_NETCONF, NULL, GFP_KERNEL);
2106 return;
2107errout:
2108 if (err < 0)
2109 rtnl_set_sk_err(net, RTNLGRP_IPV4_NETCONF, err);
2110}
2111
2112static const struct nla_policy devconf_ipv4_policy[NETCONFA_MAX+1] = {
2113 [NETCONFA_IFINDEX] = { .len = sizeof(int) },
2114 [NETCONFA_FORWARDING] = { .len = sizeof(int) },
2115 [NETCONFA_RP_FILTER] = { .len = sizeof(int) },
2116 [NETCONFA_PROXY_NEIGH] = { .len = sizeof(int) },
2117 [NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN] = { .len = sizeof(int) },
2118};
2119
2120static int inet_netconf_valid_get_req(struct sk_buff *skb,
2121 const struct nlmsghdr *nlh,
2122 struct nlattr **tb,
2123 struct netlink_ext_ack *extack)
2124{
2125 int i, err;
2126
2127 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(struct netconfmsg))) {
2128 NL_SET_ERR_MSG(extack, "ipv4: Invalid header for netconf get request");
2129 return -EINVAL;
2130 }
2131
2132 if (!netlink_strict_get_check(skb))
2133 return nlmsg_parse_deprecated(nlh, sizeof(struct netconfmsg),
2134 tb, NETCONFA_MAX,
2135 devconf_ipv4_policy, extack);
2136
2137 err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct netconfmsg),
2138 tb, NETCONFA_MAX,
2139 devconf_ipv4_policy, extack);
2140 if (err)
2141 return err;
2142
2143 for (i = 0; i <= NETCONFA_MAX; i++) {
2144 if (!tb[i])
2145 continue;
2146
2147 switch (i) {
2148 case NETCONFA_IFINDEX:
2149 break;
2150 default:
2151 NL_SET_ERR_MSG(extack, "ipv4: Unsupported attribute in netconf get request");
2152 return -EINVAL;
2153 }
2154 }
2155
2156 return 0;
2157}
2158
2159static int inet_netconf_get_devconf(struct sk_buff *in_skb,
2160 struct nlmsghdr *nlh,
2161 struct netlink_ext_ack *extack)
2162{
2163 struct net *net = sock_net(in_skb->sk);
2164 struct nlattr *tb[NETCONFA_MAX+1];
2165 struct sk_buff *skb;
2166 struct ipv4_devconf *devconf;
2167 struct in_device *in_dev;
2168 struct net_device *dev;
2169 int ifindex;
2170 int err;
2171
2172 err = inet_netconf_valid_get_req(in_skb, nlh, tb, extack);
2173 if (err)
2174 goto errout;
2175
2176 err = -EINVAL;
2177 if (!tb[NETCONFA_IFINDEX])
2178 goto errout;
2179
2180 ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
2181 switch (ifindex) {
2182 case NETCONFA_IFINDEX_ALL:
2183 devconf = net->ipv4.devconf_all;
2184 break;
2185 case NETCONFA_IFINDEX_DEFAULT:
2186 devconf = net->ipv4.devconf_dflt;
2187 break;
2188 default:
2189 dev = __dev_get_by_index(net, ifindex);
2190 if (!dev)
2191 goto errout;
2192 in_dev = __in_dev_get_rtnl(dev);
2193 if (!in_dev)
2194 goto errout;
2195 devconf = &in_dev->cnf;
2196 break;
2197 }
2198
2199 err = -ENOBUFS;
2200 skb = nlmsg_new(inet_netconf_msgsize_devconf(NETCONFA_ALL), GFP_KERNEL);
2201 if (!skb)
2202 goto errout;
2203
2204 err = inet_netconf_fill_devconf(skb, ifindex, devconf,
2205 NETLINK_CB(in_skb).portid,
2206 nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
2207 NETCONFA_ALL);
2208 if (err < 0) {
2209 /* -EMSGSIZE implies BUG in inet_netconf_msgsize_devconf() */
2210 WARN_ON(err == -EMSGSIZE);
2211 kfree_skb(skb);
2212 goto errout;
2213 }
2214 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
2215errout:
2216 return err;
2217}
2218
2219static int inet_netconf_dump_devconf(struct sk_buff *skb,
2220 struct netlink_callback *cb)
2221{
2222 const struct nlmsghdr *nlh = cb->nlh;
2223 struct net *net = sock_net(skb->sk);
2224 int h, s_h;
2225 int idx, s_idx;
2226 struct net_device *dev;
2227 struct in_device *in_dev;
2228 struct hlist_head *head;
2229
2230 if (cb->strict_check) {
2231 struct netlink_ext_ack *extack = cb->extack;
2232 struct netconfmsg *ncm;
2233
2234 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ncm))) {
2235 NL_SET_ERR_MSG(extack, "ipv4: Invalid header for netconf dump request");
2236 return -EINVAL;
2237 }
2238
2239 if (nlmsg_attrlen(nlh, sizeof(*ncm))) {
2240 NL_SET_ERR_MSG(extack, "ipv4: Invalid data after header in netconf dump request");
2241 return -EINVAL;
2242 }
2243 }
2244
2245 s_h = cb->args[0];
2246 s_idx = idx = cb->args[1];
2247
2248 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
2249 idx = 0;
2250 head = &net->dev_index_head[h];
2251 rcu_read_lock();
2252 cb->seq = atomic_read(&net->ipv4.dev_addr_genid) ^
2253 net->dev_base_seq;
2254 hlist_for_each_entry_rcu(dev, head, index_hlist) {
2255 if (idx < s_idx)
2256 goto cont;
2257 in_dev = __in_dev_get_rcu(dev);
2258 if (!in_dev)
2259 goto cont;
2260
2261 if (inet_netconf_fill_devconf(skb, dev->ifindex,
2262 &in_dev->cnf,
2263 NETLINK_CB(cb->skb).portid,
2264 nlh->nlmsg_seq,
2265 RTM_NEWNETCONF,
2266 NLM_F_MULTI,
2267 NETCONFA_ALL) < 0) {
2268 rcu_read_unlock();
2269 goto done;
2270 }
2271 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
2272cont:
2273 idx++;
2274 }
2275 rcu_read_unlock();
2276 }
2277 if (h == NETDEV_HASHENTRIES) {
2278 if (inet_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
2279 net->ipv4.devconf_all,
2280 NETLINK_CB(cb->skb).portid,
2281 nlh->nlmsg_seq,
2282 RTM_NEWNETCONF, NLM_F_MULTI,
2283 NETCONFA_ALL) < 0)
2284 goto done;
2285 else
2286 h++;
2287 }
2288 if (h == NETDEV_HASHENTRIES + 1) {
2289 if (inet_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
2290 net->ipv4.devconf_dflt,
2291 NETLINK_CB(cb->skb).portid,
2292 nlh->nlmsg_seq,
2293 RTM_NEWNETCONF, NLM_F_MULTI,
2294 NETCONFA_ALL) < 0)
2295 goto done;
2296 else
2297 h++;
2298 }
2299done:
2300 cb->args[0] = h;
2301 cb->args[1] = idx;
2302
2303 return skb->len;
2304}
2305
2306#ifdef CONFIG_SYSCTL
2307
2308static void devinet_copy_dflt_conf(struct net *net, int i)
2309{
2310 struct net_device *dev;
2311
2312 rcu_read_lock();
2313 for_each_netdev_rcu(net, dev) {
2314 struct in_device *in_dev;
2315
2316 in_dev = __in_dev_get_rcu(dev);
2317 if (in_dev && !test_bit(i, in_dev->cnf.state))
2318 in_dev->cnf.data[i] = net->ipv4.devconf_dflt->data[i];
2319 }
2320 rcu_read_unlock();
2321}
2322
2323/* called with RTNL locked */
2324static void inet_forward_change(struct net *net)
2325{
2326 struct net_device *dev;
2327 int on = IPV4_DEVCONF_ALL(net, FORWARDING);
2328
2329 IPV4_DEVCONF_ALL(net, ACCEPT_REDIRECTS) = !on;
2330 IPV4_DEVCONF_DFLT(net, FORWARDING) = on;
2331 inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
2332 NETCONFA_FORWARDING,
2333 NETCONFA_IFINDEX_ALL,
2334 net->ipv4.devconf_all);
2335 inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
2336 NETCONFA_FORWARDING,
2337 NETCONFA_IFINDEX_DEFAULT,
2338 net->ipv4.devconf_dflt);
2339
2340 for_each_netdev(net, dev) {
2341 struct in_device *in_dev;
2342
2343 if (on)
2344 dev_disable_lro(dev);
2345
2346 in_dev = __in_dev_get_rtnl(dev);
2347 if (in_dev) {
2348 IN_DEV_CONF_SET(in_dev, FORWARDING, on);
2349 inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
2350 NETCONFA_FORWARDING,
2351 dev->ifindex, &in_dev->cnf);
2352 }
2353 }
2354}
2355
2356static int devinet_conf_ifindex(struct net *net, struct ipv4_devconf *cnf)
2357{
2358 if (cnf == net->ipv4.devconf_dflt)
2359 return NETCONFA_IFINDEX_DEFAULT;
2360 else if (cnf == net->ipv4.devconf_all)
2361 return NETCONFA_IFINDEX_ALL;
2362 else {
2363 struct in_device *idev
2364 = container_of(cnf, struct in_device, cnf);
2365 return idev->dev->ifindex;
2366 }
2367}
2368
2369static int devinet_conf_proc(struct ctl_table *ctl, int write,
2370 void *buffer, size_t *lenp, loff_t *ppos)
2371{
2372 int old_value = *(int *)ctl->data;
2373 int ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
2374 int new_value = *(int *)ctl->data;
2375
2376 if (write) {
2377 struct ipv4_devconf *cnf = ctl->extra1;
2378 struct net *net = ctl->extra2;
2379 int i = (int *)ctl->data - cnf->data;
2380 int ifindex;
2381
2382 set_bit(i, cnf->state);
2383
2384 if (cnf == net->ipv4.devconf_dflt)
2385 devinet_copy_dflt_conf(net, i);
2386 if (i == IPV4_DEVCONF_ACCEPT_LOCAL - 1 ||
2387 i == IPV4_DEVCONF_ROUTE_LOCALNET - 1)
2388 if ((new_value == 0) && (old_value != 0))
2389 rt_cache_flush(net);
2390
2391 if (i == IPV4_DEVCONF_BC_FORWARDING - 1 &&
2392 new_value != old_value)
2393 rt_cache_flush(net);
2394
2395 if (i == IPV4_DEVCONF_RP_FILTER - 1 &&
2396 new_value != old_value) {
2397 ifindex = devinet_conf_ifindex(net, cnf);
2398 inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
2399 NETCONFA_RP_FILTER,
2400 ifindex, cnf);
2401 }
2402 if (i == IPV4_DEVCONF_PROXY_ARP - 1 &&
2403 new_value != old_value) {
2404 ifindex = devinet_conf_ifindex(net, cnf);
2405 inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
2406 NETCONFA_PROXY_NEIGH,
2407 ifindex, cnf);
2408 }
2409 if (i == IPV4_DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN - 1 &&
2410 new_value != old_value) {
2411 ifindex = devinet_conf_ifindex(net, cnf);
2412 inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
2413 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
2414 ifindex, cnf);
2415 }
2416 }
2417
2418 return ret;
2419}
2420
2421static int devinet_sysctl_forward(struct ctl_table *ctl, int write,
2422 void *buffer, size_t *lenp, loff_t *ppos)
2423{
2424 int *valp = ctl->data;
2425 int val = *valp;
2426 loff_t pos = *ppos;
2427 int ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
2428
2429 if (write && *valp != val) {
2430 struct net *net = ctl->extra2;
2431
2432 if (valp != &IPV4_DEVCONF_DFLT(net, FORWARDING)) {
2433 if (!rtnl_trylock()) {
2434 /* Restore the original values before restarting */
2435 *valp = val;
2436 *ppos = pos;
2437 return restart_syscall();
2438 }
2439 if (valp == &IPV4_DEVCONF_ALL(net, FORWARDING)) {
2440 inet_forward_change(net);
2441 } else {
2442 struct ipv4_devconf *cnf = ctl->extra1;
2443 struct in_device *idev =
2444 container_of(cnf, struct in_device, cnf);
2445 if (*valp)
2446 dev_disable_lro(idev->dev);
2447 inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
2448 NETCONFA_FORWARDING,
2449 idev->dev->ifindex,
2450 cnf);
2451 }
2452 rtnl_unlock();
2453 rt_cache_flush(net);
2454 } else
2455 inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
2456 NETCONFA_FORWARDING,
2457 NETCONFA_IFINDEX_DEFAULT,
2458 net->ipv4.devconf_dflt);
2459 }
2460
2461 return ret;
2462}
2463
2464static int ipv4_doint_and_flush(struct ctl_table *ctl, int write,
2465 void *buffer, size_t *lenp, loff_t *ppos)
2466{
2467 int *valp = ctl->data;
2468 int val = *valp;
2469 int ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
2470 struct net *net = ctl->extra2;
2471
2472 if (write && *valp != val)
2473 rt_cache_flush(net);
2474
2475 return ret;
2476}
2477
2478#define DEVINET_SYSCTL_ENTRY(attr, name, mval, proc) \
2479 { \
2480 .procname = name, \
2481 .data = ipv4_devconf.data + \
2482 IPV4_DEVCONF_ ## attr - 1, \
2483 .maxlen = sizeof(int), \
2484 .mode = mval, \
2485 .proc_handler = proc, \
2486 .extra1 = &ipv4_devconf, \
2487 }
2488
2489#define DEVINET_SYSCTL_RW_ENTRY(attr, name) \
2490 DEVINET_SYSCTL_ENTRY(attr, name, 0644, devinet_conf_proc)
2491
2492#define DEVINET_SYSCTL_RO_ENTRY(attr, name) \
2493 DEVINET_SYSCTL_ENTRY(attr, name, 0444, devinet_conf_proc)
2494
2495#define DEVINET_SYSCTL_COMPLEX_ENTRY(attr, name, proc) \
2496 DEVINET_SYSCTL_ENTRY(attr, name, 0644, proc)
2497
2498#define DEVINET_SYSCTL_FLUSHING_ENTRY(attr, name) \
2499 DEVINET_SYSCTL_COMPLEX_ENTRY(attr, name, ipv4_doint_and_flush)
2500
2501static struct devinet_sysctl_table {
2502 struct ctl_table_header *sysctl_header;
2503 struct ctl_table devinet_vars[__IPV4_DEVCONF_MAX];
2504} devinet_sysctl = {
2505 .devinet_vars = {
2506 DEVINET_SYSCTL_COMPLEX_ENTRY(FORWARDING, "forwarding",
2507 devinet_sysctl_forward),
2508 DEVINET_SYSCTL_RO_ENTRY(MC_FORWARDING, "mc_forwarding"),
2509 DEVINET_SYSCTL_RW_ENTRY(BC_FORWARDING, "bc_forwarding"),
2510
2511 DEVINET_SYSCTL_RW_ENTRY(ACCEPT_REDIRECTS, "accept_redirects"),
2512 DEVINET_SYSCTL_RW_ENTRY(SECURE_REDIRECTS, "secure_redirects"),
2513 DEVINET_SYSCTL_RW_ENTRY(SHARED_MEDIA, "shared_media"),
2514 DEVINET_SYSCTL_RW_ENTRY(RP_FILTER, "rp_filter"),
2515 DEVINET_SYSCTL_RW_ENTRY(SEND_REDIRECTS, "send_redirects"),
2516 DEVINET_SYSCTL_RW_ENTRY(ACCEPT_SOURCE_ROUTE,
2517 "accept_source_route"),
2518 DEVINET_SYSCTL_RW_ENTRY(ACCEPT_LOCAL, "accept_local"),
2519 DEVINET_SYSCTL_RW_ENTRY(SRC_VMARK, "src_valid_mark"),
2520 DEVINET_SYSCTL_RW_ENTRY(PROXY_ARP, "proxy_arp"),
2521 DEVINET_SYSCTL_RW_ENTRY(MEDIUM_ID, "medium_id"),
2522 DEVINET_SYSCTL_RW_ENTRY(BOOTP_RELAY, "bootp_relay"),
2523 DEVINET_SYSCTL_RW_ENTRY(LOG_MARTIANS, "log_martians"),
2524 DEVINET_SYSCTL_RW_ENTRY(TAG, "tag"),
2525 DEVINET_SYSCTL_RW_ENTRY(ARPFILTER, "arp_filter"),
2526 DEVINET_SYSCTL_RW_ENTRY(ARP_ANNOUNCE, "arp_announce"),
2527 DEVINET_SYSCTL_RW_ENTRY(ARP_IGNORE, "arp_ignore"),
2528 DEVINET_SYSCTL_RW_ENTRY(ARP_ACCEPT, "arp_accept"),
2529 DEVINET_SYSCTL_RW_ENTRY(ARP_NOTIFY, "arp_notify"),
2530 DEVINET_SYSCTL_RW_ENTRY(PROXY_ARP_PVLAN, "proxy_arp_pvlan"),
2531 DEVINET_SYSCTL_RW_ENTRY(FORCE_IGMP_VERSION,
2532 "force_igmp_version"),
2533 DEVINET_SYSCTL_RW_ENTRY(IGMPV2_UNSOLICITED_REPORT_INTERVAL,
2534 "igmpv2_unsolicited_report_interval"),
2535 DEVINET_SYSCTL_RW_ENTRY(IGMPV3_UNSOLICITED_REPORT_INTERVAL,
2536 "igmpv3_unsolicited_report_interval"),
2537 DEVINET_SYSCTL_RW_ENTRY(IGNORE_ROUTES_WITH_LINKDOWN,
2538 "ignore_routes_with_linkdown"),
2539 DEVINET_SYSCTL_RW_ENTRY(DROP_GRATUITOUS_ARP,
2540 "drop_gratuitous_arp"),
2541
2542 DEVINET_SYSCTL_FLUSHING_ENTRY(NOXFRM, "disable_xfrm"),
2543 DEVINET_SYSCTL_FLUSHING_ENTRY(NOPOLICY, "disable_policy"),
2544 DEVINET_SYSCTL_FLUSHING_ENTRY(PROMOTE_SECONDARIES,
2545 "promote_secondaries"),
2546 DEVINET_SYSCTL_FLUSHING_ENTRY(ROUTE_LOCALNET,
2547 "route_localnet"),
2548 DEVINET_SYSCTL_FLUSHING_ENTRY(DROP_UNICAST_IN_L2_MULTICAST,
2549 "drop_unicast_in_l2_multicast"),
2550 },
2551};
2552
2553static int __devinet_sysctl_register(struct net *net, char *dev_name,
2554 int ifindex, struct ipv4_devconf *p)
2555{
2556 int i;
2557 struct devinet_sysctl_table *t;
2558 char path[sizeof("net/ipv4/conf/") + IFNAMSIZ];
2559
2560 t = kmemdup(&devinet_sysctl, sizeof(*t), GFP_KERNEL);
2561 if (!t)
2562 goto out;
2563
2564 for (i = 0; i < ARRAY_SIZE(t->devinet_vars) - 1; i++) {
2565 t->devinet_vars[i].data += (char *)p - (char *)&ipv4_devconf;
2566 t->devinet_vars[i].extra1 = p;
2567 t->devinet_vars[i].extra2 = net;
2568 }
2569
2570 snprintf(path, sizeof(path), "net/ipv4/conf/%s", dev_name);
2571
2572 t->sysctl_header = register_net_sysctl(net, path, t->devinet_vars);
2573 if (!t->sysctl_header)
2574 goto free;
2575
2576 p->sysctl = t;
2577
2578 inet_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
2579 ifindex, p);
2580 return 0;
2581
2582free:
2583 kfree(t);
2584out:
2585 return -ENOBUFS;
2586}
2587
2588static void __devinet_sysctl_unregister(struct net *net,
2589 struct ipv4_devconf *cnf, int ifindex)
2590{
2591 struct devinet_sysctl_table *t = cnf->sysctl;
2592
2593 if (t) {
2594 cnf->sysctl = NULL;
2595 unregister_net_sysctl_table(t->sysctl_header);
2596 kfree(t);
2597 }
2598
2599 inet_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
2600}
2601
2602static int devinet_sysctl_register(struct in_device *idev)
2603{
2604 int err;
2605
2606 if (!sysctl_dev_name_is_allowed(idev->dev->name))
2607 return -EINVAL;
2608
2609 err = neigh_sysctl_register(idev->dev, idev->arp_parms, NULL);
2610 if (err)
2611 return err;
2612 err = __devinet_sysctl_register(dev_net(idev->dev), idev->dev->name,
2613 idev->dev->ifindex, &idev->cnf);
2614 if (err)
2615 neigh_sysctl_unregister(idev->arp_parms);
2616 return err;
2617}
2618
2619static void devinet_sysctl_unregister(struct in_device *idev)
2620{
2621 struct net *net = dev_net(idev->dev);
2622
2623 __devinet_sysctl_unregister(net, &idev->cnf, idev->dev->ifindex);
2624 neigh_sysctl_unregister(idev->arp_parms);
2625}
2626
2627static struct ctl_table ctl_forward_entry[] = {
2628 {
2629 .procname = "ip_forward",
2630 .data = &ipv4_devconf.data[
2631 IPV4_DEVCONF_FORWARDING - 1],
2632 .maxlen = sizeof(int),
2633 .mode = 0644,
2634 .proc_handler = devinet_sysctl_forward,
2635 .extra1 = &ipv4_devconf,
2636 .extra2 = &init_net,
2637 },
2638 { },
2639};
2640#endif
2641
2642static __net_init int devinet_init_net(struct net *net)
2643{
2644 int err;
2645 struct ipv4_devconf *all, *dflt;
2646#ifdef CONFIG_SYSCTL
2647 struct ctl_table *tbl;
2648 struct ctl_table_header *forw_hdr;
2649#endif
2650
2651 err = -ENOMEM;
2652 all = kmemdup(&ipv4_devconf, sizeof(ipv4_devconf), GFP_KERNEL);
2653 if (!all)
2654 goto err_alloc_all;
2655
2656 dflt = kmemdup(&ipv4_devconf_dflt, sizeof(ipv4_devconf_dflt), GFP_KERNEL);
2657 if (!dflt)
2658 goto err_alloc_dflt;
2659
2660#ifdef CONFIG_SYSCTL
2661 tbl = kmemdup(ctl_forward_entry, sizeof(ctl_forward_entry), GFP_KERNEL);
2662 if (!tbl)
2663 goto err_alloc_ctl;
2664
2665 tbl[0].data = &all->data[IPV4_DEVCONF_FORWARDING - 1];
2666 tbl[0].extra1 = all;
2667 tbl[0].extra2 = net;
2668#endif
2669
2670 if (!net_eq(net, &init_net)) {
2671 if (IS_ENABLED(CONFIG_SYSCTL) &&
2672 sysctl_devconf_inherit_init_net == 3) {
2673 /* copy from the current netns */
2674 memcpy(all, current->nsproxy->net_ns->ipv4.devconf_all,
2675 sizeof(ipv4_devconf));
2676 memcpy(dflt,
2677 current->nsproxy->net_ns->ipv4.devconf_dflt,
2678 sizeof(ipv4_devconf_dflt));
2679 } else if (!IS_ENABLED(CONFIG_SYSCTL) ||
2680 sysctl_devconf_inherit_init_net != 2) {
2681 /* inherit == 0 or 1: copy from init_net */
2682 memcpy(all, init_net.ipv4.devconf_all,
2683 sizeof(ipv4_devconf));
2684 memcpy(dflt, init_net.ipv4.devconf_dflt,
2685 sizeof(ipv4_devconf_dflt));
2686 }
2687 /* else inherit == 2: use compiled values */
2688 }
2689
2690#ifdef CONFIG_SYSCTL
2691 err = __devinet_sysctl_register(net, "all", NETCONFA_IFINDEX_ALL, all);
2692 if (err < 0)
2693 goto err_reg_all;
2694
2695 err = __devinet_sysctl_register(net, "default",
2696 NETCONFA_IFINDEX_DEFAULT, dflt);
2697 if (err < 0)
2698 goto err_reg_dflt;
2699
2700 err = -ENOMEM;
2701 forw_hdr = register_net_sysctl(net, "net/ipv4", tbl);
2702 if (!forw_hdr)
2703 goto err_reg_ctl;
2704 net->ipv4.forw_hdr = forw_hdr;
2705#endif
2706
2707 net->ipv4.devconf_all = all;
2708 net->ipv4.devconf_dflt = dflt;
2709 return 0;
2710
2711#ifdef CONFIG_SYSCTL
2712err_reg_ctl:
2713 __devinet_sysctl_unregister(net, dflt, NETCONFA_IFINDEX_DEFAULT);
2714err_reg_dflt:
2715 __devinet_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
2716err_reg_all:
2717 kfree(tbl);
2718err_alloc_ctl:
2719#endif
2720 kfree(dflt);
2721err_alloc_dflt:
2722 kfree(all);
2723err_alloc_all:
2724 return err;
2725}
2726
2727static __net_exit void devinet_exit_net(struct net *net)
2728{
2729#ifdef CONFIG_SYSCTL
2730 struct ctl_table *tbl;
2731
2732 tbl = net->ipv4.forw_hdr->ctl_table_arg;
2733 unregister_net_sysctl_table(net->ipv4.forw_hdr);
2734 __devinet_sysctl_unregister(net, net->ipv4.devconf_dflt,
2735 NETCONFA_IFINDEX_DEFAULT);
2736 __devinet_sysctl_unregister(net, net->ipv4.devconf_all,
2737 NETCONFA_IFINDEX_ALL);
2738 kfree(tbl);
2739#endif
2740 kfree(net->ipv4.devconf_dflt);
2741 kfree(net->ipv4.devconf_all);
2742}
2743
2744static __net_initdata struct pernet_operations devinet_ops = {
2745 .init = devinet_init_net,
2746 .exit = devinet_exit_net,
2747};
2748
2749static struct rtnl_af_ops inet_af_ops __read_mostly = {
2750 .family = AF_INET,
2751 .fill_link_af = inet_fill_link_af,
2752 .get_link_af_size = inet_get_link_af_size,
2753 .validate_link_af = inet_validate_link_af,
2754 .set_link_af = inet_set_link_af,
2755};
2756
2757void __init devinet_init(void)
2758{
2759 int i;
2760
2761 for (i = 0; i < IN4_ADDR_HSIZE; i++)
2762 INIT_HLIST_HEAD(&inet_addr_lst[i]);
2763
2764 register_pernet_subsys(&devinet_ops);
2765
2766 register_gifconf(PF_INET, inet_gifconf);
2767 register_netdevice_notifier(&ip_netdev_notifier);
2768
2769 queue_delayed_work(system_power_efficient_wq, &check_lifetime_work, 0);
2770
2771 rtnl_af_register(&inet_af_ops);
2772
2773 rtnl_register(PF_INET, RTM_NEWADDR, inet_rtm_newaddr, NULL, 0);
2774 rtnl_register(PF_INET, RTM_DELADDR, inet_rtm_deladdr, NULL, 0);
2775 rtnl_register(PF_INET, RTM_GETADDR, NULL, inet_dump_ifaddr, 0);
2776 rtnl_register(PF_INET, RTM_GETNETCONF, inet_netconf_get_devconf,
2777 inet_netconf_dump_devconf, 0);
2778}