Loading...
1/*
2 * NET3 IP device support routines.
3 *
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public License
6 * as published by the Free Software Foundation; either version
7 * 2 of the License, or (at your option) any later version.
8 *
9 * Derived from the IP parts of dev.c 1.0.19
10 * Authors: Ross Biro
11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Mark Evans, <evansmp@uhura.aston.ac.uk>
13 *
14 * Additional Authors:
15 * Alan Cox, <gw4pts@gw4pts.ampr.org>
16 * Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
17 *
18 * Changes:
19 * Alexey Kuznetsov: pa_* fields are replaced with ifaddr
20 * lists.
21 * Cyrus Durgin: updated for kmod
22 * Matthias Andree: in devinet_ioctl, compare label and
23 * address (4.4BSD alias style support),
24 * fall back to comparing just the label
25 * if no match found.
26 */
27
28
29#include <linux/uaccess.h>
30#include <linux/bitops.h>
31#include <linux/capability.h>
32#include <linux/module.h>
33#include <linux/types.h>
34#include <linux/kernel.h>
35#include <linux/sched/signal.h>
36#include <linux/string.h>
37#include <linux/mm.h>
38#include <linux/socket.h>
39#include <linux/sockios.h>
40#include <linux/in.h>
41#include <linux/errno.h>
42#include <linux/interrupt.h>
43#include <linux/if_addr.h>
44#include <linux/if_ether.h>
45#include <linux/inet.h>
46#include <linux/netdevice.h>
47#include <linux/etherdevice.h>
48#include <linux/skbuff.h>
49#include <linux/init.h>
50#include <linux/notifier.h>
51#include <linux/inetdevice.h>
52#include <linux/igmp.h>
53#include <linux/slab.h>
54#include <linux/hash.h>
55#ifdef CONFIG_SYSCTL
56#include <linux/sysctl.h>
57#endif
58#include <linux/kmod.h>
59#include <linux/netconf.h>
60
61#include <net/arp.h>
62#include <net/ip.h>
63#include <net/route.h>
64#include <net/ip_fib.h>
65#include <net/rtnetlink.h>
66#include <net/net_namespace.h>
67#include <net/addrconf.h>
68
69static struct ipv4_devconf ipv4_devconf = {
70 .data = {
71 [IPV4_DEVCONF_ACCEPT_REDIRECTS - 1] = 1,
72 [IPV4_DEVCONF_SEND_REDIRECTS - 1] = 1,
73 [IPV4_DEVCONF_SECURE_REDIRECTS - 1] = 1,
74 [IPV4_DEVCONF_SHARED_MEDIA - 1] = 1,
75 [IPV4_DEVCONF_IGMPV2_UNSOLICITED_REPORT_INTERVAL - 1] = 10000 /*ms*/,
76 [IPV4_DEVCONF_IGMPV3_UNSOLICITED_REPORT_INTERVAL - 1] = 1000 /*ms*/,
77 },
78};
79
80static struct ipv4_devconf ipv4_devconf_dflt = {
81 .data = {
82 [IPV4_DEVCONF_ACCEPT_REDIRECTS - 1] = 1,
83 [IPV4_DEVCONF_SEND_REDIRECTS - 1] = 1,
84 [IPV4_DEVCONF_SECURE_REDIRECTS - 1] = 1,
85 [IPV4_DEVCONF_SHARED_MEDIA - 1] = 1,
86 [IPV4_DEVCONF_ACCEPT_SOURCE_ROUTE - 1] = 1,
87 [IPV4_DEVCONF_IGMPV2_UNSOLICITED_REPORT_INTERVAL - 1] = 10000 /*ms*/,
88 [IPV4_DEVCONF_IGMPV3_UNSOLICITED_REPORT_INTERVAL - 1] = 1000 /*ms*/,
89 },
90};
91
92#define IPV4_DEVCONF_DFLT(net, attr) \
93 IPV4_DEVCONF((*net->ipv4.devconf_dflt), attr)
94
95static const struct nla_policy ifa_ipv4_policy[IFA_MAX+1] = {
96 [IFA_LOCAL] = { .type = NLA_U32 },
97 [IFA_ADDRESS] = { .type = NLA_U32 },
98 [IFA_BROADCAST] = { .type = NLA_U32 },
99 [IFA_LABEL] = { .type = NLA_STRING, .len = IFNAMSIZ - 1 },
100 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
101 [IFA_FLAGS] = { .type = NLA_U32 },
102};
103
104#define IN4_ADDR_HSIZE_SHIFT 8
105#define IN4_ADDR_HSIZE (1U << IN4_ADDR_HSIZE_SHIFT)
106
107static struct hlist_head inet_addr_lst[IN4_ADDR_HSIZE];
108
109static u32 inet_addr_hash(const struct net *net, __be32 addr)
110{
111 u32 val = (__force u32) addr ^ net_hash_mix(net);
112
113 return hash_32(val, IN4_ADDR_HSIZE_SHIFT);
114}
115
116static void inet_hash_insert(struct net *net, struct in_ifaddr *ifa)
117{
118 u32 hash = inet_addr_hash(net, ifa->ifa_local);
119
120 ASSERT_RTNL();
121 hlist_add_head_rcu(&ifa->hash, &inet_addr_lst[hash]);
122}
123
124static void inet_hash_remove(struct in_ifaddr *ifa)
125{
126 ASSERT_RTNL();
127 hlist_del_init_rcu(&ifa->hash);
128}
129
130/**
131 * __ip_dev_find - find the first device with a given source address.
132 * @net: the net namespace
133 * @addr: the source address
134 * @devref: if true, take a reference on the found device
135 *
136 * If a caller uses devref=false, it should be protected by RCU, or RTNL
137 */
138struct net_device *__ip_dev_find(struct net *net, __be32 addr, bool devref)
139{
140 struct net_device *result = NULL;
141 struct in_ifaddr *ifa;
142
143 rcu_read_lock();
144 ifa = inet_lookup_ifaddr_rcu(net, addr);
145 if (!ifa) {
146 struct flowi4 fl4 = { .daddr = addr };
147 struct fib_result res = { 0 };
148 struct fib_table *local;
149
150 /* Fallback to FIB local table so that communication
151 * over loopback subnets work.
152 */
153 local = fib_get_table(net, RT_TABLE_LOCAL);
154 if (local &&
155 !fib_table_lookup(local, &fl4, &res, FIB_LOOKUP_NOREF) &&
156 res.type == RTN_LOCAL)
157 result = FIB_RES_DEV(res);
158 } else {
159 result = ifa->ifa_dev->dev;
160 }
161 if (result && devref)
162 dev_hold(result);
163 rcu_read_unlock();
164 return result;
165}
166EXPORT_SYMBOL(__ip_dev_find);
167
168/* called under RCU lock */
169struct in_ifaddr *inet_lookup_ifaddr_rcu(struct net *net, __be32 addr)
170{
171 u32 hash = inet_addr_hash(net, addr);
172 struct in_ifaddr *ifa;
173
174 hlist_for_each_entry_rcu(ifa, &inet_addr_lst[hash], hash)
175 if (ifa->ifa_local == addr &&
176 net_eq(dev_net(ifa->ifa_dev->dev), net))
177 return ifa;
178
179 return NULL;
180}
181
182static void rtmsg_ifa(int event, struct in_ifaddr *, struct nlmsghdr *, u32);
183
184static BLOCKING_NOTIFIER_HEAD(inetaddr_chain);
185static BLOCKING_NOTIFIER_HEAD(inetaddr_validator_chain);
186static void inet_del_ifa(struct in_device *in_dev, struct in_ifaddr **ifap,
187 int destroy);
188#ifdef CONFIG_SYSCTL
189static int devinet_sysctl_register(struct in_device *idev);
190static void devinet_sysctl_unregister(struct in_device *idev);
191#else
192static int devinet_sysctl_register(struct in_device *idev)
193{
194 return 0;
195}
196static void devinet_sysctl_unregister(struct in_device *idev)
197{
198}
199#endif
200
201/* Locks all the inet devices. */
202
203static struct in_ifaddr *inet_alloc_ifa(void)
204{
205 return kzalloc(sizeof(struct in_ifaddr), GFP_KERNEL);
206}
207
208static void inet_rcu_free_ifa(struct rcu_head *head)
209{
210 struct in_ifaddr *ifa = container_of(head, struct in_ifaddr, rcu_head);
211 if (ifa->ifa_dev)
212 in_dev_put(ifa->ifa_dev);
213 kfree(ifa);
214}
215
216static void inet_free_ifa(struct in_ifaddr *ifa)
217{
218 call_rcu(&ifa->rcu_head, inet_rcu_free_ifa);
219}
220
221void in_dev_finish_destroy(struct in_device *idev)
222{
223 struct net_device *dev = idev->dev;
224
225 WARN_ON(idev->ifa_list);
226 WARN_ON(idev->mc_list);
227 kfree(rcu_dereference_protected(idev->mc_hash, 1));
228#ifdef NET_REFCNT_DEBUG
229 pr_debug("%s: %p=%s\n", __func__, idev, dev ? dev->name : "NIL");
230#endif
231 dev_put(dev);
232 if (!idev->dead)
233 pr_err("Freeing alive in_device %p\n", idev);
234 else
235 kfree(idev);
236}
237EXPORT_SYMBOL(in_dev_finish_destroy);
238
239static struct in_device *inetdev_init(struct net_device *dev)
240{
241 struct in_device *in_dev;
242 int err = -ENOMEM;
243
244 ASSERT_RTNL();
245
246 in_dev = kzalloc(sizeof(*in_dev), GFP_KERNEL);
247 if (!in_dev)
248 goto out;
249 memcpy(&in_dev->cnf, dev_net(dev)->ipv4.devconf_dflt,
250 sizeof(in_dev->cnf));
251 in_dev->cnf.sysctl = NULL;
252 in_dev->dev = dev;
253 in_dev->arp_parms = neigh_parms_alloc(dev, &arp_tbl);
254 if (!in_dev->arp_parms)
255 goto out_kfree;
256 if (IPV4_DEVCONF(in_dev->cnf, FORWARDING))
257 dev_disable_lro(dev);
258 /* Reference in_dev->dev */
259 dev_hold(dev);
260 /* Account for reference dev->ip_ptr (below) */
261 refcount_set(&in_dev->refcnt, 1);
262
263 err = devinet_sysctl_register(in_dev);
264 if (err) {
265 in_dev->dead = 1;
266 in_dev_put(in_dev);
267 in_dev = NULL;
268 goto out;
269 }
270 ip_mc_init_dev(in_dev);
271 if (dev->flags & IFF_UP)
272 ip_mc_up(in_dev);
273
274 /* we can receive as soon as ip_ptr is set -- do this last */
275 rcu_assign_pointer(dev->ip_ptr, in_dev);
276out:
277 return in_dev ?: ERR_PTR(err);
278out_kfree:
279 kfree(in_dev);
280 in_dev = NULL;
281 goto out;
282}
283
284static void in_dev_rcu_put(struct rcu_head *head)
285{
286 struct in_device *idev = container_of(head, struct in_device, rcu_head);
287 in_dev_put(idev);
288}
289
290static void inetdev_destroy(struct in_device *in_dev)
291{
292 struct in_ifaddr *ifa;
293 struct net_device *dev;
294
295 ASSERT_RTNL();
296
297 dev = in_dev->dev;
298
299 in_dev->dead = 1;
300
301 ip_mc_destroy_dev(in_dev);
302
303 while ((ifa = in_dev->ifa_list) != NULL) {
304 inet_del_ifa(in_dev, &in_dev->ifa_list, 0);
305 inet_free_ifa(ifa);
306 }
307
308 RCU_INIT_POINTER(dev->ip_ptr, NULL);
309
310 devinet_sysctl_unregister(in_dev);
311 neigh_parms_release(&arp_tbl, in_dev->arp_parms);
312 arp_ifdown(dev);
313
314 call_rcu(&in_dev->rcu_head, in_dev_rcu_put);
315}
316
317int inet_addr_onlink(struct in_device *in_dev, __be32 a, __be32 b)
318{
319 rcu_read_lock();
320 for_primary_ifa(in_dev) {
321 if (inet_ifa_match(a, ifa)) {
322 if (!b || inet_ifa_match(b, ifa)) {
323 rcu_read_unlock();
324 return 1;
325 }
326 }
327 } endfor_ifa(in_dev);
328 rcu_read_unlock();
329 return 0;
330}
331
332static void __inet_del_ifa(struct in_device *in_dev, struct in_ifaddr **ifap,
333 int destroy, struct nlmsghdr *nlh, u32 portid)
334{
335 struct in_ifaddr *promote = NULL;
336 struct in_ifaddr *ifa, *ifa1 = *ifap;
337 struct in_ifaddr *last_prim = in_dev->ifa_list;
338 struct in_ifaddr *prev_prom = NULL;
339 int do_promote = IN_DEV_PROMOTE_SECONDARIES(in_dev);
340
341 ASSERT_RTNL();
342
343 if (in_dev->dead)
344 goto no_promotions;
345
346 /* 1. Deleting primary ifaddr forces deletion all secondaries
347 * unless alias promotion is set
348 **/
349
350 if (!(ifa1->ifa_flags & IFA_F_SECONDARY)) {
351 struct in_ifaddr **ifap1 = &ifa1->ifa_next;
352
353 while ((ifa = *ifap1) != NULL) {
354 if (!(ifa->ifa_flags & IFA_F_SECONDARY) &&
355 ifa1->ifa_scope <= ifa->ifa_scope)
356 last_prim = ifa;
357
358 if (!(ifa->ifa_flags & IFA_F_SECONDARY) ||
359 ifa1->ifa_mask != ifa->ifa_mask ||
360 !inet_ifa_match(ifa1->ifa_address, ifa)) {
361 ifap1 = &ifa->ifa_next;
362 prev_prom = ifa;
363 continue;
364 }
365
366 if (!do_promote) {
367 inet_hash_remove(ifa);
368 *ifap1 = ifa->ifa_next;
369
370 rtmsg_ifa(RTM_DELADDR, ifa, nlh, portid);
371 blocking_notifier_call_chain(&inetaddr_chain,
372 NETDEV_DOWN, ifa);
373 inet_free_ifa(ifa);
374 } else {
375 promote = ifa;
376 break;
377 }
378 }
379 }
380
381 /* On promotion all secondaries from subnet are changing
382 * the primary IP, we must remove all their routes silently
383 * and later to add them back with new prefsrc. Do this
384 * while all addresses are on the device list.
385 */
386 for (ifa = promote; ifa; ifa = ifa->ifa_next) {
387 if (ifa1->ifa_mask == ifa->ifa_mask &&
388 inet_ifa_match(ifa1->ifa_address, ifa))
389 fib_del_ifaddr(ifa, ifa1);
390 }
391
392no_promotions:
393 /* 2. Unlink it */
394
395 *ifap = ifa1->ifa_next;
396 inet_hash_remove(ifa1);
397
398 /* 3. Announce address deletion */
399
400 /* Send message first, then call notifier.
401 At first sight, FIB update triggered by notifier
402 will refer to already deleted ifaddr, that could confuse
403 netlink listeners. It is not true: look, gated sees
404 that route deleted and if it still thinks that ifaddr
405 is valid, it will try to restore deleted routes... Grr.
406 So that, this order is correct.
407 */
408 rtmsg_ifa(RTM_DELADDR, ifa1, nlh, portid);
409 blocking_notifier_call_chain(&inetaddr_chain, NETDEV_DOWN, ifa1);
410
411 if (promote) {
412 struct in_ifaddr *next_sec = promote->ifa_next;
413
414 if (prev_prom) {
415 prev_prom->ifa_next = promote->ifa_next;
416 promote->ifa_next = last_prim->ifa_next;
417 last_prim->ifa_next = promote;
418 }
419
420 promote->ifa_flags &= ~IFA_F_SECONDARY;
421 rtmsg_ifa(RTM_NEWADDR, promote, nlh, portid);
422 blocking_notifier_call_chain(&inetaddr_chain,
423 NETDEV_UP, promote);
424 for (ifa = next_sec; ifa; ifa = ifa->ifa_next) {
425 if (ifa1->ifa_mask != ifa->ifa_mask ||
426 !inet_ifa_match(ifa1->ifa_address, ifa))
427 continue;
428 fib_add_ifaddr(ifa);
429 }
430
431 }
432 if (destroy)
433 inet_free_ifa(ifa1);
434}
435
436static void inet_del_ifa(struct in_device *in_dev, struct in_ifaddr **ifap,
437 int destroy)
438{
439 __inet_del_ifa(in_dev, ifap, destroy, NULL, 0);
440}
441
442static void check_lifetime(struct work_struct *work);
443
444static DECLARE_DELAYED_WORK(check_lifetime_work, check_lifetime);
445
446static int __inet_insert_ifa(struct in_ifaddr *ifa, struct nlmsghdr *nlh,
447 u32 portid, struct netlink_ext_ack *extack)
448{
449 struct in_device *in_dev = ifa->ifa_dev;
450 struct in_ifaddr *ifa1, **ifap, **last_primary;
451 struct in_validator_info ivi;
452 int ret;
453
454 ASSERT_RTNL();
455
456 if (!ifa->ifa_local) {
457 inet_free_ifa(ifa);
458 return 0;
459 }
460
461 ifa->ifa_flags &= ~IFA_F_SECONDARY;
462 last_primary = &in_dev->ifa_list;
463
464 for (ifap = &in_dev->ifa_list; (ifa1 = *ifap) != NULL;
465 ifap = &ifa1->ifa_next) {
466 if (!(ifa1->ifa_flags & IFA_F_SECONDARY) &&
467 ifa->ifa_scope <= ifa1->ifa_scope)
468 last_primary = &ifa1->ifa_next;
469 if (ifa1->ifa_mask == ifa->ifa_mask &&
470 inet_ifa_match(ifa1->ifa_address, ifa)) {
471 if (ifa1->ifa_local == ifa->ifa_local) {
472 inet_free_ifa(ifa);
473 return -EEXIST;
474 }
475 if (ifa1->ifa_scope != ifa->ifa_scope) {
476 inet_free_ifa(ifa);
477 return -EINVAL;
478 }
479 ifa->ifa_flags |= IFA_F_SECONDARY;
480 }
481 }
482
483 /* Allow any devices that wish to register ifaddr validtors to weigh
484 * in now, before changes are committed. The rntl lock is serializing
485 * access here, so the state should not change between a validator call
486 * and a final notify on commit. This isn't invoked on promotion under
487 * the assumption that validators are checking the address itself, and
488 * not the flags.
489 */
490 ivi.ivi_addr = ifa->ifa_address;
491 ivi.ivi_dev = ifa->ifa_dev;
492 ivi.extack = extack;
493 ret = blocking_notifier_call_chain(&inetaddr_validator_chain,
494 NETDEV_UP, &ivi);
495 ret = notifier_to_errno(ret);
496 if (ret) {
497 inet_free_ifa(ifa);
498 return ret;
499 }
500
501 if (!(ifa->ifa_flags & IFA_F_SECONDARY)) {
502 prandom_seed((__force u32) ifa->ifa_local);
503 ifap = last_primary;
504 }
505
506 ifa->ifa_next = *ifap;
507 *ifap = ifa;
508
509 inet_hash_insert(dev_net(in_dev->dev), ifa);
510
511 cancel_delayed_work(&check_lifetime_work);
512 queue_delayed_work(system_power_efficient_wq, &check_lifetime_work, 0);
513
514 /* Send message first, then call notifier.
515 Notifier will trigger FIB update, so that
516 listeners of netlink will know about new ifaddr */
517 rtmsg_ifa(RTM_NEWADDR, ifa, nlh, portid);
518 blocking_notifier_call_chain(&inetaddr_chain, NETDEV_UP, ifa);
519
520 return 0;
521}
522
523static int inet_insert_ifa(struct in_ifaddr *ifa)
524{
525 return __inet_insert_ifa(ifa, NULL, 0, NULL);
526}
527
528static int inet_set_ifa(struct net_device *dev, struct in_ifaddr *ifa)
529{
530 struct in_device *in_dev = __in_dev_get_rtnl(dev);
531
532 ASSERT_RTNL();
533
534 if (!in_dev) {
535 inet_free_ifa(ifa);
536 return -ENOBUFS;
537 }
538 ipv4_devconf_setall(in_dev);
539 neigh_parms_data_state_setall(in_dev->arp_parms);
540 if (ifa->ifa_dev != in_dev) {
541 WARN_ON(ifa->ifa_dev);
542 in_dev_hold(in_dev);
543 ifa->ifa_dev = in_dev;
544 }
545 if (ipv4_is_loopback(ifa->ifa_local))
546 ifa->ifa_scope = RT_SCOPE_HOST;
547 return inet_insert_ifa(ifa);
548}
549
550/* Caller must hold RCU or RTNL :
551 * We dont take a reference on found in_device
552 */
553struct in_device *inetdev_by_index(struct net *net, int ifindex)
554{
555 struct net_device *dev;
556 struct in_device *in_dev = NULL;
557
558 rcu_read_lock();
559 dev = dev_get_by_index_rcu(net, ifindex);
560 if (dev)
561 in_dev = rcu_dereference_rtnl(dev->ip_ptr);
562 rcu_read_unlock();
563 return in_dev;
564}
565EXPORT_SYMBOL(inetdev_by_index);
566
567/* Called only from RTNL semaphored context. No locks. */
568
569struct in_ifaddr *inet_ifa_byprefix(struct in_device *in_dev, __be32 prefix,
570 __be32 mask)
571{
572 ASSERT_RTNL();
573
574 for_primary_ifa(in_dev) {
575 if (ifa->ifa_mask == mask && inet_ifa_match(prefix, ifa))
576 return ifa;
577 } endfor_ifa(in_dev);
578 return NULL;
579}
580
581static int ip_mc_config(struct sock *sk, bool join, const struct in_ifaddr *ifa)
582{
583 struct ip_mreqn mreq = {
584 .imr_multiaddr.s_addr = ifa->ifa_address,
585 .imr_ifindex = ifa->ifa_dev->dev->ifindex,
586 };
587 int ret;
588
589 ASSERT_RTNL();
590
591 lock_sock(sk);
592 if (join)
593 ret = ip_mc_join_group(sk, &mreq);
594 else
595 ret = ip_mc_leave_group(sk, &mreq);
596 release_sock(sk);
597
598 return ret;
599}
600
601static int inet_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
602 struct netlink_ext_ack *extack)
603{
604 struct net *net = sock_net(skb->sk);
605 struct nlattr *tb[IFA_MAX+1];
606 struct in_device *in_dev;
607 struct ifaddrmsg *ifm;
608 struct in_ifaddr *ifa, **ifap;
609 int err = -EINVAL;
610
611 ASSERT_RTNL();
612
613 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv4_policy,
614 extack);
615 if (err < 0)
616 goto errout;
617
618 ifm = nlmsg_data(nlh);
619 in_dev = inetdev_by_index(net, ifm->ifa_index);
620 if (!in_dev) {
621 err = -ENODEV;
622 goto errout;
623 }
624
625 for (ifap = &in_dev->ifa_list; (ifa = *ifap) != NULL;
626 ifap = &ifa->ifa_next) {
627 if (tb[IFA_LOCAL] &&
628 ifa->ifa_local != nla_get_in_addr(tb[IFA_LOCAL]))
629 continue;
630
631 if (tb[IFA_LABEL] && nla_strcmp(tb[IFA_LABEL], ifa->ifa_label))
632 continue;
633
634 if (tb[IFA_ADDRESS] &&
635 (ifm->ifa_prefixlen != ifa->ifa_prefixlen ||
636 !inet_ifa_match(nla_get_in_addr(tb[IFA_ADDRESS]), ifa)))
637 continue;
638
639 if (ipv4_is_multicast(ifa->ifa_address))
640 ip_mc_config(net->ipv4.mc_autojoin_sk, false, ifa);
641 __inet_del_ifa(in_dev, ifap, 1, nlh, NETLINK_CB(skb).portid);
642 return 0;
643 }
644
645 err = -EADDRNOTAVAIL;
646errout:
647 return err;
648}
649
650#define INFINITY_LIFE_TIME 0xFFFFFFFF
651
652static void check_lifetime(struct work_struct *work)
653{
654 unsigned long now, next, next_sec, next_sched;
655 struct in_ifaddr *ifa;
656 struct hlist_node *n;
657 int i;
658
659 now = jiffies;
660 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
661
662 for (i = 0; i < IN4_ADDR_HSIZE; i++) {
663 bool change_needed = false;
664
665 rcu_read_lock();
666 hlist_for_each_entry_rcu(ifa, &inet_addr_lst[i], hash) {
667 unsigned long age;
668
669 if (ifa->ifa_flags & IFA_F_PERMANENT)
670 continue;
671
672 /* We try to batch several events at once. */
673 age = (now - ifa->ifa_tstamp +
674 ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
675
676 if (ifa->ifa_valid_lft != INFINITY_LIFE_TIME &&
677 age >= ifa->ifa_valid_lft) {
678 change_needed = true;
679 } else if (ifa->ifa_preferred_lft ==
680 INFINITY_LIFE_TIME) {
681 continue;
682 } else if (age >= ifa->ifa_preferred_lft) {
683 if (time_before(ifa->ifa_tstamp +
684 ifa->ifa_valid_lft * HZ, next))
685 next = ifa->ifa_tstamp +
686 ifa->ifa_valid_lft * HZ;
687
688 if (!(ifa->ifa_flags & IFA_F_DEPRECATED))
689 change_needed = true;
690 } else if (time_before(ifa->ifa_tstamp +
691 ifa->ifa_preferred_lft * HZ,
692 next)) {
693 next = ifa->ifa_tstamp +
694 ifa->ifa_preferred_lft * HZ;
695 }
696 }
697 rcu_read_unlock();
698 if (!change_needed)
699 continue;
700 rtnl_lock();
701 hlist_for_each_entry_safe(ifa, n, &inet_addr_lst[i], hash) {
702 unsigned long age;
703
704 if (ifa->ifa_flags & IFA_F_PERMANENT)
705 continue;
706
707 /* We try to batch several events at once. */
708 age = (now - ifa->ifa_tstamp +
709 ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
710
711 if (ifa->ifa_valid_lft != INFINITY_LIFE_TIME &&
712 age >= ifa->ifa_valid_lft) {
713 struct in_ifaddr **ifap;
714
715 for (ifap = &ifa->ifa_dev->ifa_list;
716 *ifap != NULL; ifap = &(*ifap)->ifa_next) {
717 if (*ifap == ifa) {
718 inet_del_ifa(ifa->ifa_dev,
719 ifap, 1);
720 break;
721 }
722 }
723 } else if (ifa->ifa_preferred_lft !=
724 INFINITY_LIFE_TIME &&
725 age >= ifa->ifa_preferred_lft &&
726 !(ifa->ifa_flags & IFA_F_DEPRECATED)) {
727 ifa->ifa_flags |= IFA_F_DEPRECATED;
728 rtmsg_ifa(RTM_NEWADDR, ifa, NULL, 0);
729 }
730 }
731 rtnl_unlock();
732 }
733
734 next_sec = round_jiffies_up(next);
735 next_sched = next;
736
737 /* If rounded timeout is accurate enough, accept it. */
738 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
739 next_sched = next_sec;
740
741 now = jiffies;
742 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
743 if (time_before(next_sched, now + ADDRCONF_TIMER_FUZZ_MAX))
744 next_sched = now + ADDRCONF_TIMER_FUZZ_MAX;
745
746 queue_delayed_work(system_power_efficient_wq, &check_lifetime_work,
747 next_sched - now);
748}
749
750static void set_ifa_lifetime(struct in_ifaddr *ifa, __u32 valid_lft,
751 __u32 prefered_lft)
752{
753 unsigned long timeout;
754
755 ifa->ifa_flags &= ~(IFA_F_PERMANENT | IFA_F_DEPRECATED);
756
757 timeout = addrconf_timeout_fixup(valid_lft, HZ);
758 if (addrconf_finite_timeout(timeout))
759 ifa->ifa_valid_lft = timeout;
760 else
761 ifa->ifa_flags |= IFA_F_PERMANENT;
762
763 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
764 if (addrconf_finite_timeout(timeout)) {
765 if (timeout == 0)
766 ifa->ifa_flags |= IFA_F_DEPRECATED;
767 ifa->ifa_preferred_lft = timeout;
768 }
769 ifa->ifa_tstamp = jiffies;
770 if (!ifa->ifa_cstamp)
771 ifa->ifa_cstamp = ifa->ifa_tstamp;
772}
773
774static struct in_ifaddr *rtm_to_ifaddr(struct net *net, struct nlmsghdr *nlh,
775 __u32 *pvalid_lft, __u32 *pprefered_lft)
776{
777 struct nlattr *tb[IFA_MAX+1];
778 struct in_ifaddr *ifa;
779 struct ifaddrmsg *ifm;
780 struct net_device *dev;
781 struct in_device *in_dev;
782 int err;
783
784 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv4_policy,
785 NULL);
786 if (err < 0)
787 goto errout;
788
789 ifm = nlmsg_data(nlh);
790 err = -EINVAL;
791 if (ifm->ifa_prefixlen > 32 || !tb[IFA_LOCAL])
792 goto errout;
793
794 dev = __dev_get_by_index(net, ifm->ifa_index);
795 err = -ENODEV;
796 if (!dev)
797 goto errout;
798
799 in_dev = __in_dev_get_rtnl(dev);
800 err = -ENOBUFS;
801 if (!in_dev)
802 goto errout;
803
804 ifa = inet_alloc_ifa();
805 if (!ifa)
806 /*
807 * A potential indev allocation can be left alive, it stays
808 * assigned to its device and is destroy with it.
809 */
810 goto errout;
811
812 ipv4_devconf_setall(in_dev);
813 neigh_parms_data_state_setall(in_dev->arp_parms);
814 in_dev_hold(in_dev);
815
816 if (!tb[IFA_ADDRESS])
817 tb[IFA_ADDRESS] = tb[IFA_LOCAL];
818
819 INIT_HLIST_NODE(&ifa->hash);
820 ifa->ifa_prefixlen = ifm->ifa_prefixlen;
821 ifa->ifa_mask = inet_make_mask(ifm->ifa_prefixlen);
822 ifa->ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) :
823 ifm->ifa_flags;
824 ifa->ifa_scope = ifm->ifa_scope;
825 ifa->ifa_dev = in_dev;
826
827 ifa->ifa_local = nla_get_in_addr(tb[IFA_LOCAL]);
828 ifa->ifa_address = nla_get_in_addr(tb[IFA_ADDRESS]);
829
830 if (tb[IFA_BROADCAST])
831 ifa->ifa_broadcast = nla_get_in_addr(tb[IFA_BROADCAST]);
832
833 if (tb[IFA_LABEL])
834 nla_strlcpy(ifa->ifa_label, tb[IFA_LABEL], IFNAMSIZ);
835 else
836 memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
837
838 if (tb[IFA_CACHEINFO]) {
839 struct ifa_cacheinfo *ci;
840
841 ci = nla_data(tb[IFA_CACHEINFO]);
842 if (!ci->ifa_valid || ci->ifa_prefered > ci->ifa_valid) {
843 err = -EINVAL;
844 goto errout_free;
845 }
846 *pvalid_lft = ci->ifa_valid;
847 *pprefered_lft = ci->ifa_prefered;
848 }
849
850 return ifa;
851
852errout_free:
853 inet_free_ifa(ifa);
854errout:
855 return ERR_PTR(err);
856}
857
858static struct in_ifaddr *find_matching_ifa(struct in_ifaddr *ifa)
859{
860 struct in_device *in_dev = ifa->ifa_dev;
861 struct in_ifaddr *ifa1, **ifap;
862
863 if (!ifa->ifa_local)
864 return NULL;
865
866 for (ifap = &in_dev->ifa_list; (ifa1 = *ifap) != NULL;
867 ifap = &ifa1->ifa_next) {
868 if (ifa1->ifa_mask == ifa->ifa_mask &&
869 inet_ifa_match(ifa1->ifa_address, ifa) &&
870 ifa1->ifa_local == ifa->ifa_local)
871 return ifa1;
872 }
873 return NULL;
874}
875
876static int inet_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
877 struct netlink_ext_ack *extack)
878{
879 struct net *net = sock_net(skb->sk);
880 struct in_ifaddr *ifa;
881 struct in_ifaddr *ifa_existing;
882 __u32 valid_lft = INFINITY_LIFE_TIME;
883 __u32 prefered_lft = INFINITY_LIFE_TIME;
884
885 ASSERT_RTNL();
886
887 ifa = rtm_to_ifaddr(net, nlh, &valid_lft, &prefered_lft);
888 if (IS_ERR(ifa))
889 return PTR_ERR(ifa);
890
891 ifa_existing = find_matching_ifa(ifa);
892 if (!ifa_existing) {
893 /* It would be best to check for !NLM_F_CREATE here but
894 * userspace already relies on not having to provide this.
895 */
896 set_ifa_lifetime(ifa, valid_lft, prefered_lft);
897 if (ifa->ifa_flags & IFA_F_MCAUTOJOIN) {
898 int ret = ip_mc_config(net->ipv4.mc_autojoin_sk,
899 true, ifa);
900
901 if (ret < 0) {
902 inet_free_ifa(ifa);
903 return ret;
904 }
905 }
906 return __inet_insert_ifa(ifa, nlh, NETLINK_CB(skb).portid,
907 extack);
908 } else {
909 inet_free_ifa(ifa);
910
911 if (nlh->nlmsg_flags & NLM_F_EXCL ||
912 !(nlh->nlmsg_flags & NLM_F_REPLACE))
913 return -EEXIST;
914 ifa = ifa_existing;
915 set_ifa_lifetime(ifa, valid_lft, prefered_lft);
916 cancel_delayed_work(&check_lifetime_work);
917 queue_delayed_work(system_power_efficient_wq,
918 &check_lifetime_work, 0);
919 rtmsg_ifa(RTM_NEWADDR, ifa, nlh, NETLINK_CB(skb).portid);
920 }
921 return 0;
922}
923
924/*
925 * Determine a default network mask, based on the IP address.
926 */
927
928static int inet_abc_len(__be32 addr)
929{
930 int rc = -1; /* Something else, probably a multicast. */
931
932 if (ipv4_is_zeronet(addr))
933 rc = 0;
934 else {
935 __u32 haddr = ntohl(addr);
936
937 if (IN_CLASSA(haddr))
938 rc = 8;
939 else if (IN_CLASSB(haddr))
940 rc = 16;
941 else if (IN_CLASSC(haddr))
942 rc = 24;
943 }
944
945 return rc;
946}
947
948
949int devinet_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr)
950{
951 struct sockaddr_in sin_orig;
952 struct sockaddr_in *sin = (struct sockaddr_in *)&ifr->ifr_addr;
953 struct in_device *in_dev;
954 struct in_ifaddr **ifap = NULL;
955 struct in_ifaddr *ifa = NULL;
956 struct net_device *dev;
957 char *colon;
958 int ret = -EFAULT;
959 int tryaddrmatch = 0;
960
961 ifr->ifr_name[IFNAMSIZ - 1] = 0;
962
963 /* save original address for comparison */
964 memcpy(&sin_orig, sin, sizeof(*sin));
965
966 colon = strchr(ifr->ifr_name, ':');
967 if (colon)
968 *colon = 0;
969
970 dev_load(net, ifr->ifr_name);
971
972 switch (cmd) {
973 case SIOCGIFADDR: /* Get interface address */
974 case SIOCGIFBRDADDR: /* Get the broadcast address */
975 case SIOCGIFDSTADDR: /* Get the destination address */
976 case SIOCGIFNETMASK: /* Get the netmask for the interface */
977 /* Note that these ioctls will not sleep,
978 so that we do not impose a lock.
979 One day we will be forced to put shlock here (I mean SMP)
980 */
981 tryaddrmatch = (sin_orig.sin_family == AF_INET);
982 memset(sin, 0, sizeof(*sin));
983 sin->sin_family = AF_INET;
984 break;
985
986 case SIOCSIFFLAGS:
987 ret = -EPERM;
988 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
989 goto out;
990 break;
991 case SIOCSIFADDR: /* Set interface address (and family) */
992 case SIOCSIFBRDADDR: /* Set the broadcast address */
993 case SIOCSIFDSTADDR: /* Set the destination address */
994 case SIOCSIFNETMASK: /* Set the netmask for the interface */
995 ret = -EPERM;
996 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
997 goto out;
998 ret = -EINVAL;
999 if (sin->sin_family != AF_INET)
1000 goto out;
1001 break;
1002 default:
1003 ret = -EINVAL;
1004 goto out;
1005 }
1006
1007 rtnl_lock();
1008
1009 ret = -ENODEV;
1010 dev = __dev_get_by_name(net, ifr->ifr_name);
1011 if (!dev)
1012 goto done;
1013
1014 if (colon)
1015 *colon = ':';
1016
1017 in_dev = __in_dev_get_rtnl(dev);
1018 if (in_dev) {
1019 if (tryaddrmatch) {
1020 /* Matthias Andree */
1021 /* compare label and address (4.4BSD style) */
1022 /* note: we only do this for a limited set of ioctls
1023 and only if the original address family was AF_INET.
1024 This is checked above. */
1025 for (ifap = &in_dev->ifa_list; (ifa = *ifap) != NULL;
1026 ifap = &ifa->ifa_next) {
1027 if (!strcmp(ifr->ifr_name, ifa->ifa_label) &&
1028 sin_orig.sin_addr.s_addr ==
1029 ifa->ifa_local) {
1030 break; /* found */
1031 }
1032 }
1033 }
1034 /* we didn't get a match, maybe the application is
1035 4.3BSD-style and passed in junk so we fall back to
1036 comparing just the label */
1037 if (!ifa) {
1038 for (ifap = &in_dev->ifa_list; (ifa = *ifap) != NULL;
1039 ifap = &ifa->ifa_next)
1040 if (!strcmp(ifr->ifr_name, ifa->ifa_label))
1041 break;
1042 }
1043 }
1044
1045 ret = -EADDRNOTAVAIL;
1046 if (!ifa && cmd != SIOCSIFADDR && cmd != SIOCSIFFLAGS)
1047 goto done;
1048
1049 switch (cmd) {
1050 case SIOCGIFADDR: /* Get interface address */
1051 ret = 0;
1052 sin->sin_addr.s_addr = ifa->ifa_local;
1053 break;
1054
1055 case SIOCGIFBRDADDR: /* Get the broadcast address */
1056 ret = 0;
1057 sin->sin_addr.s_addr = ifa->ifa_broadcast;
1058 break;
1059
1060 case SIOCGIFDSTADDR: /* Get the destination address */
1061 ret = 0;
1062 sin->sin_addr.s_addr = ifa->ifa_address;
1063 break;
1064
1065 case SIOCGIFNETMASK: /* Get the netmask for the interface */
1066 ret = 0;
1067 sin->sin_addr.s_addr = ifa->ifa_mask;
1068 break;
1069
1070 case SIOCSIFFLAGS:
1071 if (colon) {
1072 ret = -EADDRNOTAVAIL;
1073 if (!ifa)
1074 break;
1075 ret = 0;
1076 if (!(ifr->ifr_flags & IFF_UP))
1077 inet_del_ifa(in_dev, ifap, 1);
1078 break;
1079 }
1080 ret = dev_change_flags(dev, ifr->ifr_flags);
1081 break;
1082
1083 case SIOCSIFADDR: /* Set interface address (and family) */
1084 ret = -EINVAL;
1085 if (inet_abc_len(sin->sin_addr.s_addr) < 0)
1086 break;
1087
1088 if (!ifa) {
1089 ret = -ENOBUFS;
1090 ifa = inet_alloc_ifa();
1091 if (!ifa)
1092 break;
1093 INIT_HLIST_NODE(&ifa->hash);
1094 if (colon)
1095 memcpy(ifa->ifa_label, ifr->ifr_name, IFNAMSIZ);
1096 else
1097 memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
1098 } else {
1099 ret = 0;
1100 if (ifa->ifa_local == sin->sin_addr.s_addr)
1101 break;
1102 inet_del_ifa(in_dev, ifap, 0);
1103 ifa->ifa_broadcast = 0;
1104 ifa->ifa_scope = 0;
1105 }
1106
1107 ifa->ifa_address = ifa->ifa_local = sin->sin_addr.s_addr;
1108
1109 if (!(dev->flags & IFF_POINTOPOINT)) {
1110 ifa->ifa_prefixlen = inet_abc_len(ifa->ifa_address);
1111 ifa->ifa_mask = inet_make_mask(ifa->ifa_prefixlen);
1112 if ((dev->flags & IFF_BROADCAST) &&
1113 ifa->ifa_prefixlen < 31)
1114 ifa->ifa_broadcast = ifa->ifa_address |
1115 ~ifa->ifa_mask;
1116 } else {
1117 ifa->ifa_prefixlen = 32;
1118 ifa->ifa_mask = inet_make_mask(32);
1119 }
1120 set_ifa_lifetime(ifa, INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
1121 ret = inet_set_ifa(dev, ifa);
1122 break;
1123
1124 case SIOCSIFBRDADDR: /* Set the broadcast address */
1125 ret = 0;
1126 if (ifa->ifa_broadcast != sin->sin_addr.s_addr) {
1127 inet_del_ifa(in_dev, ifap, 0);
1128 ifa->ifa_broadcast = sin->sin_addr.s_addr;
1129 inet_insert_ifa(ifa);
1130 }
1131 break;
1132
1133 case SIOCSIFDSTADDR: /* Set the destination address */
1134 ret = 0;
1135 if (ifa->ifa_address == sin->sin_addr.s_addr)
1136 break;
1137 ret = -EINVAL;
1138 if (inet_abc_len(sin->sin_addr.s_addr) < 0)
1139 break;
1140 ret = 0;
1141 inet_del_ifa(in_dev, ifap, 0);
1142 ifa->ifa_address = sin->sin_addr.s_addr;
1143 inet_insert_ifa(ifa);
1144 break;
1145
1146 case SIOCSIFNETMASK: /* Set the netmask for the interface */
1147
1148 /*
1149 * The mask we set must be legal.
1150 */
1151 ret = -EINVAL;
1152 if (bad_mask(sin->sin_addr.s_addr, 0))
1153 break;
1154 ret = 0;
1155 if (ifa->ifa_mask != sin->sin_addr.s_addr) {
1156 __be32 old_mask = ifa->ifa_mask;
1157 inet_del_ifa(in_dev, ifap, 0);
1158 ifa->ifa_mask = sin->sin_addr.s_addr;
1159 ifa->ifa_prefixlen = inet_mask_len(ifa->ifa_mask);
1160
1161 /* See if current broadcast address matches
1162 * with current netmask, then recalculate
1163 * the broadcast address. Otherwise it's a
1164 * funny address, so don't touch it since
1165 * the user seems to know what (s)he's doing...
1166 */
1167 if ((dev->flags & IFF_BROADCAST) &&
1168 (ifa->ifa_prefixlen < 31) &&
1169 (ifa->ifa_broadcast ==
1170 (ifa->ifa_local|~old_mask))) {
1171 ifa->ifa_broadcast = (ifa->ifa_local |
1172 ~sin->sin_addr.s_addr);
1173 }
1174 inet_insert_ifa(ifa);
1175 }
1176 break;
1177 }
1178done:
1179 rtnl_unlock();
1180out:
1181 return ret;
1182}
1183
1184static int inet_gifconf(struct net_device *dev, char __user *buf, int len, int size)
1185{
1186 struct in_device *in_dev = __in_dev_get_rtnl(dev);
1187 struct in_ifaddr *ifa;
1188 struct ifreq ifr;
1189 int done = 0;
1190
1191 if (WARN_ON(size > sizeof(struct ifreq)))
1192 goto out;
1193
1194 if (!in_dev)
1195 goto out;
1196
1197 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
1198 if (!buf) {
1199 done += size;
1200 continue;
1201 }
1202 if (len < size)
1203 break;
1204 memset(&ifr, 0, sizeof(struct ifreq));
1205 strcpy(ifr.ifr_name, ifa->ifa_label);
1206
1207 (*(struct sockaddr_in *)&ifr.ifr_addr).sin_family = AF_INET;
1208 (*(struct sockaddr_in *)&ifr.ifr_addr).sin_addr.s_addr =
1209 ifa->ifa_local;
1210
1211 if (copy_to_user(buf + done, &ifr, size)) {
1212 done = -EFAULT;
1213 break;
1214 }
1215 len -= size;
1216 done += size;
1217 }
1218out:
1219 return done;
1220}
1221
1222static __be32 in_dev_select_addr(const struct in_device *in_dev,
1223 int scope)
1224{
1225 for_primary_ifa(in_dev) {
1226 if (ifa->ifa_scope != RT_SCOPE_LINK &&
1227 ifa->ifa_scope <= scope)
1228 return ifa->ifa_local;
1229 } endfor_ifa(in_dev);
1230
1231 return 0;
1232}
1233
1234__be32 inet_select_addr(const struct net_device *dev, __be32 dst, int scope)
1235{
1236 __be32 addr = 0;
1237 struct in_device *in_dev;
1238 struct net *net = dev_net(dev);
1239 int master_idx;
1240
1241 rcu_read_lock();
1242 in_dev = __in_dev_get_rcu(dev);
1243 if (!in_dev)
1244 goto no_in_dev;
1245
1246 for_primary_ifa(in_dev) {
1247 if (ifa->ifa_scope > scope)
1248 continue;
1249 if (!dst || inet_ifa_match(dst, ifa)) {
1250 addr = ifa->ifa_local;
1251 break;
1252 }
1253 if (!addr)
1254 addr = ifa->ifa_local;
1255 } endfor_ifa(in_dev);
1256
1257 if (addr)
1258 goto out_unlock;
1259no_in_dev:
1260 master_idx = l3mdev_master_ifindex_rcu(dev);
1261
1262 /* For VRFs, the VRF device takes the place of the loopback device,
1263 * with addresses on it being preferred. Note in such cases the
1264 * loopback device will be among the devices that fail the master_idx
1265 * equality check in the loop below.
1266 */
1267 if (master_idx &&
1268 (dev = dev_get_by_index_rcu(net, master_idx)) &&
1269 (in_dev = __in_dev_get_rcu(dev))) {
1270 addr = in_dev_select_addr(in_dev, scope);
1271 if (addr)
1272 goto out_unlock;
1273 }
1274
1275 /* Not loopback addresses on loopback should be preferred
1276 in this case. It is important that lo is the first interface
1277 in dev_base list.
1278 */
1279 for_each_netdev_rcu(net, dev) {
1280 if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1281 continue;
1282
1283 in_dev = __in_dev_get_rcu(dev);
1284 if (!in_dev)
1285 continue;
1286
1287 addr = in_dev_select_addr(in_dev, scope);
1288 if (addr)
1289 goto out_unlock;
1290 }
1291out_unlock:
1292 rcu_read_unlock();
1293 return addr;
1294}
1295EXPORT_SYMBOL(inet_select_addr);
1296
1297static __be32 confirm_addr_indev(struct in_device *in_dev, __be32 dst,
1298 __be32 local, int scope)
1299{
1300 int same = 0;
1301 __be32 addr = 0;
1302
1303 for_ifa(in_dev) {
1304 if (!addr &&
1305 (local == ifa->ifa_local || !local) &&
1306 ifa->ifa_scope <= scope) {
1307 addr = ifa->ifa_local;
1308 if (same)
1309 break;
1310 }
1311 if (!same) {
1312 same = (!local || inet_ifa_match(local, ifa)) &&
1313 (!dst || inet_ifa_match(dst, ifa));
1314 if (same && addr) {
1315 if (local || !dst)
1316 break;
1317 /* Is the selected addr into dst subnet? */
1318 if (inet_ifa_match(addr, ifa))
1319 break;
1320 /* No, then can we use new local src? */
1321 if (ifa->ifa_scope <= scope) {
1322 addr = ifa->ifa_local;
1323 break;
1324 }
1325 /* search for large dst subnet for addr */
1326 same = 0;
1327 }
1328 }
1329 } endfor_ifa(in_dev);
1330
1331 return same ? addr : 0;
1332}
1333
1334/*
1335 * Confirm that local IP address exists using wildcards:
1336 * - net: netns to check, cannot be NULL
1337 * - in_dev: only on this interface, NULL=any interface
1338 * - dst: only in the same subnet as dst, 0=any dst
1339 * - local: address, 0=autoselect the local address
1340 * - scope: maximum allowed scope value for the local address
1341 */
1342__be32 inet_confirm_addr(struct net *net, struct in_device *in_dev,
1343 __be32 dst, __be32 local, int scope)
1344{
1345 __be32 addr = 0;
1346 struct net_device *dev;
1347
1348 if (in_dev)
1349 return confirm_addr_indev(in_dev, dst, local, scope);
1350
1351 rcu_read_lock();
1352 for_each_netdev_rcu(net, dev) {
1353 in_dev = __in_dev_get_rcu(dev);
1354 if (in_dev) {
1355 addr = confirm_addr_indev(in_dev, dst, local, scope);
1356 if (addr)
1357 break;
1358 }
1359 }
1360 rcu_read_unlock();
1361
1362 return addr;
1363}
1364EXPORT_SYMBOL(inet_confirm_addr);
1365
1366/*
1367 * Device notifier
1368 */
1369
1370int register_inetaddr_notifier(struct notifier_block *nb)
1371{
1372 return blocking_notifier_chain_register(&inetaddr_chain, nb);
1373}
1374EXPORT_SYMBOL(register_inetaddr_notifier);
1375
1376int unregister_inetaddr_notifier(struct notifier_block *nb)
1377{
1378 return blocking_notifier_chain_unregister(&inetaddr_chain, nb);
1379}
1380EXPORT_SYMBOL(unregister_inetaddr_notifier);
1381
1382int register_inetaddr_validator_notifier(struct notifier_block *nb)
1383{
1384 return blocking_notifier_chain_register(&inetaddr_validator_chain, nb);
1385}
1386EXPORT_SYMBOL(register_inetaddr_validator_notifier);
1387
1388int unregister_inetaddr_validator_notifier(struct notifier_block *nb)
1389{
1390 return blocking_notifier_chain_unregister(&inetaddr_validator_chain,
1391 nb);
1392}
1393EXPORT_SYMBOL(unregister_inetaddr_validator_notifier);
1394
1395/* Rename ifa_labels for a device name change. Make some effort to preserve
1396 * existing alias numbering and to create unique labels if possible.
1397*/
1398static void inetdev_changename(struct net_device *dev, struct in_device *in_dev)
1399{
1400 struct in_ifaddr *ifa;
1401 int named = 0;
1402
1403 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
1404 char old[IFNAMSIZ], *dot;
1405
1406 memcpy(old, ifa->ifa_label, IFNAMSIZ);
1407 memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
1408 if (named++ == 0)
1409 goto skip;
1410 dot = strchr(old, ':');
1411 if (!dot) {
1412 sprintf(old, ":%d", named);
1413 dot = old;
1414 }
1415 if (strlen(dot) + strlen(dev->name) < IFNAMSIZ)
1416 strcat(ifa->ifa_label, dot);
1417 else
1418 strcpy(ifa->ifa_label + (IFNAMSIZ - strlen(dot) - 1), dot);
1419skip:
1420 rtmsg_ifa(RTM_NEWADDR, ifa, NULL, 0);
1421 }
1422}
1423
1424static bool inetdev_valid_mtu(unsigned int mtu)
1425{
1426 return mtu >= IPV4_MIN_MTU;
1427}
1428
1429static void inetdev_send_gratuitous_arp(struct net_device *dev,
1430 struct in_device *in_dev)
1431
1432{
1433 struct in_ifaddr *ifa;
1434
1435 for (ifa = in_dev->ifa_list; ifa;
1436 ifa = ifa->ifa_next) {
1437 arp_send(ARPOP_REQUEST, ETH_P_ARP,
1438 ifa->ifa_local, dev,
1439 ifa->ifa_local, NULL,
1440 dev->dev_addr, NULL);
1441 }
1442}
1443
1444/* Called only under RTNL semaphore */
1445
1446static int inetdev_event(struct notifier_block *this, unsigned long event,
1447 void *ptr)
1448{
1449 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1450 struct in_device *in_dev = __in_dev_get_rtnl(dev);
1451
1452 ASSERT_RTNL();
1453
1454 if (!in_dev) {
1455 if (event == NETDEV_REGISTER) {
1456 in_dev = inetdev_init(dev);
1457 if (IS_ERR(in_dev))
1458 return notifier_from_errno(PTR_ERR(in_dev));
1459 if (dev->flags & IFF_LOOPBACK) {
1460 IN_DEV_CONF_SET(in_dev, NOXFRM, 1);
1461 IN_DEV_CONF_SET(in_dev, NOPOLICY, 1);
1462 }
1463 } else if (event == NETDEV_CHANGEMTU) {
1464 /* Re-enabling IP */
1465 if (inetdev_valid_mtu(dev->mtu))
1466 in_dev = inetdev_init(dev);
1467 }
1468 goto out;
1469 }
1470
1471 switch (event) {
1472 case NETDEV_REGISTER:
1473 pr_debug("%s: bug\n", __func__);
1474 RCU_INIT_POINTER(dev->ip_ptr, NULL);
1475 break;
1476 case NETDEV_UP:
1477 if (!inetdev_valid_mtu(dev->mtu))
1478 break;
1479 if (dev->flags & IFF_LOOPBACK) {
1480 struct in_ifaddr *ifa = inet_alloc_ifa();
1481
1482 if (ifa) {
1483 INIT_HLIST_NODE(&ifa->hash);
1484 ifa->ifa_local =
1485 ifa->ifa_address = htonl(INADDR_LOOPBACK);
1486 ifa->ifa_prefixlen = 8;
1487 ifa->ifa_mask = inet_make_mask(8);
1488 in_dev_hold(in_dev);
1489 ifa->ifa_dev = in_dev;
1490 ifa->ifa_scope = RT_SCOPE_HOST;
1491 memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
1492 set_ifa_lifetime(ifa, INFINITY_LIFE_TIME,
1493 INFINITY_LIFE_TIME);
1494 ipv4_devconf_setall(in_dev);
1495 neigh_parms_data_state_setall(in_dev->arp_parms);
1496 inet_insert_ifa(ifa);
1497 }
1498 }
1499 ip_mc_up(in_dev);
1500 /* fall through */
1501 case NETDEV_CHANGEADDR:
1502 if (!IN_DEV_ARP_NOTIFY(in_dev))
1503 break;
1504 /* fall through */
1505 case NETDEV_NOTIFY_PEERS:
1506 /* Send gratuitous ARP to notify of link change */
1507 inetdev_send_gratuitous_arp(dev, in_dev);
1508 break;
1509 case NETDEV_DOWN:
1510 ip_mc_down(in_dev);
1511 break;
1512 case NETDEV_PRE_TYPE_CHANGE:
1513 ip_mc_unmap(in_dev);
1514 break;
1515 case NETDEV_POST_TYPE_CHANGE:
1516 ip_mc_remap(in_dev);
1517 break;
1518 case NETDEV_CHANGEMTU:
1519 if (inetdev_valid_mtu(dev->mtu))
1520 break;
1521 /* disable IP when MTU is not enough */
1522 /* fall through */
1523 case NETDEV_UNREGISTER:
1524 inetdev_destroy(in_dev);
1525 break;
1526 case NETDEV_CHANGENAME:
1527 /* Do not notify about label change, this event is
1528 * not interesting to applications using netlink.
1529 */
1530 inetdev_changename(dev, in_dev);
1531
1532 devinet_sysctl_unregister(in_dev);
1533 devinet_sysctl_register(in_dev);
1534 break;
1535 }
1536out:
1537 return NOTIFY_DONE;
1538}
1539
1540static struct notifier_block ip_netdev_notifier = {
1541 .notifier_call = inetdev_event,
1542};
1543
1544static size_t inet_nlmsg_size(void)
1545{
1546 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
1547 + nla_total_size(4) /* IFA_ADDRESS */
1548 + nla_total_size(4) /* IFA_LOCAL */
1549 + nla_total_size(4) /* IFA_BROADCAST */
1550 + nla_total_size(IFNAMSIZ) /* IFA_LABEL */
1551 + nla_total_size(4) /* IFA_FLAGS */
1552 + nla_total_size(sizeof(struct ifa_cacheinfo)); /* IFA_CACHEINFO */
1553}
1554
1555static inline u32 cstamp_delta(unsigned long cstamp)
1556{
1557 return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
1558}
1559
1560static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
1561 unsigned long tstamp, u32 preferred, u32 valid)
1562{
1563 struct ifa_cacheinfo ci;
1564
1565 ci.cstamp = cstamp_delta(cstamp);
1566 ci.tstamp = cstamp_delta(tstamp);
1567 ci.ifa_prefered = preferred;
1568 ci.ifa_valid = valid;
1569
1570 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
1571}
1572
1573static int inet_fill_ifaddr(struct sk_buff *skb, struct in_ifaddr *ifa,
1574 u32 portid, u32 seq, int event, unsigned int flags)
1575{
1576 struct ifaddrmsg *ifm;
1577 struct nlmsghdr *nlh;
1578 u32 preferred, valid;
1579
1580 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*ifm), flags);
1581 if (!nlh)
1582 return -EMSGSIZE;
1583
1584 ifm = nlmsg_data(nlh);
1585 ifm->ifa_family = AF_INET;
1586 ifm->ifa_prefixlen = ifa->ifa_prefixlen;
1587 ifm->ifa_flags = ifa->ifa_flags;
1588 ifm->ifa_scope = ifa->ifa_scope;
1589 ifm->ifa_index = ifa->ifa_dev->dev->ifindex;
1590
1591 if (!(ifm->ifa_flags & IFA_F_PERMANENT)) {
1592 preferred = ifa->ifa_preferred_lft;
1593 valid = ifa->ifa_valid_lft;
1594 if (preferred != INFINITY_LIFE_TIME) {
1595 long tval = (jiffies - ifa->ifa_tstamp) / HZ;
1596
1597 if (preferred > tval)
1598 preferred -= tval;
1599 else
1600 preferred = 0;
1601 if (valid != INFINITY_LIFE_TIME) {
1602 if (valid > tval)
1603 valid -= tval;
1604 else
1605 valid = 0;
1606 }
1607 }
1608 } else {
1609 preferred = INFINITY_LIFE_TIME;
1610 valid = INFINITY_LIFE_TIME;
1611 }
1612 if ((ifa->ifa_address &&
1613 nla_put_in_addr(skb, IFA_ADDRESS, ifa->ifa_address)) ||
1614 (ifa->ifa_local &&
1615 nla_put_in_addr(skb, IFA_LOCAL, ifa->ifa_local)) ||
1616 (ifa->ifa_broadcast &&
1617 nla_put_in_addr(skb, IFA_BROADCAST, ifa->ifa_broadcast)) ||
1618 (ifa->ifa_label[0] &&
1619 nla_put_string(skb, IFA_LABEL, ifa->ifa_label)) ||
1620 nla_put_u32(skb, IFA_FLAGS, ifa->ifa_flags) ||
1621 put_cacheinfo(skb, ifa->ifa_cstamp, ifa->ifa_tstamp,
1622 preferred, valid))
1623 goto nla_put_failure;
1624
1625 nlmsg_end(skb, nlh);
1626 return 0;
1627
1628nla_put_failure:
1629 nlmsg_cancel(skb, nlh);
1630 return -EMSGSIZE;
1631}
1632
1633static int inet_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
1634{
1635 struct net *net = sock_net(skb->sk);
1636 int h, s_h;
1637 int idx, s_idx;
1638 int ip_idx, s_ip_idx;
1639 struct net_device *dev;
1640 struct in_device *in_dev;
1641 struct in_ifaddr *ifa;
1642 struct hlist_head *head;
1643
1644 s_h = cb->args[0];
1645 s_idx = idx = cb->args[1];
1646 s_ip_idx = ip_idx = cb->args[2];
1647
1648 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1649 idx = 0;
1650 head = &net->dev_index_head[h];
1651 rcu_read_lock();
1652 cb->seq = atomic_read(&net->ipv4.dev_addr_genid) ^
1653 net->dev_base_seq;
1654 hlist_for_each_entry_rcu(dev, head, index_hlist) {
1655 if (idx < s_idx)
1656 goto cont;
1657 if (h > s_h || idx > s_idx)
1658 s_ip_idx = 0;
1659 in_dev = __in_dev_get_rcu(dev);
1660 if (!in_dev)
1661 goto cont;
1662
1663 for (ifa = in_dev->ifa_list, ip_idx = 0; ifa;
1664 ifa = ifa->ifa_next, ip_idx++) {
1665 if (ip_idx < s_ip_idx)
1666 continue;
1667 if (inet_fill_ifaddr(skb, ifa,
1668 NETLINK_CB(cb->skb).portid,
1669 cb->nlh->nlmsg_seq,
1670 RTM_NEWADDR, NLM_F_MULTI) < 0) {
1671 rcu_read_unlock();
1672 goto done;
1673 }
1674 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1675 }
1676cont:
1677 idx++;
1678 }
1679 rcu_read_unlock();
1680 }
1681
1682done:
1683 cb->args[0] = h;
1684 cb->args[1] = idx;
1685 cb->args[2] = ip_idx;
1686
1687 return skb->len;
1688}
1689
1690static void rtmsg_ifa(int event, struct in_ifaddr *ifa, struct nlmsghdr *nlh,
1691 u32 portid)
1692{
1693 struct sk_buff *skb;
1694 u32 seq = nlh ? nlh->nlmsg_seq : 0;
1695 int err = -ENOBUFS;
1696 struct net *net;
1697
1698 net = dev_net(ifa->ifa_dev->dev);
1699 skb = nlmsg_new(inet_nlmsg_size(), GFP_KERNEL);
1700 if (!skb)
1701 goto errout;
1702
1703 err = inet_fill_ifaddr(skb, ifa, portid, seq, event, 0);
1704 if (err < 0) {
1705 /* -EMSGSIZE implies BUG in inet_nlmsg_size() */
1706 WARN_ON(err == -EMSGSIZE);
1707 kfree_skb(skb);
1708 goto errout;
1709 }
1710 rtnl_notify(skb, net, portid, RTNLGRP_IPV4_IFADDR, nlh, GFP_KERNEL);
1711 return;
1712errout:
1713 if (err < 0)
1714 rtnl_set_sk_err(net, RTNLGRP_IPV4_IFADDR, err);
1715}
1716
1717static size_t inet_get_link_af_size(const struct net_device *dev,
1718 u32 ext_filter_mask)
1719{
1720 struct in_device *in_dev = rcu_dereference_rtnl(dev->ip_ptr);
1721
1722 if (!in_dev)
1723 return 0;
1724
1725 return nla_total_size(IPV4_DEVCONF_MAX * 4); /* IFLA_INET_CONF */
1726}
1727
1728static int inet_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
1729 u32 ext_filter_mask)
1730{
1731 struct in_device *in_dev = rcu_dereference_rtnl(dev->ip_ptr);
1732 struct nlattr *nla;
1733 int i;
1734
1735 if (!in_dev)
1736 return -ENODATA;
1737
1738 nla = nla_reserve(skb, IFLA_INET_CONF, IPV4_DEVCONF_MAX * 4);
1739 if (!nla)
1740 return -EMSGSIZE;
1741
1742 for (i = 0; i < IPV4_DEVCONF_MAX; i++)
1743 ((u32 *) nla_data(nla))[i] = in_dev->cnf.data[i];
1744
1745 return 0;
1746}
1747
1748static const struct nla_policy inet_af_policy[IFLA_INET_MAX+1] = {
1749 [IFLA_INET_CONF] = { .type = NLA_NESTED },
1750};
1751
1752static int inet_validate_link_af(const struct net_device *dev,
1753 const struct nlattr *nla)
1754{
1755 struct nlattr *a, *tb[IFLA_INET_MAX+1];
1756 int err, rem;
1757
1758 if (dev && !__in_dev_get_rcu(dev))
1759 return -EAFNOSUPPORT;
1760
1761 err = nla_parse_nested(tb, IFLA_INET_MAX, nla, inet_af_policy, NULL);
1762 if (err < 0)
1763 return err;
1764
1765 if (tb[IFLA_INET_CONF]) {
1766 nla_for_each_nested(a, tb[IFLA_INET_CONF], rem) {
1767 int cfgid = nla_type(a);
1768
1769 if (nla_len(a) < 4)
1770 return -EINVAL;
1771
1772 if (cfgid <= 0 || cfgid > IPV4_DEVCONF_MAX)
1773 return -EINVAL;
1774 }
1775 }
1776
1777 return 0;
1778}
1779
1780static int inet_set_link_af(struct net_device *dev, const struct nlattr *nla)
1781{
1782 struct in_device *in_dev = __in_dev_get_rcu(dev);
1783 struct nlattr *a, *tb[IFLA_INET_MAX+1];
1784 int rem;
1785
1786 if (!in_dev)
1787 return -EAFNOSUPPORT;
1788
1789 if (nla_parse_nested(tb, IFLA_INET_MAX, nla, NULL, NULL) < 0)
1790 BUG();
1791
1792 if (tb[IFLA_INET_CONF]) {
1793 nla_for_each_nested(a, tb[IFLA_INET_CONF], rem)
1794 ipv4_devconf_set(in_dev, nla_type(a), nla_get_u32(a));
1795 }
1796
1797 return 0;
1798}
1799
1800static int inet_netconf_msgsize_devconf(int type)
1801{
1802 int size = NLMSG_ALIGN(sizeof(struct netconfmsg))
1803 + nla_total_size(4); /* NETCONFA_IFINDEX */
1804 bool all = false;
1805
1806 if (type == NETCONFA_ALL)
1807 all = true;
1808
1809 if (all || type == NETCONFA_FORWARDING)
1810 size += nla_total_size(4);
1811 if (all || type == NETCONFA_RP_FILTER)
1812 size += nla_total_size(4);
1813 if (all || type == NETCONFA_MC_FORWARDING)
1814 size += nla_total_size(4);
1815 if (all || type == NETCONFA_PROXY_NEIGH)
1816 size += nla_total_size(4);
1817 if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
1818 size += nla_total_size(4);
1819
1820 return size;
1821}
1822
1823static int inet_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
1824 struct ipv4_devconf *devconf, u32 portid,
1825 u32 seq, int event, unsigned int flags,
1826 int type)
1827{
1828 struct nlmsghdr *nlh;
1829 struct netconfmsg *ncm;
1830 bool all = false;
1831
1832 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
1833 flags);
1834 if (!nlh)
1835 return -EMSGSIZE;
1836
1837 if (type == NETCONFA_ALL)
1838 all = true;
1839
1840 ncm = nlmsg_data(nlh);
1841 ncm->ncm_family = AF_INET;
1842
1843 if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
1844 goto nla_put_failure;
1845
1846 if (!devconf)
1847 goto out;
1848
1849 if ((all || type == NETCONFA_FORWARDING) &&
1850 nla_put_s32(skb, NETCONFA_FORWARDING,
1851 IPV4_DEVCONF(*devconf, FORWARDING)) < 0)
1852 goto nla_put_failure;
1853 if ((all || type == NETCONFA_RP_FILTER) &&
1854 nla_put_s32(skb, NETCONFA_RP_FILTER,
1855 IPV4_DEVCONF(*devconf, RP_FILTER)) < 0)
1856 goto nla_put_failure;
1857 if ((all || type == NETCONFA_MC_FORWARDING) &&
1858 nla_put_s32(skb, NETCONFA_MC_FORWARDING,
1859 IPV4_DEVCONF(*devconf, MC_FORWARDING)) < 0)
1860 goto nla_put_failure;
1861 if ((all || type == NETCONFA_PROXY_NEIGH) &&
1862 nla_put_s32(skb, NETCONFA_PROXY_NEIGH,
1863 IPV4_DEVCONF(*devconf, PROXY_ARP)) < 0)
1864 goto nla_put_failure;
1865 if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
1866 nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
1867 IPV4_DEVCONF(*devconf, IGNORE_ROUTES_WITH_LINKDOWN)) < 0)
1868 goto nla_put_failure;
1869
1870out:
1871 nlmsg_end(skb, nlh);
1872 return 0;
1873
1874nla_put_failure:
1875 nlmsg_cancel(skb, nlh);
1876 return -EMSGSIZE;
1877}
1878
1879void inet_netconf_notify_devconf(struct net *net, int event, int type,
1880 int ifindex, struct ipv4_devconf *devconf)
1881{
1882 struct sk_buff *skb;
1883 int err = -ENOBUFS;
1884
1885 skb = nlmsg_new(inet_netconf_msgsize_devconf(type), GFP_KERNEL);
1886 if (!skb)
1887 goto errout;
1888
1889 err = inet_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
1890 event, 0, type);
1891 if (err < 0) {
1892 /* -EMSGSIZE implies BUG in inet_netconf_msgsize_devconf() */
1893 WARN_ON(err == -EMSGSIZE);
1894 kfree_skb(skb);
1895 goto errout;
1896 }
1897 rtnl_notify(skb, net, 0, RTNLGRP_IPV4_NETCONF, NULL, GFP_KERNEL);
1898 return;
1899errout:
1900 if (err < 0)
1901 rtnl_set_sk_err(net, RTNLGRP_IPV4_NETCONF, err);
1902}
1903
1904static const struct nla_policy devconf_ipv4_policy[NETCONFA_MAX+1] = {
1905 [NETCONFA_IFINDEX] = { .len = sizeof(int) },
1906 [NETCONFA_FORWARDING] = { .len = sizeof(int) },
1907 [NETCONFA_RP_FILTER] = { .len = sizeof(int) },
1908 [NETCONFA_PROXY_NEIGH] = { .len = sizeof(int) },
1909 [NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN] = { .len = sizeof(int) },
1910};
1911
1912static int inet_netconf_get_devconf(struct sk_buff *in_skb,
1913 struct nlmsghdr *nlh,
1914 struct netlink_ext_ack *extack)
1915{
1916 struct net *net = sock_net(in_skb->sk);
1917 struct nlattr *tb[NETCONFA_MAX+1];
1918 struct netconfmsg *ncm;
1919 struct sk_buff *skb;
1920 struct ipv4_devconf *devconf;
1921 struct in_device *in_dev;
1922 struct net_device *dev;
1923 int ifindex;
1924 int err;
1925
1926 err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
1927 devconf_ipv4_policy, extack);
1928 if (err < 0)
1929 goto errout;
1930
1931 err = -EINVAL;
1932 if (!tb[NETCONFA_IFINDEX])
1933 goto errout;
1934
1935 ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
1936 switch (ifindex) {
1937 case NETCONFA_IFINDEX_ALL:
1938 devconf = net->ipv4.devconf_all;
1939 break;
1940 case NETCONFA_IFINDEX_DEFAULT:
1941 devconf = net->ipv4.devconf_dflt;
1942 break;
1943 default:
1944 dev = __dev_get_by_index(net, ifindex);
1945 if (!dev)
1946 goto errout;
1947 in_dev = __in_dev_get_rtnl(dev);
1948 if (!in_dev)
1949 goto errout;
1950 devconf = &in_dev->cnf;
1951 break;
1952 }
1953
1954 err = -ENOBUFS;
1955 skb = nlmsg_new(inet_netconf_msgsize_devconf(NETCONFA_ALL), GFP_KERNEL);
1956 if (!skb)
1957 goto errout;
1958
1959 err = inet_netconf_fill_devconf(skb, ifindex, devconf,
1960 NETLINK_CB(in_skb).portid,
1961 nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
1962 NETCONFA_ALL);
1963 if (err < 0) {
1964 /* -EMSGSIZE implies BUG in inet_netconf_msgsize_devconf() */
1965 WARN_ON(err == -EMSGSIZE);
1966 kfree_skb(skb);
1967 goto errout;
1968 }
1969 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
1970errout:
1971 return err;
1972}
1973
1974static int inet_netconf_dump_devconf(struct sk_buff *skb,
1975 struct netlink_callback *cb)
1976{
1977 struct net *net = sock_net(skb->sk);
1978 int h, s_h;
1979 int idx, s_idx;
1980 struct net_device *dev;
1981 struct in_device *in_dev;
1982 struct hlist_head *head;
1983
1984 s_h = cb->args[0];
1985 s_idx = idx = cb->args[1];
1986
1987 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1988 idx = 0;
1989 head = &net->dev_index_head[h];
1990 rcu_read_lock();
1991 cb->seq = atomic_read(&net->ipv4.dev_addr_genid) ^
1992 net->dev_base_seq;
1993 hlist_for_each_entry_rcu(dev, head, index_hlist) {
1994 if (idx < s_idx)
1995 goto cont;
1996 in_dev = __in_dev_get_rcu(dev);
1997 if (!in_dev)
1998 goto cont;
1999
2000 if (inet_netconf_fill_devconf(skb, dev->ifindex,
2001 &in_dev->cnf,
2002 NETLINK_CB(cb->skb).portid,
2003 cb->nlh->nlmsg_seq,
2004 RTM_NEWNETCONF,
2005 NLM_F_MULTI,
2006 NETCONFA_ALL) < 0) {
2007 rcu_read_unlock();
2008 goto done;
2009 }
2010 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
2011cont:
2012 idx++;
2013 }
2014 rcu_read_unlock();
2015 }
2016 if (h == NETDEV_HASHENTRIES) {
2017 if (inet_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
2018 net->ipv4.devconf_all,
2019 NETLINK_CB(cb->skb).portid,
2020 cb->nlh->nlmsg_seq,
2021 RTM_NEWNETCONF, NLM_F_MULTI,
2022 NETCONFA_ALL) < 0)
2023 goto done;
2024 else
2025 h++;
2026 }
2027 if (h == NETDEV_HASHENTRIES + 1) {
2028 if (inet_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
2029 net->ipv4.devconf_dflt,
2030 NETLINK_CB(cb->skb).portid,
2031 cb->nlh->nlmsg_seq,
2032 RTM_NEWNETCONF, NLM_F_MULTI,
2033 NETCONFA_ALL) < 0)
2034 goto done;
2035 else
2036 h++;
2037 }
2038done:
2039 cb->args[0] = h;
2040 cb->args[1] = idx;
2041
2042 return skb->len;
2043}
2044
2045#ifdef CONFIG_SYSCTL
2046
2047static void devinet_copy_dflt_conf(struct net *net, int i)
2048{
2049 struct net_device *dev;
2050
2051 rcu_read_lock();
2052 for_each_netdev_rcu(net, dev) {
2053 struct in_device *in_dev;
2054
2055 in_dev = __in_dev_get_rcu(dev);
2056 if (in_dev && !test_bit(i, in_dev->cnf.state))
2057 in_dev->cnf.data[i] = net->ipv4.devconf_dflt->data[i];
2058 }
2059 rcu_read_unlock();
2060}
2061
2062/* called with RTNL locked */
2063static void inet_forward_change(struct net *net)
2064{
2065 struct net_device *dev;
2066 int on = IPV4_DEVCONF_ALL(net, FORWARDING);
2067
2068 IPV4_DEVCONF_ALL(net, ACCEPT_REDIRECTS) = !on;
2069 IPV4_DEVCONF_DFLT(net, FORWARDING) = on;
2070 inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
2071 NETCONFA_FORWARDING,
2072 NETCONFA_IFINDEX_ALL,
2073 net->ipv4.devconf_all);
2074 inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
2075 NETCONFA_FORWARDING,
2076 NETCONFA_IFINDEX_DEFAULT,
2077 net->ipv4.devconf_dflt);
2078
2079 for_each_netdev(net, dev) {
2080 struct in_device *in_dev;
2081
2082 if (on)
2083 dev_disable_lro(dev);
2084
2085 in_dev = __in_dev_get_rtnl(dev);
2086 if (in_dev) {
2087 IN_DEV_CONF_SET(in_dev, FORWARDING, on);
2088 inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
2089 NETCONFA_FORWARDING,
2090 dev->ifindex, &in_dev->cnf);
2091 }
2092 }
2093}
2094
2095static int devinet_conf_ifindex(struct net *net, struct ipv4_devconf *cnf)
2096{
2097 if (cnf == net->ipv4.devconf_dflt)
2098 return NETCONFA_IFINDEX_DEFAULT;
2099 else if (cnf == net->ipv4.devconf_all)
2100 return NETCONFA_IFINDEX_ALL;
2101 else {
2102 struct in_device *idev
2103 = container_of(cnf, struct in_device, cnf);
2104 return idev->dev->ifindex;
2105 }
2106}
2107
2108static int devinet_conf_proc(struct ctl_table *ctl, int write,
2109 void __user *buffer,
2110 size_t *lenp, loff_t *ppos)
2111{
2112 int old_value = *(int *)ctl->data;
2113 int ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
2114 int new_value = *(int *)ctl->data;
2115
2116 if (write) {
2117 struct ipv4_devconf *cnf = ctl->extra1;
2118 struct net *net = ctl->extra2;
2119 int i = (int *)ctl->data - cnf->data;
2120 int ifindex;
2121
2122 set_bit(i, cnf->state);
2123
2124 if (cnf == net->ipv4.devconf_dflt)
2125 devinet_copy_dflt_conf(net, i);
2126 if (i == IPV4_DEVCONF_ACCEPT_LOCAL - 1 ||
2127 i == IPV4_DEVCONF_ROUTE_LOCALNET - 1)
2128 if ((new_value == 0) && (old_value != 0))
2129 rt_cache_flush(net);
2130
2131 if (i == IPV4_DEVCONF_RP_FILTER - 1 &&
2132 new_value != old_value) {
2133 ifindex = devinet_conf_ifindex(net, cnf);
2134 inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
2135 NETCONFA_RP_FILTER,
2136 ifindex, cnf);
2137 }
2138 if (i == IPV4_DEVCONF_PROXY_ARP - 1 &&
2139 new_value != old_value) {
2140 ifindex = devinet_conf_ifindex(net, cnf);
2141 inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
2142 NETCONFA_PROXY_NEIGH,
2143 ifindex, cnf);
2144 }
2145 if (i == IPV4_DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN - 1 &&
2146 new_value != old_value) {
2147 ifindex = devinet_conf_ifindex(net, cnf);
2148 inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
2149 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
2150 ifindex, cnf);
2151 }
2152 }
2153
2154 return ret;
2155}
2156
2157static int devinet_sysctl_forward(struct ctl_table *ctl, int write,
2158 void __user *buffer,
2159 size_t *lenp, loff_t *ppos)
2160{
2161 int *valp = ctl->data;
2162 int val = *valp;
2163 loff_t pos = *ppos;
2164 int ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
2165
2166 if (write && *valp != val) {
2167 struct net *net = ctl->extra2;
2168
2169 if (valp != &IPV4_DEVCONF_DFLT(net, FORWARDING)) {
2170 if (!rtnl_trylock()) {
2171 /* Restore the original values before restarting */
2172 *valp = val;
2173 *ppos = pos;
2174 return restart_syscall();
2175 }
2176 if (valp == &IPV4_DEVCONF_ALL(net, FORWARDING)) {
2177 inet_forward_change(net);
2178 } else {
2179 struct ipv4_devconf *cnf = ctl->extra1;
2180 struct in_device *idev =
2181 container_of(cnf, struct in_device, cnf);
2182 if (*valp)
2183 dev_disable_lro(idev->dev);
2184 inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
2185 NETCONFA_FORWARDING,
2186 idev->dev->ifindex,
2187 cnf);
2188 }
2189 rtnl_unlock();
2190 rt_cache_flush(net);
2191 } else
2192 inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
2193 NETCONFA_FORWARDING,
2194 NETCONFA_IFINDEX_DEFAULT,
2195 net->ipv4.devconf_dflt);
2196 }
2197
2198 return ret;
2199}
2200
2201static int ipv4_doint_and_flush(struct ctl_table *ctl, int write,
2202 void __user *buffer,
2203 size_t *lenp, loff_t *ppos)
2204{
2205 int *valp = ctl->data;
2206 int val = *valp;
2207 int ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
2208 struct net *net = ctl->extra2;
2209
2210 if (write && *valp != val)
2211 rt_cache_flush(net);
2212
2213 return ret;
2214}
2215
2216#define DEVINET_SYSCTL_ENTRY(attr, name, mval, proc) \
2217 { \
2218 .procname = name, \
2219 .data = ipv4_devconf.data + \
2220 IPV4_DEVCONF_ ## attr - 1, \
2221 .maxlen = sizeof(int), \
2222 .mode = mval, \
2223 .proc_handler = proc, \
2224 .extra1 = &ipv4_devconf, \
2225 }
2226
2227#define DEVINET_SYSCTL_RW_ENTRY(attr, name) \
2228 DEVINET_SYSCTL_ENTRY(attr, name, 0644, devinet_conf_proc)
2229
2230#define DEVINET_SYSCTL_RO_ENTRY(attr, name) \
2231 DEVINET_SYSCTL_ENTRY(attr, name, 0444, devinet_conf_proc)
2232
2233#define DEVINET_SYSCTL_COMPLEX_ENTRY(attr, name, proc) \
2234 DEVINET_SYSCTL_ENTRY(attr, name, 0644, proc)
2235
2236#define DEVINET_SYSCTL_FLUSHING_ENTRY(attr, name) \
2237 DEVINET_SYSCTL_COMPLEX_ENTRY(attr, name, ipv4_doint_and_flush)
2238
2239static struct devinet_sysctl_table {
2240 struct ctl_table_header *sysctl_header;
2241 struct ctl_table devinet_vars[__IPV4_DEVCONF_MAX];
2242} devinet_sysctl = {
2243 .devinet_vars = {
2244 DEVINET_SYSCTL_COMPLEX_ENTRY(FORWARDING, "forwarding",
2245 devinet_sysctl_forward),
2246 DEVINET_SYSCTL_RO_ENTRY(MC_FORWARDING, "mc_forwarding"),
2247
2248 DEVINET_SYSCTL_RW_ENTRY(ACCEPT_REDIRECTS, "accept_redirects"),
2249 DEVINET_SYSCTL_RW_ENTRY(SECURE_REDIRECTS, "secure_redirects"),
2250 DEVINET_SYSCTL_RW_ENTRY(SHARED_MEDIA, "shared_media"),
2251 DEVINET_SYSCTL_RW_ENTRY(RP_FILTER, "rp_filter"),
2252 DEVINET_SYSCTL_RW_ENTRY(SEND_REDIRECTS, "send_redirects"),
2253 DEVINET_SYSCTL_RW_ENTRY(ACCEPT_SOURCE_ROUTE,
2254 "accept_source_route"),
2255 DEVINET_SYSCTL_RW_ENTRY(ACCEPT_LOCAL, "accept_local"),
2256 DEVINET_SYSCTL_RW_ENTRY(SRC_VMARK, "src_valid_mark"),
2257 DEVINET_SYSCTL_RW_ENTRY(PROXY_ARP, "proxy_arp"),
2258 DEVINET_SYSCTL_RW_ENTRY(MEDIUM_ID, "medium_id"),
2259 DEVINET_SYSCTL_RW_ENTRY(BOOTP_RELAY, "bootp_relay"),
2260 DEVINET_SYSCTL_RW_ENTRY(LOG_MARTIANS, "log_martians"),
2261 DEVINET_SYSCTL_RW_ENTRY(TAG, "tag"),
2262 DEVINET_SYSCTL_RW_ENTRY(ARPFILTER, "arp_filter"),
2263 DEVINET_SYSCTL_RW_ENTRY(ARP_ANNOUNCE, "arp_announce"),
2264 DEVINET_SYSCTL_RW_ENTRY(ARP_IGNORE, "arp_ignore"),
2265 DEVINET_SYSCTL_RW_ENTRY(ARP_ACCEPT, "arp_accept"),
2266 DEVINET_SYSCTL_RW_ENTRY(ARP_NOTIFY, "arp_notify"),
2267 DEVINET_SYSCTL_RW_ENTRY(PROXY_ARP_PVLAN, "proxy_arp_pvlan"),
2268 DEVINET_SYSCTL_RW_ENTRY(FORCE_IGMP_VERSION,
2269 "force_igmp_version"),
2270 DEVINET_SYSCTL_RW_ENTRY(IGMPV2_UNSOLICITED_REPORT_INTERVAL,
2271 "igmpv2_unsolicited_report_interval"),
2272 DEVINET_SYSCTL_RW_ENTRY(IGMPV3_UNSOLICITED_REPORT_INTERVAL,
2273 "igmpv3_unsolicited_report_interval"),
2274 DEVINET_SYSCTL_RW_ENTRY(IGNORE_ROUTES_WITH_LINKDOWN,
2275 "ignore_routes_with_linkdown"),
2276 DEVINET_SYSCTL_RW_ENTRY(DROP_GRATUITOUS_ARP,
2277 "drop_gratuitous_arp"),
2278
2279 DEVINET_SYSCTL_FLUSHING_ENTRY(NOXFRM, "disable_xfrm"),
2280 DEVINET_SYSCTL_FLUSHING_ENTRY(NOPOLICY, "disable_policy"),
2281 DEVINET_SYSCTL_FLUSHING_ENTRY(PROMOTE_SECONDARIES,
2282 "promote_secondaries"),
2283 DEVINET_SYSCTL_FLUSHING_ENTRY(ROUTE_LOCALNET,
2284 "route_localnet"),
2285 DEVINET_SYSCTL_FLUSHING_ENTRY(DROP_UNICAST_IN_L2_MULTICAST,
2286 "drop_unicast_in_l2_multicast"),
2287 },
2288};
2289
2290static int __devinet_sysctl_register(struct net *net, char *dev_name,
2291 int ifindex, struct ipv4_devconf *p)
2292{
2293 int i;
2294 struct devinet_sysctl_table *t;
2295 char path[sizeof("net/ipv4/conf/") + IFNAMSIZ];
2296
2297 t = kmemdup(&devinet_sysctl, sizeof(*t), GFP_KERNEL);
2298 if (!t)
2299 goto out;
2300
2301 for (i = 0; i < ARRAY_SIZE(t->devinet_vars) - 1; i++) {
2302 t->devinet_vars[i].data += (char *)p - (char *)&ipv4_devconf;
2303 t->devinet_vars[i].extra1 = p;
2304 t->devinet_vars[i].extra2 = net;
2305 }
2306
2307 snprintf(path, sizeof(path), "net/ipv4/conf/%s", dev_name);
2308
2309 t->sysctl_header = register_net_sysctl(net, path, t->devinet_vars);
2310 if (!t->sysctl_header)
2311 goto free;
2312
2313 p->sysctl = t;
2314
2315 inet_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
2316 ifindex, p);
2317 return 0;
2318
2319free:
2320 kfree(t);
2321out:
2322 return -ENOBUFS;
2323}
2324
2325static void __devinet_sysctl_unregister(struct net *net,
2326 struct ipv4_devconf *cnf, int ifindex)
2327{
2328 struct devinet_sysctl_table *t = cnf->sysctl;
2329
2330 if (t) {
2331 cnf->sysctl = NULL;
2332 unregister_net_sysctl_table(t->sysctl_header);
2333 kfree(t);
2334 }
2335
2336 inet_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
2337}
2338
2339static int devinet_sysctl_register(struct in_device *idev)
2340{
2341 int err;
2342
2343 if (!sysctl_dev_name_is_allowed(idev->dev->name))
2344 return -EINVAL;
2345
2346 err = neigh_sysctl_register(idev->dev, idev->arp_parms, NULL);
2347 if (err)
2348 return err;
2349 err = __devinet_sysctl_register(dev_net(idev->dev), idev->dev->name,
2350 idev->dev->ifindex, &idev->cnf);
2351 if (err)
2352 neigh_sysctl_unregister(idev->arp_parms);
2353 return err;
2354}
2355
2356static void devinet_sysctl_unregister(struct in_device *idev)
2357{
2358 struct net *net = dev_net(idev->dev);
2359
2360 __devinet_sysctl_unregister(net, &idev->cnf, idev->dev->ifindex);
2361 neigh_sysctl_unregister(idev->arp_parms);
2362}
2363
2364static struct ctl_table ctl_forward_entry[] = {
2365 {
2366 .procname = "ip_forward",
2367 .data = &ipv4_devconf.data[
2368 IPV4_DEVCONF_FORWARDING - 1],
2369 .maxlen = sizeof(int),
2370 .mode = 0644,
2371 .proc_handler = devinet_sysctl_forward,
2372 .extra1 = &ipv4_devconf,
2373 .extra2 = &init_net,
2374 },
2375 { },
2376};
2377#endif
2378
2379static __net_init int devinet_init_net(struct net *net)
2380{
2381 int err;
2382 struct ipv4_devconf *all, *dflt;
2383#ifdef CONFIG_SYSCTL
2384 struct ctl_table *tbl = ctl_forward_entry;
2385 struct ctl_table_header *forw_hdr;
2386#endif
2387
2388 err = -ENOMEM;
2389 all = &ipv4_devconf;
2390 dflt = &ipv4_devconf_dflt;
2391
2392 if (!net_eq(net, &init_net)) {
2393 all = kmemdup(all, sizeof(ipv4_devconf), GFP_KERNEL);
2394 if (!all)
2395 goto err_alloc_all;
2396
2397 dflt = kmemdup(dflt, sizeof(ipv4_devconf_dflt), GFP_KERNEL);
2398 if (!dflt)
2399 goto err_alloc_dflt;
2400
2401#ifdef CONFIG_SYSCTL
2402 tbl = kmemdup(tbl, sizeof(ctl_forward_entry), GFP_KERNEL);
2403 if (!tbl)
2404 goto err_alloc_ctl;
2405
2406 tbl[0].data = &all->data[IPV4_DEVCONF_FORWARDING - 1];
2407 tbl[0].extra1 = all;
2408 tbl[0].extra2 = net;
2409#endif
2410 }
2411
2412#ifdef CONFIG_SYSCTL
2413 err = __devinet_sysctl_register(net, "all", NETCONFA_IFINDEX_ALL, all);
2414 if (err < 0)
2415 goto err_reg_all;
2416
2417 err = __devinet_sysctl_register(net, "default",
2418 NETCONFA_IFINDEX_DEFAULT, dflt);
2419 if (err < 0)
2420 goto err_reg_dflt;
2421
2422 err = -ENOMEM;
2423 forw_hdr = register_net_sysctl(net, "net/ipv4", tbl);
2424 if (!forw_hdr)
2425 goto err_reg_ctl;
2426 net->ipv4.forw_hdr = forw_hdr;
2427#endif
2428
2429 net->ipv4.devconf_all = all;
2430 net->ipv4.devconf_dflt = dflt;
2431 return 0;
2432
2433#ifdef CONFIG_SYSCTL
2434err_reg_ctl:
2435 __devinet_sysctl_unregister(net, dflt, NETCONFA_IFINDEX_DEFAULT);
2436err_reg_dflt:
2437 __devinet_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
2438err_reg_all:
2439 if (tbl != ctl_forward_entry)
2440 kfree(tbl);
2441err_alloc_ctl:
2442#endif
2443 if (dflt != &ipv4_devconf_dflt)
2444 kfree(dflt);
2445err_alloc_dflt:
2446 if (all != &ipv4_devconf)
2447 kfree(all);
2448err_alloc_all:
2449 return err;
2450}
2451
2452static __net_exit void devinet_exit_net(struct net *net)
2453{
2454#ifdef CONFIG_SYSCTL
2455 struct ctl_table *tbl;
2456
2457 tbl = net->ipv4.forw_hdr->ctl_table_arg;
2458 unregister_net_sysctl_table(net->ipv4.forw_hdr);
2459 __devinet_sysctl_unregister(net, net->ipv4.devconf_dflt,
2460 NETCONFA_IFINDEX_DEFAULT);
2461 __devinet_sysctl_unregister(net, net->ipv4.devconf_all,
2462 NETCONFA_IFINDEX_ALL);
2463 kfree(tbl);
2464#endif
2465 kfree(net->ipv4.devconf_dflt);
2466 kfree(net->ipv4.devconf_all);
2467}
2468
2469static __net_initdata struct pernet_operations devinet_ops = {
2470 .init = devinet_init_net,
2471 .exit = devinet_exit_net,
2472};
2473
2474static struct rtnl_af_ops inet_af_ops __read_mostly = {
2475 .family = AF_INET,
2476 .fill_link_af = inet_fill_link_af,
2477 .get_link_af_size = inet_get_link_af_size,
2478 .validate_link_af = inet_validate_link_af,
2479 .set_link_af = inet_set_link_af,
2480};
2481
2482void __init devinet_init(void)
2483{
2484 int i;
2485
2486 for (i = 0; i < IN4_ADDR_HSIZE; i++)
2487 INIT_HLIST_HEAD(&inet_addr_lst[i]);
2488
2489 register_pernet_subsys(&devinet_ops);
2490
2491 register_gifconf(PF_INET, inet_gifconf);
2492 register_netdevice_notifier(&ip_netdev_notifier);
2493
2494 queue_delayed_work(system_power_efficient_wq, &check_lifetime_work, 0);
2495
2496 rtnl_af_register(&inet_af_ops);
2497
2498 rtnl_register(PF_INET, RTM_NEWADDR, inet_rtm_newaddr, NULL, 0);
2499 rtnl_register(PF_INET, RTM_DELADDR, inet_rtm_deladdr, NULL, 0);
2500 rtnl_register(PF_INET, RTM_GETADDR, NULL, inet_dump_ifaddr, 0);
2501 rtnl_register(PF_INET, RTM_GETNETCONF, inet_netconf_get_devconf,
2502 inet_netconf_dump_devconf, 0);
2503}
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 struct hlist_head inet_addr_lst[IN4_ADDR_HSIZE];
123
124static u32 inet_addr_hash(const struct net *net, __be32 addr)
125{
126 u32 val = (__force u32) addr ^ net_hash_mix(net);
127
128 return hash_32(val, IN4_ADDR_HSIZE_SHIFT);
129}
130
131static void inet_hash_insert(struct net *net, struct in_ifaddr *ifa)
132{
133 u32 hash = inet_addr_hash(net, ifa->ifa_local);
134
135 ASSERT_RTNL();
136 hlist_add_head_rcu(&ifa->hash, &inet_addr_lst[hash]);
137}
138
139static void inet_hash_remove(struct in_ifaddr *ifa)
140{
141 ASSERT_RTNL();
142 hlist_del_init_rcu(&ifa->hash);
143}
144
145/**
146 * __ip_dev_find - find the first device with a given source address.
147 * @net: the net namespace
148 * @addr: the source address
149 * @devref: if true, take a reference on the found device
150 *
151 * If a caller uses devref=false, it should be protected by RCU, or RTNL
152 */
153struct net_device *__ip_dev_find(struct net *net, __be32 addr, bool devref)
154{
155 struct net_device *result = NULL;
156 struct in_ifaddr *ifa;
157
158 rcu_read_lock();
159 ifa = inet_lookup_ifaddr_rcu(net, addr);
160 if (!ifa) {
161 struct flowi4 fl4 = { .daddr = addr };
162 struct fib_result res = { 0 };
163 struct fib_table *local;
164
165 /* Fallback to FIB local table so that communication
166 * over loopback subnets work.
167 */
168 local = fib_get_table(net, RT_TABLE_LOCAL);
169 if (local &&
170 !fib_table_lookup(local, &fl4, &res, FIB_LOOKUP_NOREF) &&
171 res.type == RTN_LOCAL)
172 result = FIB_RES_DEV(res);
173 } else {
174 result = ifa->ifa_dev->dev;
175 }
176 if (result && devref)
177 dev_hold(result);
178 rcu_read_unlock();
179 return result;
180}
181EXPORT_SYMBOL(__ip_dev_find);
182
183/* called under RCU lock */
184struct in_ifaddr *inet_lookup_ifaddr_rcu(struct net *net, __be32 addr)
185{
186 u32 hash = inet_addr_hash(net, addr);
187 struct in_ifaddr *ifa;
188
189 hlist_for_each_entry_rcu(ifa, &inet_addr_lst[hash], hash)
190 if (ifa->ifa_local == addr &&
191 net_eq(dev_net(ifa->ifa_dev->dev), net))
192 return ifa;
193
194 return NULL;
195}
196
197static void rtmsg_ifa(int event, struct in_ifaddr *, struct nlmsghdr *, u32);
198
199static BLOCKING_NOTIFIER_HEAD(inetaddr_chain);
200static BLOCKING_NOTIFIER_HEAD(inetaddr_validator_chain);
201static void inet_del_ifa(struct in_device *in_dev,
202 struct in_ifaddr __rcu **ifap,
203 int destroy);
204#ifdef CONFIG_SYSCTL
205static int devinet_sysctl_register(struct in_device *idev);
206static void devinet_sysctl_unregister(struct in_device *idev);
207#else
208static int devinet_sysctl_register(struct in_device *idev)
209{
210 return 0;
211}
212static void devinet_sysctl_unregister(struct in_device *idev)
213{
214}
215#endif
216
217/* Locks all the inet devices. */
218
219static struct in_ifaddr *inet_alloc_ifa(void)
220{
221 return kzalloc(sizeof(struct in_ifaddr), GFP_KERNEL_ACCOUNT);
222}
223
224static void inet_rcu_free_ifa(struct rcu_head *head)
225{
226 struct in_ifaddr *ifa = container_of(head, struct in_ifaddr, rcu_head);
227 if (ifa->ifa_dev)
228 in_dev_put(ifa->ifa_dev);
229 kfree(ifa);
230}
231
232static void inet_free_ifa(struct in_ifaddr *ifa)
233{
234 call_rcu(&ifa->rcu_head, inet_rcu_free_ifa);
235}
236
237static void in_dev_free_rcu(struct rcu_head *head)
238{
239 struct in_device *idev = container_of(head, struct in_device, rcu_head);
240
241 kfree(rcu_dereference_protected(idev->mc_hash, 1));
242 kfree(idev);
243}
244
245void in_dev_finish_destroy(struct in_device *idev)
246{
247 struct net_device *dev = idev->dev;
248
249 WARN_ON(idev->ifa_list);
250 WARN_ON(idev->mc_list);
251#ifdef NET_REFCNT_DEBUG
252 pr_debug("%s: %p=%s\n", __func__, idev, dev ? dev->name : "NIL");
253#endif
254 netdev_put(dev, &idev->dev_tracker);
255 if (!idev->dead)
256 pr_err("Freeing alive in_device %p\n", idev);
257 else
258 call_rcu(&idev->rcu_head, in_dev_free_rcu);
259}
260EXPORT_SYMBOL(in_dev_finish_destroy);
261
262static struct in_device *inetdev_init(struct net_device *dev)
263{
264 struct in_device *in_dev;
265 int err = -ENOMEM;
266
267 ASSERT_RTNL();
268
269 in_dev = kzalloc(sizeof(*in_dev), GFP_KERNEL);
270 if (!in_dev)
271 goto out;
272 memcpy(&in_dev->cnf, dev_net(dev)->ipv4.devconf_dflt,
273 sizeof(in_dev->cnf));
274 in_dev->cnf.sysctl = NULL;
275 in_dev->dev = dev;
276 in_dev->arp_parms = neigh_parms_alloc(dev, &arp_tbl);
277 if (!in_dev->arp_parms)
278 goto out_kfree;
279 if (IPV4_DEVCONF(in_dev->cnf, FORWARDING))
280 dev_disable_lro(dev);
281 /* Reference in_dev->dev */
282 netdev_hold(dev, &in_dev->dev_tracker, GFP_KERNEL);
283 /* Account for reference dev->ip_ptr (below) */
284 refcount_set(&in_dev->refcnt, 1);
285
286 err = devinet_sysctl_register(in_dev);
287 if (err) {
288 in_dev->dead = 1;
289 neigh_parms_release(&arp_tbl, in_dev->arp_parms);
290 in_dev_put(in_dev);
291 in_dev = NULL;
292 goto out;
293 }
294 ip_mc_init_dev(in_dev);
295 if (dev->flags & IFF_UP)
296 ip_mc_up(in_dev);
297
298 /* we can receive as soon as ip_ptr is set -- do this last */
299 rcu_assign_pointer(dev->ip_ptr, in_dev);
300out:
301 return in_dev ?: ERR_PTR(err);
302out_kfree:
303 kfree(in_dev);
304 in_dev = NULL;
305 goto out;
306}
307
308static void inetdev_destroy(struct in_device *in_dev)
309{
310 struct net_device *dev;
311 struct in_ifaddr *ifa;
312
313 ASSERT_RTNL();
314
315 dev = in_dev->dev;
316
317 in_dev->dead = 1;
318
319 ip_mc_destroy_dev(in_dev);
320
321 while ((ifa = rtnl_dereference(in_dev->ifa_list)) != NULL) {
322 inet_del_ifa(in_dev, &in_dev->ifa_list, 0);
323 inet_free_ifa(ifa);
324 }
325
326 RCU_INIT_POINTER(dev->ip_ptr, NULL);
327
328 devinet_sysctl_unregister(in_dev);
329 neigh_parms_release(&arp_tbl, in_dev->arp_parms);
330 arp_ifdown(dev);
331
332 in_dev_put(in_dev);
333}
334
335int inet_addr_onlink(struct in_device *in_dev, __be32 a, __be32 b)
336{
337 const struct in_ifaddr *ifa;
338
339 rcu_read_lock();
340 in_dev_for_each_ifa_rcu(ifa, in_dev) {
341 if (inet_ifa_match(a, ifa)) {
342 if (!b || inet_ifa_match(b, ifa)) {
343 rcu_read_unlock();
344 return 1;
345 }
346 }
347 }
348 rcu_read_unlock();
349 return 0;
350}
351
352static void __inet_del_ifa(struct in_device *in_dev,
353 struct in_ifaddr __rcu **ifap,
354 int destroy, struct nlmsghdr *nlh, u32 portid)
355{
356 struct in_ifaddr *promote = NULL;
357 struct in_ifaddr *ifa, *ifa1;
358 struct in_ifaddr __rcu **last_prim;
359 struct in_ifaddr *prev_prom = NULL;
360 int do_promote = IN_DEV_PROMOTE_SECONDARIES(in_dev);
361
362 ASSERT_RTNL();
363
364 ifa1 = rtnl_dereference(*ifap);
365 last_prim = ifap;
366 if (in_dev->dead)
367 goto no_promotions;
368
369 /* 1. Deleting primary ifaddr forces deletion all secondaries
370 * unless alias promotion is set
371 **/
372
373 if (!(ifa1->ifa_flags & IFA_F_SECONDARY)) {
374 struct in_ifaddr __rcu **ifap1 = &ifa1->ifa_next;
375
376 while ((ifa = rtnl_dereference(*ifap1)) != NULL) {
377 if (!(ifa->ifa_flags & IFA_F_SECONDARY) &&
378 ifa1->ifa_scope <= ifa->ifa_scope)
379 last_prim = &ifa->ifa_next;
380
381 if (!(ifa->ifa_flags & IFA_F_SECONDARY) ||
382 ifa1->ifa_mask != ifa->ifa_mask ||
383 !inet_ifa_match(ifa1->ifa_address, ifa)) {
384 ifap1 = &ifa->ifa_next;
385 prev_prom = ifa;
386 continue;
387 }
388
389 if (!do_promote) {
390 inet_hash_remove(ifa);
391 *ifap1 = ifa->ifa_next;
392
393 rtmsg_ifa(RTM_DELADDR, ifa, nlh, portid);
394 blocking_notifier_call_chain(&inetaddr_chain,
395 NETDEV_DOWN, ifa);
396 inet_free_ifa(ifa);
397 } else {
398 promote = ifa;
399 break;
400 }
401 }
402 }
403
404 /* On promotion all secondaries from subnet are changing
405 * the primary IP, we must remove all their routes silently
406 * and later to add them back with new prefsrc. Do this
407 * while all addresses are on the device list.
408 */
409 for (ifa = promote; ifa; ifa = rtnl_dereference(ifa->ifa_next)) {
410 if (ifa1->ifa_mask == ifa->ifa_mask &&
411 inet_ifa_match(ifa1->ifa_address, ifa))
412 fib_del_ifaddr(ifa, ifa1);
413 }
414
415no_promotions:
416 /* 2. Unlink it */
417
418 *ifap = ifa1->ifa_next;
419 inet_hash_remove(ifa1);
420
421 /* 3. Announce address deletion */
422
423 /* Send message first, then call notifier.
424 At first sight, FIB update triggered by notifier
425 will refer to already deleted ifaddr, that could confuse
426 netlink listeners. It is not true: look, gated sees
427 that route deleted and if it still thinks that ifaddr
428 is valid, it will try to restore deleted routes... Grr.
429 So that, this order is correct.
430 */
431 rtmsg_ifa(RTM_DELADDR, ifa1, nlh, portid);
432 blocking_notifier_call_chain(&inetaddr_chain, NETDEV_DOWN, ifa1);
433
434 if (promote) {
435 struct in_ifaddr *next_sec;
436
437 next_sec = rtnl_dereference(promote->ifa_next);
438 if (prev_prom) {
439 struct in_ifaddr *last_sec;
440
441 rcu_assign_pointer(prev_prom->ifa_next, next_sec);
442
443 last_sec = rtnl_dereference(*last_prim);
444 rcu_assign_pointer(promote->ifa_next, last_sec);
445 rcu_assign_pointer(*last_prim, promote);
446 }
447
448 promote->ifa_flags &= ~IFA_F_SECONDARY;
449 rtmsg_ifa(RTM_NEWADDR, promote, nlh, portid);
450 blocking_notifier_call_chain(&inetaddr_chain,
451 NETDEV_UP, promote);
452 for (ifa = next_sec; ifa;
453 ifa = rtnl_dereference(ifa->ifa_next)) {
454 if (ifa1->ifa_mask != ifa->ifa_mask ||
455 !inet_ifa_match(ifa1->ifa_address, ifa))
456 continue;
457 fib_add_ifaddr(ifa);
458 }
459
460 }
461 if (destroy)
462 inet_free_ifa(ifa1);
463}
464
465static void inet_del_ifa(struct in_device *in_dev,
466 struct in_ifaddr __rcu **ifap,
467 int destroy)
468{
469 __inet_del_ifa(in_dev, ifap, destroy, NULL, 0);
470}
471
472static void check_lifetime(struct work_struct *work);
473
474static DECLARE_DELAYED_WORK(check_lifetime_work, check_lifetime);
475
476static int __inet_insert_ifa(struct in_ifaddr *ifa, struct nlmsghdr *nlh,
477 u32 portid, struct netlink_ext_ack *extack)
478{
479 struct in_ifaddr __rcu **last_primary, **ifap;
480 struct in_device *in_dev = ifa->ifa_dev;
481 struct in_validator_info ivi;
482 struct in_ifaddr *ifa1;
483 int ret;
484
485 ASSERT_RTNL();
486
487 if (!ifa->ifa_local) {
488 inet_free_ifa(ifa);
489 return 0;
490 }
491
492 ifa->ifa_flags &= ~IFA_F_SECONDARY;
493 last_primary = &in_dev->ifa_list;
494
495 /* Don't set IPv6 only flags to IPv4 addresses */
496 ifa->ifa_flags &= ~IPV6ONLY_FLAGS;
497
498 ifap = &in_dev->ifa_list;
499 ifa1 = rtnl_dereference(*ifap);
500
501 while (ifa1) {
502 if (!(ifa1->ifa_flags & IFA_F_SECONDARY) &&
503 ifa->ifa_scope <= ifa1->ifa_scope)
504 last_primary = &ifa1->ifa_next;
505 if (ifa1->ifa_mask == ifa->ifa_mask &&
506 inet_ifa_match(ifa1->ifa_address, ifa)) {
507 if (ifa1->ifa_local == ifa->ifa_local) {
508 inet_free_ifa(ifa);
509 return -EEXIST;
510 }
511 if (ifa1->ifa_scope != ifa->ifa_scope) {
512 NL_SET_ERR_MSG(extack, "ipv4: Invalid scope value");
513 inet_free_ifa(ifa);
514 return -EINVAL;
515 }
516 ifa->ifa_flags |= IFA_F_SECONDARY;
517 }
518
519 ifap = &ifa1->ifa_next;
520 ifa1 = rtnl_dereference(*ifap);
521 }
522
523 /* Allow any devices that wish to register ifaddr validtors to weigh
524 * in now, before changes are committed. The rntl lock is serializing
525 * access here, so the state should not change between a validator call
526 * and a final notify on commit. This isn't invoked on promotion under
527 * the assumption that validators are checking the address itself, and
528 * not the flags.
529 */
530 ivi.ivi_addr = ifa->ifa_address;
531 ivi.ivi_dev = ifa->ifa_dev;
532 ivi.extack = extack;
533 ret = blocking_notifier_call_chain(&inetaddr_validator_chain,
534 NETDEV_UP, &ivi);
535 ret = notifier_to_errno(ret);
536 if (ret) {
537 inet_free_ifa(ifa);
538 return ret;
539 }
540
541 if (!(ifa->ifa_flags & IFA_F_SECONDARY))
542 ifap = last_primary;
543
544 rcu_assign_pointer(ifa->ifa_next, *ifap);
545 rcu_assign_pointer(*ifap, ifa);
546
547 inet_hash_insert(dev_net(in_dev->dev), ifa);
548
549 cancel_delayed_work(&check_lifetime_work);
550 queue_delayed_work(system_power_efficient_wq, &check_lifetime_work, 0);
551
552 /* Send message first, then call notifier.
553 Notifier will trigger FIB update, so that
554 listeners of netlink will know about new ifaddr */
555 rtmsg_ifa(RTM_NEWADDR, ifa, nlh, portid);
556 blocking_notifier_call_chain(&inetaddr_chain, NETDEV_UP, ifa);
557
558 return 0;
559}
560
561static int inet_insert_ifa(struct in_ifaddr *ifa)
562{
563 return __inet_insert_ifa(ifa, NULL, 0, NULL);
564}
565
566static int inet_set_ifa(struct net_device *dev, struct in_ifaddr *ifa)
567{
568 struct in_device *in_dev = __in_dev_get_rtnl(dev);
569
570 ASSERT_RTNL();
571
572 if (!in_dev) {
573 inet_free_ifa(ifa);
574 return -ENOBUFS;
575 }
576 ipv4_devconf_setall(in_dev);
577 neigh_parms_data_state_setall(in_dev->arp_parms);
578 if (ifa->ifa_dev != in_dev) {
579 WARN_ON(ifa->ifa_dev);
580 in_dev_hold(in_dev);
581 ifa->ifa_dev = in_dev;
582 }
583 if (ipv4_is_loopback(ifa->ifa_local))
584 ifa->ifa_scope = RT_SCOPE_HOST;
585 return inet_insert_ifa(ifa);
586}
587
588/* Caller must hold RCU or RTNL :
589 * We dont take a reference on found in_device
590 */
591struct in_device *inetdev_by_index(struct net *net, int ifindex)
592{
593 struct net_device *dev;
594 struct in_device *in_dev = NULL;
595
596 rcu_read_lock();
597 dev = dev_get_by_index_rcu(net, ifindex);
598 if (dev)
599 in_dev = rcu_dereference_rtnl(dev->ip_ptr);
600 rcu_read_unlock();
601 return in_dev;
602}
603EXPORT_SYMBOL(inetdev_by_index);
604
605/* Called only from RTNL semaphored context. No locks. */
606
607struct in_ifaddr *inet_ifa_byprefix(struct in_device *in_dev, __be32 prefix,
608 __be32 mask)
609{
610 struct in_ifaddr *ifa;
611
612 ASSERT_RTNL();
613
614 in_dev_for_each_ifa_rtnl(ifa, in_dev) {
615 if (ifa->ifa_mask == mask && inet_ifa_match(prefix, ifa))
616 return ifa;
617 }
618 return NULL;
619}
620
621static int ip_mc_autojoin_config(struct net *net, bool join,
622 const struct in_ifaddr *ifa)
623{
624#if defined(CONFIG_IP_MULTICAST)
625 struct ip_mreqn mreq = {
626 .imr_multiaddr.s_addr = ifa->ifa_address,
627 .imr_ifindex = ifa->ifa_dev->dev->ifindex,
628 };
629 struct sock *sk = net->ipv4.mc_autojoin_sk;
630 int ret;
631
632 ASSERT_RTNL();
633
634 lock_sock(sk);
635 if (join)
636 ret = ip_mc_join_group(sk, &mreq);
637 else
638 ret = ip_mc_leave_group(sk, &mreq);
639 release_sock(sk);
640
641 return ret;
642#else
643 return -EOPNOTSUPP;
644#endif
645}
646
647static int inet_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
648 struct netlink_ext_ack *extack)
649{
650 struct net *net = sock_net(skb->sk);
651 struct in_ifaddr __rcu **ifap;
652 struct nlattr *tb[IFA_MAX+1];
653 struct in_device *in_dev;
654 struct ifaddrmsg *ifm;
655 struct in_ifaddr *ifa;
656 int err;
657
658 ASSERT_RTNL();
659
660 err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
661 ifa_ipv4_policy, extack);
662 if (err < 0)
663 goto errout;
664
665 ifm = nlmsg_data(nlh);
666 in_dev = inetdev_by_index(net, ifm->ifa_index);
667 if (!in_dev) {
668 NL_SET_ERR_MSG(extack, "ipv4: Device not found");
669 err = -ENODEV;
670 goto errout;
671 }
672
673 for (ifap = &in_dev->ifa_list; (ifa = rtnl_dereference(*ifap)) != NULL;
674 ifap = &ifa->ifa_next) {
675 if (tb[IFA_LOCAL] &&
676 ifa->ifa_local != nla_get_in_addr(tb[IFA_LOCAL]))
677 continue;
678
679 if (tb[IFA_LABEL] && nla_strcmp(tb[IFA_LABEL], ifa->ifa_label))
680 continue;
681
682 if (tb[IFA_ADDRESS] &&
683 (ifm->ifa_prefixlen != ifa->ifa_prefixlen ||
684 !inet_ifa_match(nla_get_in_addr(tb[IFA_ADDRESS]), ifa)))
685 continue;
686
687 if (ipv4_is_multicast(ifa->ifa_address))
688 ip_mc_autojoin_config(net, false, ifa);
689 __inet_del_ifa(in_dev, ifap, 1, nlh, NETLINK_CB(skb).portid);
690 return 0;
691 }
692
693 NL_SET_ERR_MSG(extack, "ipv4: Address not found");
694 err = -EADDRNOTAVAIL;
695errout:
696 return err;
697}
698
699#define INFINITY_LIFE_TIME 0xFFFFFFFF
700
701static void check_lifetime(struct work_struct *work)
702{
703 unsigned long now, next, next_sec, next_sched;
704 struct in_ifaddr *ifa;
705 struct hlist_node *n;
706 int i;
707
708 now = jiffies;
709 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
710
711 for (i = 0; i < IN4_ADDR_HSIZE; i++) {
712 bool change_needed = false;
713
714 rcu_read_lock();
715 hlist_for_each_entry_rcu(ifa, &inet_addr_lst[i], hash) {
716 unsigned long age, tstamp;
717 u32 preferred_lft;
718 u32 valid_lft;
719 u32 flags;
720
721 flags = READ_ONCE(ifa->ifa_flags);
722 if (flags & IFA_F_PERMANENT)
723 continue;
724
725 preferred_lft = READ_ONCE(ifa->ifa_preferred_lft);
726 valid_lft = READ_ONCE(ifa->ifa_valid_lft);
727 tstamp = READ_ONCE(ifa->ifa_tstamp);
728 /* We try to batch several events at once. */
729 age = (now - tstamp +
730 ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
731
732 if (valid_lft != INFINITY_LIFE_TIME &&
733 age >= valid_lft) {
734 change_needed = true;
735 } else if (preferred_lft ==
736 INFINITY_LIFE_TIME) {
737 continue;
738 } else if (age >= preferred_lft) {
739 if (time_before(tstamp + valid_lft * HZ, next))
740 next = tstamp + valid_lft * HZ;
741
742 if (!(flags & IFA_F_DEPRECATED))
743 change_needed = true;
744 } else if (time_before(tstamp + preferred_lft * HZ,
745 next)) {
746 next = tstamp + preferred_lft * HZ;
747 }
748 }
749 rcu_read_unlock();
750 if (!change_needed)
751 continue;
752 rtnl_lock();
753 hlist_for_each_entry_safe(ifa, n, &inet_addr_lst[i], hash) {
754 unsigned long age;
755
756 if (ifa->ifa_flags & IFA_F_PERMANENT)
757 continue;
758
759 /* We try to batch several events at once. */
760 age = (now - ifa->ifa_tstamp +
761 ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
762
763 if (ifa->ifa_valid_lft != INFINITY_LIFE_TIME &&
764 age >= ifa->ifa_valid_lft) {
765 struct in_ifaddr __rcu **ifap;
766 struct in_ifaddr *tmp;
767
768 ifap = &ifa->ifa_dev->ifa_list;
769 tmp = rtnl_dereference(*ifap);
770 while (tmp) {
771 if (tmp == ifa) {
772 inet_del_ifa(ifa->ifa_dev,
773 ifap, 1);
774 break;
775 }
776 ifap = &tmp->ifa_next;
777 tmp = rtnl_dereference(*ifap);
778 }
779 } else if (ifa->ifa_preferred_lft !=
780 INFINITY_LIFE_TIME &&
781 age >= ifa->ifa_preferred_lft &&
782 !(ifa->ifa_flags & IFA_F_DEPRECATED)) {
783 ifa->ifa_flags |= IFA_F_DEPRECATED;
784 rtmsg_ifa(RTM_NEWADDR, ifa, NULL, 0);
785 }
786 }
787 rtnl_unlock();
788 }
789
790 next_sec = round_jiffies_up(next);
791 next_sched = next;
792
793 /* If rounded timeout is accurate enough, accept it. */
794 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
795 next_sched = next_sec;
796
797 now = jiffies;
798 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
799 if (time_before(next_sched, now + ADDRCONF_TIMER_FUZZ_MAX))
800 next_sched = now + ADDRCONF_TIMER_FUZZ_MAX;
801
802 queue_delayed_work(system_power_efficient_wq, &check_lifetime_work,
803 next_sched - now);
804}
805
806static void set_ifa_lifetime(struct in_ifaddr *ifa, __u32 valid_lft,
807 __u32 prefered_lft)
808{
809 unsigned long timeout;
810 u32 flags;
811
812 flags = ifa->ifa_flags & ~(IFA_F_PERMANENT | IFA_F_DEPRECATED);
813
814 timeout = addrconf_timeout_fixup(valid_lft, HZ);
815 if (addrconf_finite_timeout(timeout))
816 WRITE_ONCE(ifa->ifa_valid_lft, timeout);
817 else
818 flags |= IFA_F_PERMANENT;
819
820 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
821 if (addrconf_finite_timeout(timeout)) {
822 if (timeout == 0)
823 flags |= IFA_F_DEPRECATED;
824 WRITE_ONCE(ifa->ifa_preferred_lft, timeout);
825 }
826 WRITE_ONCE(ifa->ifa_flags, flags);
827 WRITE_ONCE(ifa->ifa_tstamp, jiffies);
828 if (!ifa->ifa_cstamp)
829 WRITE_ONCE(ifa->ifa_cstamp, ifa->ifa_tstamp);
830}
831
832static struct in_ifaddr *rtm_to_ifaddr(struct net *net, struct nlmsghdr *nlh,
833 __u32 *pvalid_lft, __u32 *pprefered_lft,
834 struct netlink_ext_ack *extack)
835{
836 struct nlattr *tb[IFA_MAX+1];
837 struct in_ifaddr *ifa;
838 struct ifaddrmsg *ifm;
839 struct net_device *dev;
840 struct in_device *in_dev;
841 int err;
842
843 err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
844 ifa_ipv4_policy, extack);
845 if (err < 0)
846 goto errout;
847
848 ifm = nlmsg_data(nlh);
849 err = -EINVAL;
850
851 if (ifm->ifa_prefixlen > 32) {
852 NL_SET_ERR_MSG(extack, "ipv4: Invalid prefix length");
853 goto errout;
854 }
855
856 if (!tb[IFA_LOCAL]) {
857 NL_SET_ERR_MSG(extack, "ipv4: Local address is not supplied");
858 goto errout;
859 }
860
861 dev = __dev_get_by_index(net, ifm->ifa_index);
862 err = -ENODEV;
863 if (!dev) {
864 NL_SET_ERR_MSG(extack, "ipv4: Device not found");
865 goto errout;
866 }
867
868 in_dev = __in_dev_get_rtnl(dev);
869 err = -ENOBUFS;
870 if (!in_dev)
871 goto errout;
872
873 ifa = inet_alloc_ifa();
874 if (!ifa)
875 /*
876 * A potential indev allocation can be left alive, it stays
877 * assigned to its device and is destroy with it.
878 */
879 goto errout;
880
881 ipv4_devconf_setall(in_dev);
882 neigh_parms_data_state_setall(in_dev->arp_parms);
883 in_dev_hold(in_dev);
884
885 if (!tb[IFA_ADDRESS])
886 tb[IFA_ADDRESS] = tb[IFA_LOCAL];
887
888 INIT_HLIST_NODE(&ifa->hash);
889 ifa->ifa_prefixlen = ifm->ifa_prefixlen;
890 ifa->ifa_mask = inet_make_mask(ifm->ifa_prefixlen);
891 ifa->ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) :
892 ifm->ifa_flags;
893 ifa->ifa_scope = ifm->ifa_scope;
894 ifa->ifa_dev = in_dev;
895
896 ifa->ifa_local = nla_get_in_addr(tb[IFA_LOCAL]);
897 ifa->ifa_address = nla_get_in_addr(tb[IFA_ADDRESS]);
898
899 if (tb[IFA_BROADCAST])
900 ifa->ifa_broadcast = nla_get_in_addr(tb[IFA_BROADCAST]);
901
902 if (tb[IFA_LABEL])
903 nla_strscpy(ifa->ifa_label, tb[IFA_LABEL], IFNAMSIZ);
904 else
905 memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
906
907 if (tb[IFA_RT_PRIORITY])
908 ifa->ifa_rt_priority = nla_get_u32(tb[IFA_RT_PRIORITY]);
909
910 if (tb[IFA_PROTO])
911 ifa->ifa_proto = nla_get_u8(tb[IFA_PROTO]);
912
913 if (tb[IFA_CACHEINFO]) {
914 struct ifa_cacheinfo *ci;
915
916 ci = nla_data(tb[IFA_CACHEINFO]);
917 if (!ci->ifa_valid || ci->ifa_prefered > ci->ifa_valid) {
918 NL_SET_ERR_MSG(extack, "ipv4: address lifetime invalid");
919 err = -EINVAL;
920 goto errout_free;
921 }
922 *pvalid_lft = ci->ifa_valid;
923 *pprefered_lft = ci->ifa_prefered;
924 }
925
926 return ifa;
927
928errout_free:
929 inet_free_ifa(ifa);
930errout:
931 return ERR_PTR(err);
932}
933
934static struct in_ifaddr *find_matching_ifa(struct in_ifaddr *ifa)
935{
936 struct in_device *in_dev = ifa->ifa_dev;
937 struct in_ifaddr *ifa1;
938
939 if (!ifa->ifa_local)
940 return NULL;
941
942 in_dev_for_each_ifa_rtnl(ifa1, in_dev) {
943 if (ifa1->ifa_mask == ifa->ifa_mask &&
944 inet_ifa_match(ifa1->ifa_address, ifa) &&
945 ifa1->ifa_local == ifa->ifa_local)
946 return ifa1;
947 }
948 return NULL;
949}
950
951static int inet_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
952 struct netlink_ext_ack *extack)
953{
954 struct net *net = sock_net(skb->sk);
955 struct in_ifaddr *ifa;
956 struct in_ifaddr *ifa_existing;
957 __u32 valid_lft = INFINITY_LIFE_TIME;
958 __u32 prefered_lft = INFINITY_LIFE_TIME;
959
960 ASSERT_RTNL();
961
962 ifa = rtm_to_ifaddr(net, nlh, &valid_lft, &prefered_lft, extack);
963 if (IS_ERR(ifa))
964 return PTR_ERR(ifa);
965
966 ifa_existing = find_matching_ifa(ifa);
967 if (!ifa_existing) {
968 /* It would be best to check for !NLM_F_CREATE here but
969 * userspace already relies on not having to provide this.
970 */
971 set_ifa_lifetime(ifa, valid_lft, prefered_lft);
972 if (ifa->ifa_flags & IFA_F_MCAUTOJOIN) {
973 int ret = ip_mc_autojoin_config(net, true, ifa);
974
975 if (ret < 0) {
976 NL_SET_ERR_MSG(extack, "ipv4: Multicast auto join failed");
977 inet_free_ifa(ifa);
978 return ret;
979 }
980 }
981 return __inet_insert_ifa(ifa, nlh, NETLINK_CB(skb).portid,
982 extack);
983 } else {
984 u32 new_metric = ifa->ifa_rt_priority;
985 u8 new_proto = ifa->ifa_proto;
986
987 inet_free_ifa(ifa);
988
989 if (nlh->nlmsg_flags & NLM_F_EXCL ||
990 !(nlh->nlmsg_flags & NLM_F_REPLACE)) {
991 NL_SET_ERR_MSG(extack, "ipv4: Address already assigned");
992 return -EEXIST;
993 }
994 ifa = ifa_existing;
995
996 if (ifa->ifa_rt_priority != new_metric) {
997 fib_modify_prefix_metric(ifa, new_metric);
998 ifa->ifa_rt_priority = new_metric;
999 }
1000
1001 ifa->ifa_proto = new_proto;
1002
1003 set_ifa_lifetime(ifa, valid_lft, prefered_lft);
1004 cancel_delayed_work(&check_lifetime_work);
1005 queue_delayed_work(system_power_efficient_wq,
1006 &check_lifetime_work, 0);
1007 rtmsg_ifa(RTM_NEWADDR, ifa, nlh, NETLINK_CB(skb).portid);
1008 }
1009 return 0;
1010}
1011
1012/*
1013 * Determine a default network mask, based on the IP address.
1014 */
1015
1016static int inet_abc_len(__be32 addr)
1017{
1018 int rc = -1; /* Something else, probably a multicast. */
1019
1020 if (ipv4_is_zeronet(addr) || ipv4_is_lbcast(addr))
1021 rc = 0;
1022 else {
1023 __u32 haddr = ntohl(addr);
1024 if (IN_CLASSA(haddr))
1025 rc = 8;
1026 else if (IN_CLASSB(haddr))
1027 rc = 16;
1028 else if (IN_CLASSC(haddr))
1029 rc = 24;
1030 else if (IN_CLASSE(haddr))
1031 rc = 32;
1032 }
1033
1034 return rc;
1035}
1036
1037
1038int devinet_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr)
1039{
1040 struct sockaddr_in sin_orig;
1041 struct sockaddr_in *sin = (struct sockaddr_in *)&ifr->ifr_addr;
1042 struct in_ifaddr __rcu **ifap = NULL;
1043 struct in_device *in_dev;
1044 struct in_ifaddr *ifa = NULL;
1045 struct net_device *dev;
1046 char *colon;
1047 int ret = -EFAULT;
1048 int tryaddrmatch = 0;
1049
1050 ifr->ifr_name[IFNAMSIZ - 1] = 0;
1051
1052 /* save original address for comparison */
1053 memcpy(&sin_orig, sin, sizeof(*sin));
1054
1055 colon = strchr(ifr->ifr_name, ':');
1056 if (colon)
1057 *colon = 0;
1058
1059 dev_load(net, ifr->ifr_name);
1060
1061 switch (cmd) {
1062 case SIOCGIFADDR: /* Get interface address */
1063 case SIOCGIFBRDADDR: /* Get the broadcast address */
1064 case SIOCGIFDSTADDR: /* Get the destination address */
1065 case SIOCGIFNETMASK: /* Get the netmask for the interface */
1066 /* Note that these ioctls will not sleep,
1067 so that we do not impose a lock.
1068 One day we will be forced to put shlock here (I mean SMP)
1069 */
1070 tryaddrmatch = (sin_orig.sin_family == AF_INET);
1071 memset(sin, 0, sizeof(*sin));
1072 sin->sin_family = AF_INET;
1073 break;
1074
1075 case SIOCSIFFLAGS:
1076 ret = -EPERM;
1077 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1078 goto out;
1079 break;
1080 case SIOCSIFADDR: /* Set interface address (and family) */
1081 case SIOCSIFBRDADDR: /* Set the broadcast address */
1082 case SIOCSIFDSTADDR: /* Set the destination address */
1083 case SIOCSIFNETMASK: /* Set the netmask for the interface */
1084 ret = -EPERM;
1085 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1086 goto out;
1087 ret = -EINVAL;
1088 if (sin->sin_family != AF_INET)
1089 goto out;
1090 break;
1091 default:
1092 ret = -EINVAL;
1093 goto out;
1094 }
1095
1096 rtnl_lock();
1097
1098 ret = -ENODEV;
1099 dev = __dev_get_by_name(net, ifr->ifr_name);
1100 if (!dev)
1101 goto done;
1102
1103 if (colon)
1104 *colon = ':';
1105
1106 in_dev = __in_dev_get_rtnl(dev);
1107 if (in_dev) {
1108 if (tryaddrmatch) {
1109 /* Matthias Andree */
1110 /* compare label and address (4.4BSD style) */
1111 /* note: we only do this for a limited set of ioctls
1112 and only if the original address family was AF_INET.
1113 This is checked above. */
1114
1115 for (ifap = &in_dev->ifa_list;
1116 (ifa = rtnl_dereference(*ifap)) != NULL;
1117 ifap = &ifa->ifa_next) {
1118 if (!strcmp(ifr->ifr_name, ifa->ifa_label) &&
1119 sin_orig.sin_addr.s_addr ==
1120 ifa->ifa_local) {
1121 break; /* found */
1122 }
1123 }
1124 }
1125 /* we didn't get a match, maybe the application is
1126 4.3BSD-style and passed in junk so we fall back to
1127 comparing just the label */
1128 if (!ifa) {
1129 for (ifap = &in_dev->ifa_list;
1130 (ifa = rtnl_dereference(*ifap)) != NULL;
1131 ifap = &ifa->ifa_next)
1132 if (!strcmp(ifr->ifr_name, ifa->ifa_label))
1133 break;
1134 }
1135 }
1136
1137 ret = -EADDRNOTAVAIL;
1138 if (!ifa && cmd != SIOCSIFADDR && cmd != SIOCSIFFLAGS)
1139 goto done;
1140
1141 switch (cmd) {
1142 case SIOCGIFADDR: /* Get interface address */
1143 ret = 0;
1144 sin->sin_addr.s_addr = ifa->ifa_local;
1145 break;
1146
1147 case SIOCGIFBRDADDR: /* Get the broadcast address */
1148 ret = 0;
1149 sin->sin_addr.s_addr = ifa->ifa_broadcast;
1150 break;
1151
1152 case SIOCGIFDSTADDR: /* Get the destination address */
1153 ret = 0;
1154 sin->sin_addr.s_addr = ifa->ifa_address;
1155 break;
1156
1157 case SIOCGIFNETMASK: /* Get the netmask for the interface */
1158 ret = 0;
1159 sin->sin_addr.s_addr = ifa->ifa_mask;
1160 break;
1161
1162 case SIOCSIFFLAGS:
1163 if (colon) {
1164 ret = -EADDRNOTAVAIL;
1165 if (!ifa)
1166 break;
1167 ret = 0;
1168 if (!(ifr->ifr_flags & IFF_UP))
1169 inet_del_ifa(in_dev, ifap, 1);
1170 break;
1171 }
1172 ret = dev_change_flags(dev, ifr->ifr_flags, NULL);
1173 break;
1174
1175 case SIOCSIFADDR: /* Set interface address (and family) */
1176 ret = -EINVAL;
1177 if (inet_abc_len(sin->sin_addr.s_addr) < 0)
1178 break;
1179
1180 if (!ifa) {
1181 ret = -ENOBUFS;
1182 ifa = inet_alloc_ifa();
1183 if (!ifa)
1184 break;
1185 INIT_HLIST_NODE(&ifa->hash);
1186 if (colon)
1187 memcpy(ifa->ifa_label, ifr->ifr_name, IFNAMSIZ);
1188 else
1189 memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
1190 } else {
1191 ret = 0;
1192 if (ifa->ifa_local == sin->sin_addr.s_addr)
1193 break;
1194 inet_del_ifa(in_dev, ifap, 0);
1195 ifa->ifa_broadcast = 0;
1196 ifa->ifa_scope = 0;
1197 }
1198
1199 ifa->ifa_address = ifa->ifa_local = sin->sin_addr.s_addr;
1200
1201 if (!(dev->flags & IFF_POINTOPOINT)) {
1202 ifa->ifa_prefixlen = inet_abc_len(ifa->ifa_address);
1203 ifa->ifa_mask = inet_make_mask(ifa->ifa_prefixlen);
1204 if ((dev->flags & IFF_BROADCAST) &&
1205 ifa->ifa_prefixlen < 31)
1206 ifa->ifa_broadcast = ifa->ifa_address |
1207 ~ifa->ifa_mask;
1208 } else {
1209 ifa->ifa_prefixlen = 32;
1210 ifa->ifa_mask = inet_make_mask(32);
1211 }
1212 set_ifa_lifetime(ifa, INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
1213 ret = inet_set_ifa(dev, ifa);
1214 break;
1215
1216 case SIOCSIFBRDADDR: /* Set the broadcast address */
1217 ret = 0;
1218 if (ifa->ifa_broadcast != sin->sin_addr.s_addr) {
1219 inet_del_ifa(in_dev, ifap, 0);
1220 ifa->ifa_broadcast = sin->sin_addr.s_addr;
1221 inet_insert_ifa(ifa);
1222 }
1223 break;
1224
1225 case SIOCSIFDSTADDR: /* Set the destination address */
1226 ret = 0;
1227 if (ifa->ifa_address == sin->sin_addr.s_addr)
1228 break;
1229 ret = -EINVAL;
1230 if (inet_abc_len(sin->sin_addr.s_addr) < 0)
1231 break;
1232 ret = 0;
1233 inet_del_ifa(in_dev, ifap, 0);
1234 ifa->ifa_address = sin->sin_addr.s_addr;
1235 inet_insert_ifa(ifa);
1236 break;
1237
1238 case SIOCSIFNETMASK: /* Set the netmask for the interface */
1239
1240 /*
1241 * The mask we set must be legal.
1242 */
1243 ret = -EINVAL;
1244 if (bad_mask(sin->sin_addr.s_addr, 0))
1245 break;
1246 ret = 0;
1247 if (ifa->ifa_mask != sin->sin_addr.s_addr) {
1248 __be32 old_mask = ifa->ifa_mask;
1249 inet_del_ifa(in_dev, ifap, 0);
1250 ifa->ifa_mask = sin->sin_addr.s_addr;
1251 ifa->ifa_prefixlen = inet_mask_len(ifa->ifa_mask);
1252
1253 /* See if current broadcast address matches
1254 * with current netmask, then recalculate
1255 * the broadcast address. Otherwise it's a
1256 * funny address, so don't touch it since
1257 * the user seems to know what (s)he's doing...
1258 */
1259 if ((dev->flags & IFF_BROADCAST) &&
1260 (ifa->ifa_prefixlen < 31) &&
1261 (ifa->ifa_broadcast ==
1262 (ifa->ifa_local|~old_mask))) {
1263 ifa->ifa_broadcast = (ifa->ifa_local |
1264 ~sin->sin_addr.s_addr);
1265 }
1266 inet_insert_ifa(ifa);
1267 }
1268 break;
1269 }
1270done:
1271 rtnl_unlock();
1272out:
1273 return ret;
1274}
1275
1276int inet_gifconf(struct net_device *dev, char __user *buf, int len, int size)
1277{
1278 struct in_device *in_dev = __in_dev_get_rtnl(dev);
1279 const struct in_ifaddr *ifa;
1280 struct ifreq ifr;
1281 int done = 0;
1282
1283 if (WARN_ON(size > sizeof(struct ifreq)))
1284 goto out;
1285
1286 if (!in_dev)
1287 goto out;
1288
1289 in_dev_for_each_ifa_rtnl(ifa, in_dev) {
1290 if (!buf) {
1291 done += size;
1292 continue;
1293 }
1294 if (len < size)
1295 break;
1296 memset(&ifr, 0, sizeof(struct ifreq));
1297 strcpy(ifr.ifr_name, ifa->ifa_label);
1298
1299 (*(struct sockaddr_in *)&ifr.ifr_addr).sin_family = AF_INET;
1300 (*(struct sockaddr_in *)&ifr.ifr_addr).sin_addr.s_addr =
1301 ifa->ifa_local;
1302
1303 if (copy_to_user(buf + done, &ifr, size)) {
1304 done = -EFAULT;
1305 break;
1306 }
1307 len -= size;
1308 done += size;
1309 }
1310out:
1311 return done;
1312}
1313
1314static __be32 in_dev_select_addr(const struct in_device *in_dev,
1315 int scope)
1316{
1317 const struct in_ifaddr *ifa;
1318
1319 in_dev_for_each_ifa_rcu(ifa, in_dev) {
1320 if (READ_ONCE(ifa->ifa_flags) & IFA_F_SECONDARY)
1321 continue;
1322 if (ifa->ifa_scope != RT_SCOPE_LINK &&
1323 ifa->ifa_scope <= scope)
1324 return ifa->ifa_local;
1325 }
1326
1327 return 0;
1328}
1329
1330__be32 inet_select_addr(const struct net_device *dev, __be32 dst, int scope)
1331{
1332 const struct in_ifaddr *ifa;
1333 __be32 addr = 0;
1334 unsigned char localnet_scope = RT_SCOPE_HOST;
1335 struct in_device *in_dev;
1336 struct net *net = dev_net(dev);
1337 int master_idx;
1338
1339 rcu_read_lock();
1340 in_dev = __in_dev_get_rcu(dev);
1341 if (!in_dev)
1342 goto no_in_dev;
1343
1344 if (unlikely(IN_DEV_ROUTE_LOCALNET(in_dev)))
1345 localnet_scope = RT_SCOPE_LINK;
1346
1347 in_dev_for_each_ifa_rcu(ifa, in_dev) {
1348 if (READ_ONCE(ifa->ifa_flags) & IFA_F_SECONDARY)
1349 continue;
1350 if (min(ifa->ifa_scope, localnet_scope) > scope)
1351 continue;
1352 if (!dst || inet_ifa_match(dst, ifa)) {
1353 addr = ifa->ifa_local;
1354 break;
1355 }
1356 if (!addr)
1357 addr = ifa->ifa_local;
1358 }
1359
1360 if (addr)
1361 goto out_unlock;
1362no_in_dev:
1363 master_idx = l3mdev_master_ifindex_rcu(dev);
1364
1365 /* For VRFs, the VRF device takes the place of the loopback device,
1366 * with addresses on it being preferred. Note in such cases the
1367 * loopback device will be among the devices that fail the master_idx
1368 * equality check in the loop below.
1369 */
1370 if (master_idx &&
1371 (dev = dev_get_by_index_rcu(net, master_idx)) &&
1372 (in_dev = __in_dev_get_rcu(dev))) {
1373 addr = in_dev_select_addr(in_dev, scope);
1374 if (addr)
1375 goto out_unlock;
1376 }
1377
1378 /* Not loopback addresses on loopback should be preferred
1379 in this case. It is important that lo is the first interface
1380 in dev_base list.
1381 */
1382 for_each_netdev_rcu(net, dev) {
1383 if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1384 continue;
1385
1386 in_dev = __in_dev_get_rcu(dev);
1387 if (!in_dev)
1388 continue;
1389
1390 addr = in_dev_select_addr(in_dev, scope);
1391 if (addr)
1392 goto out_unlock;
1393 }
1394out_unlock:
1395 rcu_read_unlock();
1396 return addr;
1397}
1398EXPORT_SYMBOL(inet_select_addr);
1399
1400static __be32 confirm_addr_indev(struct in_device *in_dev, __be32 dst,
1401 __be32 local, int scope)
1402{
1403 unsigned char localnet_scope = RT_SCOPE_HOST;
1404 const struct in_ifaddr *ifa;
1405 __be32 addr = 0;
1406 int same = 0;
1407
1408 if (unlikely(IN_DEV_ROUTE_LOCALNET(in_dev)))
1409 localnet_scope = RT_SCOPE_LINK;
1410
1411 in_dev_for_each_ifa_rcu(ifa, in_dev) {
1412 unsigned char min_scope = min(ifa->ifa_scope, localnet_scope);
1413
1414 if (!addr &&
1415 (local == ifa->ifa_local || !local) &&
1416 min_scope <= scope) {
1417 addr = ifa->ifa_local;
1418 if (same)
1419 break;
1420 }
1421 if (!same) {
1422 same = (!local || inet_ifa_match(local, ifa)) &&
1423 (!dst || inet_ifa_match(dst, ifa));
1424 if (same && addr) {
1425 if (local || !dst)
1426 break;
1427 /* Is the selected addr into dst subnet? */
1428 if (inet_ifa_match(addr, ifa))
1429 break;
1430 /* No, then can we use new local src? */
1431 if (min_scope <= scope) {
1432 addr = ifa->ifa_local;
1433 break;
1434 }
1435 /* search for large dst subnet for addr */
1436 same = 0;
1437 }
1438 }
1439 }
1440
1441 return same ? addr : 0;
1442}
1443
1444/*
1445 * Confirm that local IP address exists using wildcards:
1446 * - net: netns to check, cannot be NULL
1447 * - in_dev: only on this interface, NULL=any interface
1448 * - dst: only in the same subnet as dst, 0=any dst
1449 * - local: address, 0=autoselect the local address
1450 * - scope: maximum allowed scope value for the local address
1451 */
1452__be32 inet_confirm_addr(struct net *net, struct in_device *in_dev,
1453 __be32 dst, __be32 local, int scope)
1454{
1455 __be32 addr = 0;
1456 struct net_device *dev;
1457
1458 if (in_dev)
1459 return confirm_addr_indev(in_dev, dst, local, scope);
1460
1461 rcu_read_lock();
1462 for_each_netdev_rcu(net, dev) {
1463 in_dev = __in_dev_get_rcu(dev);
1464 if (in_dev) {
1465 addr = confirm_addr_indev(in_dev, dst, local, scope);
1466 if (addr)
1467 break;
1468 }
1469 }
1470 rcu_read_unlock();
1471
1472 return addr;
1473}
1474EXPORT_SYMBOL(inet_confirm_addr);
1475
1476/*
1477 * Device notifier
1478 */
1479
1480int register_inetaddr_notifier(struct notifier_block *nb)
1481{
1482 return blocking_notifier_chain_register(&inetaddr_chain, nb);
1483}
1484EXPORT_SYMBOL(register_inetaddr_notifier);
1485
1486int unregister_inetaddr_notifier(struct notifier_block *nb)
1487{
1488 return blocking_notifier_chain_unregister(&inetaddr_chain, nb);
1489}
1490EXPORT_SYMBOL(unregister_inetaddr_notifier);
1491
1492int register_inetaddr_validator_notifier(struct notifier_block *nb)
1493{
1494 return blocking_notifier_chain_register(&inetaddr_validator_chain, nb);
1495}
1496EXPORT_SYMBOL(register_inetaddr_validator_notifier);
1497
1498int unregister_inetaddr_validator_notifier(struct notifier_block *nb)
1499{
1500 return blocking_notifier_chain_unregister(&inetaddr_validator_chain,
1501 nb);
1502}
1503EXPORT_SYMBOL(unregister_inetaddr_validator_notifier);
1504
1505/* Rename ifa_labels for a device name change. Make some effort to preserve
1506 * existing alias numbering and to create unique labels if possible.
1507*/
1508static void inetdev_changename(struct net_device *dev, struct in_device *in_dev)
1509{
1510 struct in_ifaddr *ifa;
1511 int named = 0;
1512
1513 in_dev_for_each_ifa_rtnl(ifa, in_dev) {
1514 char old[IFNAMSIZ], *dot;
1515
1516 memcpy(old, ifa->ifa_label, IFNAMSIZ);
1517 memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
1518 if (named++ == 0)
1519 goto skip;
1520 dot = strchr(old, ':');
1521 if (!dot) {
1522 sprintf(old, ":%d", named);
1523 dot = old;
1524 }
1525 if (strlen(dot) + strlen(dev->name) < IFNAMSIZ)
1526 strcat(ifa->ifa_label, dot);
1527 else
1528 strcpy(ifa->ifa_label + (IFNAMSIZ - strlen(dot) - 1), dot);
1529skip:
1530 rtmsg_ifa(RTM_NEWADDR, ifa, NULL, 0);
1531 }
1532}
1533
1534static void inetdev_send_gratuitous_arp(struct net_device *dev,
1535 struct in_device *in_dev)
1536
1537{
1538 const struct in_ifaddr *ifa;
1539
1540 in_dev_for_each_ifa_rtnl(ifa, in_dev) {
1541 arp_send(ARPOP_REQUEST, ETH_P_ARP,
1542 ifa->ifa_local, dev,
1543 ifa->ifa_local, NULL,
1544 dev->dev_addr, NULL);
1545 }
1546}
1547
1548/* Called only under RTNL semaphore */
1549
1550static int inetdev_event(struct notifier_block *this, unsigned long event,
1551 void *ptr)
1552{
1553 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1554 struct in_device *in_dev = __in_dev_get_rtnl(dev);
1555
1556 ASSERT_RTNL();
1557
1558 if (!in_dev) {
1559 if (event == NETDEV_REGISTER) {
1560 in_dev = inetdev_init(dev);
1561 if (IS_ERR(in_dev))
1562 return notifier_from_errno(PTR_ERR(in_dev));
1563 if (dev->flags & IFF_LOOPBACK) {
1564 IN_DEV_CONF_SET(in_dev, NOXFRM, 1);
1565 IN_DEV_CONF_SET(in_dev, NOPOLICY, 1);
1566 }
1567 } else if (event == NETDEV_CHANGEMTU) {
1568 /* Re-enabling IP */
1569 if (inetdev_valid_mtu(dev->mtu))
1570 in_dev = inetdev_init(dev);
1571 }
1572 goto out;
1573 }
1574
1575 switch (event) {
1576 case NETDEV_REGISTER:
1577 pr_debug("%s: bug\n", __func__);
1578 RCU_INIT_POINTER(dev->ip_ptr, NULL);
1579 break;
1580 case NETDEV_UP:
1581 if (!inetdev_valid_mtu(dev->mtu))
1582 break;
1583 if (dev->flags & IFF_LOOPBACK) {
1584 struct in_ifaddr *ifa = inet_alloc_ifa();
1585
1586 if (ifa) {
1587 INIT_HLIST_NODE(&ifa->hash);
1588 ifa->ifa_local =
1589 ifa->ifa_address = htonl(INADDR_LOOPBACK);
1590 ifa->ifa_prefixlen = 8;
1591 ifa->ifa_mask = inet_make_mask(8);
1592 in_dev_hold(in_dev);
1593 ifa->ifa_dev = in_dev;
1594 ifa->ifa_scope = RT_SCOPE_HOST;
1595 memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
1596 set_ifa_lifetime(ifa, INFINITY_LIFE_TIME,
1597 INFINITY_LIFE_TIME);
1598 ipv4_devconf_setall(in_dev);
1599 neigh_parms_data_state_setall(in_dev->arp_parms);
1600 inet_insert_ifa(ifa);
1601 }
1602 }
1603 ip_mc_up(in_dev);
1604 fallthrough;
1605 case NETDEV_CHANGEADDR:
1606 if (!IN_DEV_ARP_NOTIFY(in_dev))
1607 break;
1608 fallthrough;
1609 case NETDEV_NOTIFY_PEERS:
1610 /* Send gratuitous ARP to notify of link change */
1611 inetdev_send_gratuitous_arp(dev, in_dev);
1612 break;
1613 case NETDEV_DOWN:
1614 ip_mc_down(in_dev);
1615 break;
1616 case NETDEV_PRE_TYPE_CHANGE:
1617 ip_mc_unmap(in_dev);
1618 break;
1619 case NETDEV_POST_TYPE_CHANGE:
1620 ip_mc_remap(in_dev);
1621 break;
1622 case NETDEV_CHANGEMTU:
1623 if (inetdev_valid_mtu(dev->mtu))
1624 break;
1625 /* disable IP when MTU is not enough */
1626 fallthrough;
1627 case NETDEV_UNREGISTER:
1628 inetdev_destroy(in_dev);
1629 break;
1630 case NETDEV_CHANGENAME:
1631 /* Do not notify about label change, this event is
1632 * not interesting to applications using netlink.
1633 */
1634 inetdev_changename(dev, in_dev);
1635
1636 devinet_sysctl_unregister(in_dev);
1637 devinet_sysctl_register(in_dev);
1638 break;
1639 }
1640out:
1641 return NOTIFY_DONE;
1642}
1643
1644static struct notifier_block ip_netdev_notifier = {
1645 .notifier_call = inetdev_event,
1646};
1647
1648static size_t inet_nlmsg_size(void)
1649{
1650 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
1651 + nla_total_size(4) /* IFA_ADDRESS */
1652 + nla_total_size(4) /* IFA_LOCAL */
1653 + nla_total_size(4) /* IFA_BROADCAST */
1654 + nla_total_size(IFNAMSIZ) /* IFA_LABEL */
1655 + nla_total_size(4) /* IFA_FLAGS */
1656 + nla_total_size(1) /* IFA_PROTO */
1657 + nla_total_size(4) /* IFA_RT_PRIORITY */
1658 + nla_total_size(sizeof(struct ifa_cacheinfo)); /* IFA_CACHEINFO */
1659}
1660
1661static inline u32 cstamp_delta(unsigned long cstamp)
1662{
1663 return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
1664}
1665
1666static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
1667 unsigned long tstamp, u32 preferred, u32 valid)
1668{
1669 struct ifa_cacheinfo ci;
1670
1671 ci.cstamp = cstamp_delta(cstamp);
1672 ci.tstamp = cstamp_delta(tstamp);
1673 ci.ifa_prefered = preferred;
1674 ci.ifa_valid = valid;
1675
1676 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
1677}
1678
1679static int inet_fill_ifaddr(struct sk_buff *skb, const struct in_ifaddr *ifa,
1680 struct inet_fill_args *args)
1681{
1682 struct ifaddrmsg *ifm;
1683 struct nlmsghdr *nlh;
1684 unsigned long tstamp;
1685 u32 preferred, valid;
1686 u32 flags;
1687
1688 nlh = nlmsg_put(skb, args->portid, args->seq, args->event, sizeof(*ifm),
1689 args->flags);
1690 if (!nlh)
1691 return -EMSGSIZE;
1692
1693 ifm = nlmsg_data(nlh);
1694 ifm->ifa_family = AF_INET;
1695 ifm->ifa_prefixlen = ifa->ifa_prefixlen;
1696
1697 flags = READ_ONCE(ifa->ifa_flags);
1698 /* Warning : ifm->ifa_flags is an __u8, it holds only 8 bits.
1699 * The 32bit value is given in IFA_FLAGS attribute.
1700 */
1701 ifm->ifa_flags = (__u8)flags;
1702
1703 ifm->ifa_scope = ifa->ifa_scope;
1704 ifm->ifa_index = ifa->ifa_dev->dev->ifindex;
1705
1706 if (args->netnsid >= 0 &&
1707 nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
1708 goto nla_put_failure;
1709
1710 tstamp = READ_ONCE(ifa->ifa_tstamp);
1711 if (!(flags & IFA_F_PERMANENT)) {
1712 preferred = READ_ONCE(ifa->ifa_preferred_lft);
1713 valid = READ_ONCE(ifa->ifa_valid_lft);
1714 if (preferred != INFINITY_LIFE_TIME) {
1715 long tval = (jiffies - tstamp) / HZ;
1716
1717 if (preferred > tval)
1718 preferred -= tval;
1719 else
1720 preferred = 0;
1721 if (valid != INFINITY_LIFE_TIME) {
1722 if (valid > tval)
1723 valid -= tval;
1724 else
1725 valid = 0;
1726 }
1727 }
1728 } else {
1729 preferred = INFINITY_LIFE_TIME;
1730 valid = INFINITY_LIFE_TIME;
1731 }
1732 if ((ifa->ifa_address &&
1733 nla_put_in_addr(skb, IFA_ADDRESS, ifa->ifa_address)) ||
1734 (ifa->ifa_local &&
1735 nla_put_in_addr(skb, IFA_LOCAL, ifa->ifa_local)) ||
1736 (ifa->ifa_broadcast &&
1737 nla_put_in_addr(skb, IFA_BROADCAST, ifa->ifa_broadcast)) ||
1738 (ifa->ifa_label[0] &&
1739 nla_put_string(skb, IFA_LABEL, ifa->ifa_label)) ||
1740 (ifa->ifa_proto &&
1741 nla_put_u8(skb, IFA_PROTO, ifa->ifa_proto)) ||
1742 nla_put_u32(skb, IFA_FLAGS, flags) ||
1743 (ifa->ifa_rt_priority &&
1744 nla_put_u32(skb, IFA_RT_PRIORITY, ifa->ifa_rt_priority)) ||
1745 put_cacheinfo(skb, READ_ONCE(ifa->ifa_cstamp), tstamp,
1746 preferred, valid))
1747 goto nla_put_failure;
1748
1749 nlmsg_end(skb, nlh);
1750 return 0;
1751
1752nla_put_failure:
1753 nlmsg_cancel(skb, nlh);
1754 return -EMSGSIZE;
1755}
1756
1757static int inet_valid_dump_ifaddr_req(const struct nlmsghdr *nlh,
1758 struct inet_fill_args *fillargs,
1759 struct net **tgt_net, struct sock *sk,
1760 struct netlink_callback *cb)
1761{
1762 struct netlink_ext_ack *extack = cb->extack;
1763 struct nlattr *tb[IFA_MAX+1];
1764 struct ifaddrmsg *ifm;
1765 int err, i;
1766
1767 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
1768 NL_SET_ERR_MSG(extack, "ipv4: Invalid header for address dump request");
1769 return -EINVAL;
1770 }
1771
1772 ifm = nlmsg_data(nlh);
1773 if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
1774 NL_SET_ERR_MSG(extack, "ipv4: Invalid values in header for address dump request");
1775 return -EINVAL;
1776 }
1777
1778 fillargs->ifindex = ifm->ifa_index;
1779 if (fillargs->ifindex) {
1780 cb->answer_flags |= NLM_F_DUMP_FILTERED;
1781 fillargs->flags |= NLM_F_DUMP_FILTERED;
1782 }
1783
1784 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
1785 ifa_ipv4_policy, extack);
1786 if (err < 0)
1787 return err;
1788
1789 for (i = 0; i <= IFA_MAX; ++i) {
1790 if (!tb[i])
1791 continue;
1792
1793 if (i == IFA_TARGET_NETNSID) {
1794 struct net *net;
1795
1796 fillargs->netnsid = nla_get_s32(tb[i]);
1797
1798 net = rtnl_get_net_ns_capable(sk, fillargs->netnsid);
1799 if (IS_ERR(net)) {
1800 fillargs->netnsid = -1;
1801 NL_SET_ERR_MSG(extack, "ipv4: Invalid target network namespace id");
1802 return PTR_ERR(net);
1803 }
1804 *tgt_net = net;
1805 } else {
1806 NL_SET_ERR_MSG(extack, "ipv4: Unsupported attribute in dump request");
1807 return -EINVAL;
1808 }
1809 }
1810
1811 return 0;
1812}
1813
1814static int in_dev_dump_addr(struct in_device *in_dev, struct sk_buff *skb,
1815 struct netlink_callback *cb, int *s_ip_idx,
1816 struct inet_fill_args *fillargs)
1817{
1818 struct in_ifaddr *ifa;
1819 int ip_idx = 0;
1820 int err;
1821
1822 in_dev_for_each_ifa_rcu(ifa, in_dev) {
1823 if (ip_idx < *s_ip_idx) {
1824 ip_idx++;
1825 continue;
1826 }
1827 err = inet_fill_ifaddr(skb, ifa, fillargs);
1828 if (err < 0)
1829 goto done;
1830
1831 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1832 ip_idx++;
1833 }
1834 err = 0;
1835 ip_idx = 0;
1836done:
1837 *s_ip_idx = ip_idx;
1838
1839 return err;
1840}
1841
1842/* Combine dev_addr_genid and dev_base_seq to detect changes.
1843 */
1844static u32 inet_base_seq(const struct net *net)
1845{
1846 u32 res = atomic_read(&net->ipv4.dev_addr_genid) +
1847 READ_ONCE(net->dev_base_seq);
1848
1849 /* Must not return 0 (see nl_dump_check_consistent()).
1850 * Chose a value far away from 0.
1851 */
1852 if (!res)
1853 res = 0x80000000;
1854 return res;
1855}
1856
1857static int inet_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
1858{
1859 const struct nlmsghdr *nlh = cb->nlh;
1860 struct inet_fill_args fillargs = {
1861 .portid = NETLINK_CB(cb->skb).portid,
1862 .seq = nlh->nlmsg_seq,
1863 .event = RTM_NEWADDR,
1864 .flags = NLM_F_MULTI,
1865 .netnsid = -1,
1866 };
1867 struct net *net = sock_net(skb->sk);
1868 struct net *tgt_net = net;
1869 struct {
1870 unsigned long ifindex;
1871 int ip_idx;
1872 } *ctx = (void *)cb->ctx;
1873 struct in_device *in_dev;
1874 struct net_device *dev;
1875 int err = 0;
1876
1877 rcu_read_lock();
1878 if (cb->strict_check) {
1879 err = inet_valid_dump_ifaddr_req(nlh, &fillargs, &tgt_net,
1880 skb->sk, cb);
1881 if (err < 0)
1882 goto done;
1883
1884 if (fillargs.ifindex) {
1885 dev = dev_get_by_index_rcu(tgt_net, fillargs.ifindex);
1886 if (!dev) {
1887 err = -ENODEV;
1888 goto done;
1889 }
1890 in_dev = __in_dev_get_rcu(dev);
1891 if (!in_dev)
1892 goto done;
1893 err = in_dev_dump_addr(in_dev, skb, cb, &ctx->ip_idx,
1894 &fillargs);
1895 goto done;
1896 }
1897 }
1898
1899 cb->seq = inet_base_seq(tgt_net);
1900
1901 for_each_netdev_dump(tgt_net, dev, ctx->ifindex) {
1902 in_dev = __in_dev_get_rcu(dev);
1903 if (!in_dev)
1904 continue;
1905 err = in_dev_dump_addr(in_dev, skb, cb, &ctx->ip_idx,
1906 &fillargs);
1907 if (err < 0)
1908 goto done;
1909 }
1910done:
1911 if (fillargs.netnsid >= 0)
1912 put_net(tgt_net);
1913 rcu_read_unlock();
1914 return err;
1915}
1916
1917static void rtmsg_ifa(int event, struct in_ifaddr *ifa, struct nlmsghdr *nlh,
1918 u32 portid)
1919{
1920 struct inet_fill_args fillargs = {
1921 .portid = portid,
1922 .seq = nlh ? nlh->nlmsg_seq : 0,
1923 .event = event,
1924 .flags = 0,
1925 .netnsid = -1,
1926 };
1927 struct sk_buff *skb;
1928 int err = -ENOBUFS;
1929 struct net *net;
1930
1931 net = dev_net(ifa->ifa_dev->dev);
1932 skb = nlmsg_new(inet_nlmsg_size(), GFP_KERNEL);
1933 if (!skb)
1934 goto errout;
1935
1936 err = inet_fill_ifaddr(skb, ifa, &fillargs);
1937 if (err < 0) {
1938 /* -EMSGSIZE implies BUG in inet_nlmsg_size() */
1939 WARN_ON(err == -EMSGSIZE);
1940 kfree_skb(skb);
1941 goto errout;
1942 }
1943 rtnl_notify(skb, net, portid, RTNLGRP_IPV4_IFADDR, nlh, GFP_KERNEL);
1944 return;
1945errout:
1946 if (err < 0)
1947 rtnl_set_sk_err(net, RTNLGRP_IPV4_IFADDR, err);
1948}
1949
1950static size_t inet_get_link_af_size(const struct net_device *dev,
1951 u32 ext_filter_mask)
1952{
1953 struct in_device *in_dev = rcu_dereference_rtnl(dev->ip_ptr);
1954
1955 if (!in_dev)
1956 return 0;
1957
1958 return nla_total_size(IPV4_DEVCONF_MAX * 4); /* IFLA_INET_CONF */
1959}
1960
1961static int inet_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
1962 u32 ext_filter_mask)
1963{
1964 struct in_device *in_dev = rcu_dereference_rtnl(dev->ip_ptr);
1965 struct nlattr *nla;
1966 int i;
1967
1968 if (!in_dev)
1969 return -ENODATA;
1970
1971 nla = nla_reserve(skb, IFLA_INET_CONF, IPV4_DEVCONF_MAX * 4);
1972 if (!nla)
1973 return -EMSGSIZE;
1974
1975 for (i = 0; i < IPV4_DEVCONF_MAX; i++)
1976 ((u32 *) nla_data(nla))[i] = READ_ONCE(in_dev->cnf.data[i]);
1977
1978 return 0;
1979}
1980
1981static const struct nla_policy inet_af_policy[IFLA_INET_MAX+1] = {
1982 [IFLA_INET_CONF] = { .type = NLA_NESTED },
1983};
1984
1985static int inet_validate_link_af(const struct net_device *dev,
1986 const struct nlattr *nla,
1987 struct netlink_ext_ack *extack)
1988{
1989 struct nlattr *a, *tb[IFLA_INET_MAX+1];
1990 int err, rem;
1991
1992 if (dev && !__in_dev_get_rtnl(dev))
1993 return -EAFNOSUPPORT;
1994
1995 err = nla_parse_nested_deprecated(tb, IFLA_INET_MAX, nla,
1996 inet_af_policy, extack);
1997 if (err < 0)
1998 return err;
1999
2000 if (tb[IFLA_INET_CONF]) {
2001 nla_for_each_nested(a, tb[IFLA_INET_CONF], rem) {
2002 int cfgid = nla_type(a);
2003
2004 if (nla_len(a) < 4)
2005 return -EINVAL;
2006
2007 if (cfgid <= 0 || cfgid > IPV4_DEVCONF_MAX)
2008 return -EINVAL;
2009 }
2010 }
2011
2012 return 0;
2013}
2014
2015static int inet_set_link_af(struct net_device *dev, const struct nlattr *nla,
2016 struct netlink_ext_ack *extack)
2017{
2018 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2019 struct nlattr *a, *tb[IFLA_INET_MAX+1];
2020 int rem;
2021
2022 if (!in_dev)
2023 return -EAFNOSUPPORT;
2024
2025 if (nla_parse_nested_deprecated(tb, IFLA_INET_MAX, nla, NULL, NULL) < 0)
2026 return -EINVAL;
2027
2028 if (tb[IFLA_INET_CONF]) {
2029 nla_for_each_nested(a, tb[IFLA_INET_CONF], rem)
2030 ipv4_devconf_set(in_dev, nla_type(a), nla_get_u32(a));
2031 }
2032
2033 return 0;
2034}
2035
2036static int inet_netconf_msgsize_devconf(int type)
2037{
2038 int size = NLMSG_ALIGN(sizeof(struct netconfmsg))
2039 + nla_total_size(4); /* NETCONFA_IFINDEX */
2040 bool all = false;
2041
2042 if (type == NETCONFA_ALL)
2043 all = true;
2044
2045 if (all || type == NETCONFA_FORWARDING)
2046 size += nla_total_size(4);
2047 if (all || type == NETCONFA_RP_FILTER)
2048 size += nla_total_size(4);
2049 if (all || type == NETCONFA_MC_FORWARDING)
2050 size += nla_total_size(4);
2051 if (all || type == NETCONFA_BC_FORWARDING)
2052 size += nla_total_size(4);
2053 if (all || type == NETCONFA_PROXY_NEIGH)
2054 size += nla_total_size(4);
2055 if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
2056 size += nla_total_size(4);
2057
2058 return size;
2059}
2060
2061static int inet_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
2062 const struct ipv4_devconf *devconf,
2063 u32 portid, u32 seq, int event,
2064 unsigned int flags, int type)
2065{
2066 struct nlmsghdr *nlh;
2067 struct netconfmsg *ncm;
2068 bool all = false;
2069
2070 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
2071 flags);
2072 if (!nlh)
2073 return -EMSGSIZE;
2074
2075 if (type == NETCONFA_ALL)
2076 all = true;
2077
2078 ncm = nlmsg_data(nlh);
2079 ncm->ncm_family = AF_INET;
2080
2081 if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
2082 goto nla_put_failure;
2083
2084 if (!devconf)
2085 goto out;
2086
2087 if ((all || type == NETCONFA_FORWARDING) &&
2088 nla_put_s32(skb, NETCONFA_FORWARDING,
2089 IPV4_DEVCONF_RO(*devconf, FORWARDING)) < 0)
2090 goto nla_put_failure;
2091 if ((all || type == NETCONFA_RP_FILTER) &&
2092 nla_put_s32(skb, NETCONFA_RP_FILTER,
2093 IPV4_DEVCONF_RO(*devconf, RP_FILTER)) < 0)
2094 goto nla_put_failure;
2095 if ((all || type == NETCONFA_MC_FORWARDING) &&
2096 nla_put_s32(skb, NETCONFA_MC_FORWARDING,
2097 IPV4_DEVCONF_RO(*devconf, MC_FORWARDING)) < 0)
2098 goto nla_put_failure;
2099 if ((all || type == NETCONFA_BC_FORWARDING) &&
2100 nla_put_s32(skb, NETCONFA_BC_FORWARDING,
2101 IPV4_DEVCONF_RO(*devconf, BC_FORWARDING)) < 0)
2102 goto nla_put_failure;
2103 if ((all || type == NETCONFA_PROXY_NEIGH) &&
2104 nla_put_s32(skb, NETCONFA_PROXY_NEIGH,
2105 IPV4_DEVCONF_RO(*devconf, PROXY_ARP)) < 0)
2106 goto nla_put_failure;
2107 if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
2108 nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
2109 IPV4_DEVCONF_RO(*devconf,
2110 IGNORE_ROUTES_WITH_LINKDOWN)) < 0)
2111 goto nla_put_failure;
2112
2113out:
2114 nlmsg_end(skb, nlh);
2115 return 0;
2116
2117nla_put_failure:
2118 nlmsg_cancel(skb, nlh);
2119 return -EMSGSIZE;
2120}
2121
2122void inet_netconf_notify_devconf(struct net *net, int event, int type,
2123 int ifindex, struct ipv4_devconf *devconf)
2124{
2125 struct sk_buff *skb;
2126 int err = -ENOBUFS;
2127
2128 skb = nlmsg_new(inet_netconf_msgsize_devconf(type), GFP_KERNEL);
2129 if (!skb)
2130 goto errout;
2131
2132 err = inet_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
2133 event, 0, type);
2134 if (err < 0) {
2135 /* -EMSGSIZE implies BUG in inet_netconf_msgsize_devconf() */
2136 WARN_ON(err == -EMSGSIZE);
2137 kfree_skb(skb);
2138 goto errout;
2139 }
2140 rtnl_notify(skb, net, 0, RTNLGRP_IPV4_NETCONF, NULL, GFP_KERNEL);
2141 return;
2142errout:
2143 if (err < 0)
2144 rtnl_set_sk_err(net, RTNLGRP_IPV4_NETCONF, err);
2145}
2146
2147static const struct nla_policy devconf_ipv4_policy[NETCONFA_MAX+1] = {
2148 [NETCONFA_IFINDEX] = { .len = sizeof(int) },
2149 [NETCONFA_FORWARDING] = { .len = sizeof(int) },
2150 [NETCONFA_RP_FILTER] = { .len = sizeof(int) },
2151 [NETCONFA_PROXY_NEIGH] = { .len = sizeof(int) },
2152 [NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN] = { .len = sizeof(int) },
2153};
2154
2155static int inet_netconf_valid_get_req(struct sk_buff *skb,
2156 const struct nlmsghdr *nlh,
2157 struct nlattr **tb,
2158 struct netlink_ext_ack *extack)
2159{
2160 int i, err;
2161
2162 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(struct netconfmsg))) {
2163 NL_SET_ERR_MSG(extack, "ipv4: Invalid header for netconf get request");
2164 return -EINVAL;
2165 }
2166
2167 if (!netlink_strict_get_check(skb))
2168 return nlmsg_parse_deprecated(nlh, sizeof(struct netconfmsg),
2169 tb, NETCONFA_MAX,
2170 devconf_ipv4_policy, extack);
2171
2172 err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct netconfmsg),
2173 tb, NETCONFA_MAX,
2174 devconf_ipv4_policy, extack);
2175 if (err)
2176 return err;
2177
2178 for (i = 0; i <= NETCONFA_MAX; i++) {
2179 if (!tb[i])
2180 continue;
2181
2182 switch (i) {
2183 case NETCONFA_IFINDEX:
2184 break;
2185 default:
2186 NL_SET_ERR_MSG(extack, "ipv4: Unsupported attribute in netconf get request");
2187 return -EINVAL;
2188 }
2189 }
2190
2191 return 0;
2192}
2193
2194static int inet_netconf_get_devconf(struct sk_buff *in_skb,
2195 struct nlmsghdr *nlh,
2196 struct netlink_ext_ack *extack)
2197{
2198 struct net *net = sock_net(in_skb->sk);
2199 struct nlattr *tb[NETCONFA_MAX + 1];
2200 const struct ipv4_devconf *devconf;
2201 struct in_device *in_dev = NULL;
2202 struct net_device *dev = NULL;
2203 struct sk_buff *skb;
2204 int ifindex;
2205 int err;
2206
2207 err = inet_netconf_valid_get_req(in_skb, nlh, tb, extack);
2208 if (err)
2209 return err;
2210
2211 if (!tb[NETCONFA_IFINDEX])
2212 return -EINVAL;
2213
2214 ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
2215 switch (ifindex) {
2216 case NETCONFA_IFINDEX_ALL:
2217 devconf = net->ipv4.devconf_all;
2218 break;
2219 case NETCONFA_IFINDEX_DEFAULT:
2220 devconf = net->ipv4.devconf_dflt;
2221 break;
2222 default:
2223 err = -ENODEV;
2224 dev = dev_get_by_index(net, ifindex);
2225 if (dev)
2226 in_dev = in_dev_get(dev);
2227 if (!in_dev)
2228 goto errout;
2229 devconf = &in_dev->cnf;
2230 break;
2231 }
2232
2233 err = -ENOBUFS;
2234 skb = nlmsg_new(inet_netconf_msgsize_devconf(NETCONFA_ALL), GFP_KERNEL);
2235 if (!skb)
2236 goto errout;
2237
2238 err = inet_netconf_fill_devconf(skb, ifindex, devconf,
2239 NETLINK_CB(in_skb).portid,
2240 nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
2241 NETCONFA_ALL);
2242 if (err < 0) {
2243 /* -EMSGSIZE implies BUG in inet_netconf_msgsize_devconf() */
2244 WARN_ON(err == -EMSGSIZE);
2245 kfree_skb(skb);
2246 goto errout;
2247 }
2248 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
2249errout:
2250 if (in_dev)
2251 in_dev_put(in_dev);
2252 dev_put(dev);
2253 return err;
2254}
2255
2256static int inet_netconf_dump_devconf(struct sk_buff *skb,
2257 struct netlink_callback *cb)
2258{
2259 const struct nlmsghdr *nlh = cb->nlh;
2260 struct net *net = sock_net(skb->sk);
2261 struct {
2262 unsigned long ifindex;
2263 unsigned int all_default;
2264 } *ctx = (void *)cb->ctx;
2265 const struct in_device *in_dev;
2266 struct net_device *dev;
2267 int err = 0;
2268
2269 if (cb->strict_check) {
2270 struct netlink_ext_ack *extack = cb->extack;
2271 struct netconfmsg *ncm;
2272
2273 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ncm))) {
2274 NL_SET_ERR_MSG(extack, "ipv4: Invalid header for netconf dump request");
2275 return -EINVAL;
2276 }
2277
2278 if (nlmsg_attrlen(nlh, sizeof(*ncm))) {
2279 NL_SET_ERR_MSG(extack, "ipv4: Invalid data after header in netconf dump request");
2280 return -EINVAL;
2281 }
2282 }
2283
2284 rcu_read_lock();
2285 for_each_netdev_dump(net, dev, ctx->ifindex) {
2286 in_dev = __in_dev_get_rcu(dev);
2287 if (!in_dev)
2288 continue;
2289 err = inet_netconf_fill_devconf(skb, dev->ifindex,
2290 &in_dev->cnf,
2291 NETLINK_CB(cb->skb).portid,
2292 nlh->nlmsg_seq,
2293 RTM_NEWNETCONF, NLM_F_MULTI,
2294 NETCONFA_ALL);
2295 if (err < 0)
2296 goto done;
2297 }
2298 if (ctx->all_default == 0) {
2299 err = inet_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
2300 net->ipv4.devconf_all,
2301 NETLINK_CB(cb->skb).portid,
2302 nlh->nlmsg_seq,
2303 RTM_NEWNETCONF, NLM_F_MULTI,
2304 NETCONFA_ALL);
2305 if (err < 0)
2306 goto done;
2307 ctx->all_default++;
2308 }
2309 if (ctx->all_default == 1) {
2310 err = inet_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
2311 net->ipv4.devconf_dflt,
2312 NETLINK_CB(cb->skb).portid,
2313 nlh->nlmsg_seq,
2314 RTM_NEWNETCONF, NLM_F_MULTI,
2315 NETCONFA_ALL);
2316 if (err < 0)
2317 goto done;
2318 ctx->all_default++;
2319 }
2320done:
2321 rcu_read_unlock();
2322 return err;
2323}
2324
2325#ifdef CONFIG_SYSCTL
2326
2327static void devinet_copy_dflt_conf(struct net *net, int i)
2328{
2329 struct net_device *dev;
2330
2331 rcu_read_lock();
2332 for_each_netdev_rcu(net, dev) {
2333 struct in_device *in_dev;
2334
2335 in_dev = __in_dev_get_rcu(dev);
2336 if (in_dev && !test_bit(i, in_dev->cnf.state))
2337 in_dev->cnf.data[i] = net->ipv4.devconf_dflt->data[i];
2338 }
2339 rcu_read_unlock();
2340}
2341
2342/* called with RTNL locked */
2343static void inet_forward_change(struct net *net)
2344{
2345 struct net_device *dev;
2346 int on = IPV4_DEVCONF_ALL(net, FORWARDING);
2347
2348 IPV4_DEVCONF_ALL(net, ACCEPT_REDIRECTS) = !on;
2349 IPV4_DEVCONF_DFLT(net, FORWARDING) = on;
2350 inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
2351 NETCONFA_FORWARDING,
2352 NETCONFA_IFINDEX_ALL,
2353 net->ipv4.devconf_all);
2354 inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
2355 NETCONFA_FORWARDING,
2356 NETCONFA_IFINDEX_DEFAULT,
2357 net->ipv4.devconf_dflt);
2358
2359 for_each_netdev(net, dev) {
2360 struct in_device *in_dev;
2361
2362 if (on)
2363 dev_disable_lro(dev);
2364
2365 in_dev = __in_dev_get_rtnl(dev);
2366 if (in_dev) {
2367 IN_DEV_CONF_SET(in_dev, FORWARDING, on);
2368 inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
2369 NETCONFA_FORWARDING,
2370 dev->ifindex, &in_dev->cnf);
2371 }
2372 }
2373}
2374
2375static int devinet_conf_ifindex(struct net *net, struct ipv4_devconf *cnf)
2376{
2377 if (cnf == net->ipv4.devconf_dflt)
2378 return NETCONFA_IFINDEX_DEFAULT;
2379 else if (cnf == net->ipv4.devconf_all)
2380 return NETCONFA_IFINDEX_ALL;
2381 else {
2382 struct in_device *idev
2383 = container_of(cnf, struct in_device, cnf);
2384 return idev->dev->ifindex;
2385 }
2386}
2387
2388static int devinet_conf_proc(struct ctl_table *ctl, int write,
2389 void *buffer, size_t *lenp, loff_t *ppos)
2390{
2391 int old_value = *(int *)ctl->data;
2392 int ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
2393 int new_value = *(int *)ctl->data;
2394
2395 if (write) {
2396 struct ipv4_devconf *cnf = ctl->extra1;
2397 struct net *net = ctl->extra2;
2398 int i = (int *)ctl->data - cnf->data;
2399 int ifindex;
2400
2401 set_bit(i, cnf->state);
2402
2403 if (cnf == net->ipv4.devconf_dflt)
2404 devinet_copy_dflt_conf(net, i);
2405 if (i == IPV4_DEVCONF_ACCEPT_LOCAL - 1 ||
2406 i == IPV4_DEVCONF_ROUTE_LOCALNET - 1)
2407 if ((new_value == 0) && (old_value != 0))
2408 rt_cache_flush(net);
2409
2410 if (i == IPV4_DEVCONF_BC_FORWARDING - 1 &&
2411 new_value != old_value)
2412 rt_cache_flush(net);
2413
2414 if (i == IPV4_DEVCONF_RP_FILTER - 1 &&
2415 new_value != old_value) {
2416 ifindex = devinet_conf_ifindex(net, cnf);
2417 inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
2418 NETCONFA_RP_FILTER,
2419 ifindex, cnf);
2420 }
2421 if (i == IPV4_DEVCONF_PROXY_ARP - 1 &&
2422 new_value != old_value) {
2423 ifindex = devinet_conf_ifindex(net, cnf);
2424 inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
2425 NETCONFA_PROXY_NEIGH,
2426 ifindex, cnf);
2427 }
2428 if (i == IPV4_DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN - 1 &&
2429 new_value != old_value) {
2430 ifindex = devinet_conf_ifindex(net, cnf);
2431 inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
2432 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
2433 ifindex, cnf);
2434 }
2435 }
2436
2437 return ret;
2438}
2439
2440static int devinet_sysctl_forward(struct ctl_table *ctl, int write,
2441 void *buffer, size_t *lenp, loff_t *ppos)
2442{
2443 int *valp = ctl->data;
2444 int val = *valp;
2445 loff_t pos = *ppos;
2446 struct net *net = ctl->extra2;
2447 int ret;
2448
2449 if (write && !ns_capable(net->user_ns, CAP_NET_ADMIN))
2450 return -EPERM;
2451
2452 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
2453
2454 if (write && *valp != val) {
2455 if (valp != &IPV4_DEVCONF_DFLT(net, FORWARDING)) {
2456 if (!rtnl_trylock()) {
2457 /* Restore the original values before restarting */
2458 *valp = val;
2459 *ppos = pos;
2460 return restart_syscall();
2461 }
2462 if (valp == &IPV4_DEVCONF_ALL(net, FORWARDING)) {
2463 inet_forward_change(net);
2464 } else {
2465 struct ipv4_devconf *cnf = ctl->extra1;
2466 struct in_device *idev =
2467 container_of(cnf, struct in_device, cnf);
2468 if (*valp)
2469 dev_disable_lro(idev->dev);
2470 inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
2471 NETCONFA_FORWARDING,
2472 idev->dev->ifindex,
2473 cnf);
2474 }
2475 rtnl_unlock();
2476 rt_cache_flush(net);
2477 } else
2478 inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
2479 NETCONFA_FORWARDING,
2480 NETCONFA_IFINDEX_DEFAULT,
2481 net->ipv4.devconf_dflt);
2482 }
2483
2484 return ret;
2485}
2486
2487static int ipv4_doint_and_flush(struct ctl_table *ctl, int write,
2488 void *buffer, size_t *lenp, loff_t *ppos)
2489{
2490 int *valp = ctl->data;
2491 int val = *valp;
2492 int ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
2493 struct net *net = ctl->extra2;
2494
2495 if (write && *valp != val)
2496 rt_cache_flush(net);
2497
2498 return ret;
2499}
2500
2501#define DEVINET_SYSCTL_ENTRY(attr, name, mval, proc) \
2502 { \
2503 .procname = name, \
2504 .data = ipv4_devconf.data + \
2505 IPV4_DEVCONF_ ## attr - 1, \
2506 .maxlen = sizeof(int), \
2507 .mode = mval, \
2508 .proc_handler = proc, \
2509 .extra1 = &ipv4_devconf, \
2510 }
2511
2512#define DEVINET_SYSCTL_RW_ENTRY(attr, name) \
2513 DEVINET_SYSCTL_ENTRY(attr, name, 0644, devinet_conf_proc)
2514
2515#define DEVINET_SYSCTL_RO_ENTRY(attr, name) \
2516 DEVINET_SYSCTL_ENTRY(attr, name, 0444, devinet_conf_proc)
2517
2518#define DEVINET_SYSCTL_COMPLEX_ENTRY(attr, name, proc) \
2519 DEVINET_SYSCTL_ENTRY(attr, name, 0644, proc)
2520
2521#define DEVINET_SYSCTL_FLUSHING_ENTRY(attr, name) \
2522 DEVINET_SYSCTL_COMPLEX_ENTRY(attr, name, ipv4_doint_and_flush)
2523
2524static struct devinet_sysctl_table {
2525 struct ctl_table_header *sysctl_header;
2526 struct ctl_table devinet_vars[__IPV4_DEVCONF_MAX];
2527} devinet_sysctl = {
2528 .devinet_vars = {
2529 DEVINET_SYSCTL_COMPLEX_ENTRY(FORWARDING, "forwarding",
2530 devinet_sysctl_forward),
2531 DEVINET_SYSCTL_RO_ENTRY(MC_FORWARDING, "mc_forwarding"),
2532 DEVINET_SYSCTL_RW_ENTRY(BC_FORWARDING, "bc_forwarding"),
2533
2534 DEVINET_SYSCTL_RW_ENTRY(ACCEPT_REDIRECTS, "accept_redirects"),
2535 DEVINET_SYSCTL_RW_ENTRY(SECURE_REDIRECTS, "secure_redirects"),
2536 DEVINET_SYSCTL_RW_ENTRY(SHARED_MEDIA, "shared_media"),
2537 DEVINET_SYSCTL_RW_ENTRY(RP_FILTER, "rp_filter"),
2538 DEVINET_SYSCTL_RW_ENTRY(SEND_REDIRECTS, "send_redirects"),
2539 DEVINET_SYSCTL_RW_ENTRY(ACCEPT_SOURCE_ROUTE,
2540 "accept_source_route"),
2541 DEVINET_SYSCTL_RW_ENTRY(ACCEPT_LOCAL, "accept_local"),
2542 DEVINET_SYSCTL_RW_ENTRY(SRC_VMARK, "src_valid_mark"),
2543 DEVINET_SYSCTL_RW_ENTRY(PROXY_ARP, "proxy_arp"),
2544 DEVINET_SYSCTL_RW_ENTRY(MEDIUM_ID, "medium_id"),
2545 DEVINET_SYSCTL_RW_ENTRY(BOOTP_RELAY, "bootp_relay"),
2546 DEVINET_SYSCTL_RW_ENTRY(LOG_MARTIANS, "log_martians"),
2547 DEVINET_SYSCTL_RW_ENTRY(TAG, "tag"),
2548 DEVINET_SYSCTL_RW_ENTRY(ARPFILTER, "arp_filter"),
2549 DEVINET_SYSCTL_RW_ENTRY(ARP_ANNOUNCE, "arp_announce"),
2550 DEVINET_SYSCTL_RW_ENTRY(ARP_IGNORE, "arp_ignore"),
2551 DEVINET_SYSCTL_RW_ENTRY(ARP_ACCEPT, "arp_accept"),
2552 DEVINET_SYSCTL_RW_ENTRY(ARP_NOTIFY, "arp_notify"),
2553 DEVINET_SYSCTL_RW_ENTRY(ARP_EVICT_NOCARRIER,
2554 "arp_evict_nocarrier"),
2555 DEVINET_SYSCTL_RW_ENTRY(PROXY_ARP_PVLAN, "proxy_arp_pvlan"),
2556 DEVINET_SYSCTL_RW_ENTRY(FORCE_IGMP_VERSION,
2557 "force_igmp_version"),
2558 DEVINET_SYSCTL_RW_ENTRY(IGMPV2_UNSOLICITED_REPORT_INTERVAL,
2559 "igmpv2_unsolicited_report_interval"),
2560 DEVINET_SYSCTL_RW_ENTRY(IGMPV3_UNSOLICITED_REPORT_INTERVAL,
2561 "igmpv3_unsolicited_report_interval"),
2562 DEVINET_SYSCTL_RW_ENTRY(IGNORE_ROUTES_WITH_LINKDOWN,
2563 "ignore_routes_with_linkdown"),
2564 DEVINET_SYSCTL_RW_ENTRY(DROP_GRATUITOUS_ARP,
2565 "drop_gratuitous_arp"),
2566
2567 DEVINET_SYSCTL_FLUSHING_ENTRY(NOXFRM, "disable_xfrm"),
2568 DEVINET_SYSCTL_FLUSHING_ENTRY(NOPOLICY, "disable_policy"),
2569 DEVINET_SYSCTL_FLUSHING_ENTRY(PROMOTE_SECONDARIES,
2570 "promote_secondaries"),
2571 DEVINET_SYSCTL_FLUSHING_ENTRY(ROUTE_LOCALNET,
2572 "route_localnet"),
2573 DEVINET_SYSCTL_FLUSHING_ENTRY(DROP_UNICAST_IN_L2_MULTICAST,
2574 "drop_unicast_in_l2_multicast"),
2575 },
2576};
2577
2578static int __devinet_sysctl_register(struct net *net, char *dev_name,
2579 int ifindex, struct ipv4_devconf *p)
2580{
2581 int i;
2582 struct devinet_sysctl_table *t;
2583 char path[sizeof("net/ipv4/conf/") + IFNAMSIZ];
2584
2585 t = kmemdup(&devinet_sysctl, sizeof(*t), GFP_KERNEL_ACCOUNT);
2586 if (!t)
2587 goto out;
2588
2589 for (i = 0; i < ARRAY_SIZE(t->devinet_vars) - 1; i++) {
2590 t->devinet_vars[i].data += (char *)p - (char *)&ipv4_devconf;
2591 t->devinet_vars[i].extra1 = p;
2592 t->devinet_vars[i].extra2 = net;
2593 }
2594
2595 snprintf(path, sizeof(path), "net/ipv4/conf/%s", dev_name);
2596
2597 t->sysctl_header = register_net_sysctl(net, path, t->devinet_vars);
2598 if (!t->sysctl_header)
2599 goto free;
2600
2601 p->sysctl = t;
2602
2603 inet_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
2604 ifindex, p);
2605 return 0;
2606
2607free:
2608 kfree(t);
2609out:
2610 return -ENOMEM;
2611}
2612
2613static void __devinet_sysctl_unregister(struct net *net,
2614 struct ipv4_devconf *cnf, int ifindex)
2615{
2616 struct devinet_sysctl_table *t = cnf->sysctl;
2617
2618 if (t) {
2619 cnf->sysctl = NULL;
2620 unregister_net_sysctl_table(t->sysctl_header);
2621 kfree(t);
2622 }
2623
2624 inet_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
2625}
2626
2627static int devinet_sysctl_register(struct in_device *idev)
2628{
2629 int err;
2630
2631 if (!sysctl_dev_name_is_allowed(idev->dev->name))
2632 return -EINVAL;
2633
2634 err = neigh_sysctl_register(idev->dev, idev->arp_parms, NULL);
2635 if (err)
2636 return err;
2637 err = __devinet_sysctl_register(dev_net(idev->dev), idev->dev->name,
2638 idev->dev->ifindex, &idev->cnf);
2639 if (err)
2640 neigh_sysctl_unregister(idev->arp_parms);
2641 return err;
2642}
2643
2644static void devinet_sysctl_unregister(struct in_device *idev)
2645{
2646 struct net *net = dev_net(idev->dev);
2647
2648 __devinet_sysctl_unregister(net, &idev->cnf, idev->dev->ifindex);
2649 neigh_sysctl_unregister(idev->arp_parms);
2650}
2651
2652static struct ctl_table ctl_forward_entry[] = {
2653 {
2654 .procname = "ip_forward",
2655 .data = &ipv4_devconf.data[
2656 IPV4_DEVCONF_FORWARDING - 1],
2657 .maxlen = sizeof(int),
2658 .mode = 0644,
2659 .proc_handler = devinet_sysctl_forward,
2660 .extra1 = &ipv4_devconf,
2661 .extra2 = &init_net,
2662 },
2663 { },
2664};
2665#endif
2666
2667static __net_init int devinet_init_net(struct net *net)
2668{
2669 int err;
2670 struct ipv4_devconf *all, *dflt;
2671#ifdef CONFIG_SYSCTL
2672 struct ctl_table *tbl;
2673 struct ctl_table_header *forw_hdr;
2674#endif
2675
2676 err = -ENOMEM;
2677 all = kmemdup(&ipv4_devconf, sizeof(ipv4_devconf), GFP_KERNEL);
2678 if (!all)
2679 goto err_alloc_all;
2680
2681 dflt = kmemdup(&ipv4_devconf_dflt, sizeof(ipv4_devconf_dflt), GFP_KERNEL);
2682 if (!dflt)
2683 goto err_alloc_dflt;
2684
2685#ifdef CONFIG_SYSCTL
2686 tbl = kmemdup(ctl_forward_entry, sizeof(ctl_forward_entry), GFP_KERNEL);
2687 if (!tbl)
2688 goto err_alloc_ctl;
2689
2690 tbl[0].data = &all->data[IPV4_DEVCONF_FORWARDING - 1];
2691 tbl[0].extra1 = all;
2692 tbl[0].extra2 = net;
2693#endif
2694
2695 if (!net_eq(net, &init_net)) {
2696 switch (net_inherit_devconf()) {
2697 case 3:
2698 /* copy from the current netns */
2699 memcpy(all, current->nsproxy->net_ns->ipv4.devconf_all,
2700 sizeof(ipv4_devconf));
2701 memcpy(dflt,
2702 current->nsproxy->net_ns->ipv4.devconf_dflt,
2703 sizeof(ipv4_devconf_dflt));
2704 break;
2705 case 0:
2706 case 1:
2707 /* copy from init_net */
2708 memcpy(all, init_net.ipv4.devconf_all,
2709 sizeof(ipv4_devconf));
2710 memcpy(dflt, init_net.ipv4.devconf_dflt,
2711 sizeof(ipv4_devconf_dflt));
2712 break;
2713 case 2:
2714 /* use compiled values */
2715 break;
2716 }
2717 }
2718
2719#ifdef CONFIG_SYSCTL
2720 err = __devinet_sysctl_register(net, "all", NETCONFA_IFINDEX_ALL, all);
2721 if (err < 0)
2722 goto err_reg_all;
2723
2724 err = __devinet_sysctl_register(net, "default",
2725 NETCONFA_IFINDEX_DEFAULT, dflt);
2726 if (err < 0)
2727 goto err_reg_dflt;
2728
2729 err = -ENOMEM;
2730 forw_hdr = register_net_sysctl_sz(net, "net/ipv4", tbl,
2731 ARRAY_SIZE(ctl_forward_entry));
2732 if (!forw_hdr)
2733 goto err_reg_ctl;
2734 net->ipv4.forw_hdr = forw_hdr;
2735#endif
2736
2737 net->ipv4.devconf_all = all;
2738 net->ipv4.devconf_dflt = dflt;
2739 return 0;
2740
2741#ifdef CONFIG_SYSCTL
2742err_reg_ctl:
2743 __devinet_sysctl_unregister(net, dflt, NETCONFA_IFINDEX_DEFAULT);
2744err_reg_dflt:
2745 __devinet_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
2746err_reg_all:
2747 kfree(tbl);
2748err_alloc_ctl:
2749#endif
2750 kfree(dflt);
2751err_alloc_dflt:
2752 kfree(all);
2753err_alloc_all:
2754 return err;
2755}
2756
2757static __net_exit void devinet_exit_net(struct net *net)
2758{
2759#ifdef CONFIG_SYSCTL
2760 struct ctl_table *tbl;
2761
2762 tbl = net->ipv4.forw_hdr->ctl_table_arg;
2763 unregister_net_sysctl_table(net->ipv4.forw_hdr);
2764 __devinet_sysctl_unregister(net, net->ipv4.devconf_dflt,
2765 NETCONFA_IFINDEX_DEFAULT);
2766 __devinet_sysctl_unregister(net, net->ipv4.devconf_all,
2767 NETCONFA_IFINDEX_ALL);
2768 kfree(tbl);
2769#endif
2770 kfree(net->ipv4.devconf_dflt);
2771 kfree(net->ipv4.devconf_all);
2772}
2773
2774static __net_initdata struct pernet_operations devinet_ops = {
2775 .init = devinet_init_net,
2776 .exit = devinet_exit_net,
2777};
2778
2779static struct rtnl_af_ops inet_af_ops __read_mostly = {
2780 .family = AF_INET,
2781 .fill_link_af = inet_fill_link_af,
2782 .get_link_af_size = inet_get_link_af_size,
2783 .validate_link_af = inet_validate_link_af,
2784 .set_link_af = inet_set_link_af,
2785};
2786
2787void __init devinet_init(void)
2788{
2789 int i;
2790
2791 for (i = 0; i < IN4_ADDR_HSIZE; i++)
2792 INIT_HLIST_HEAD(&inet_addr_lst[i]);
2793
2794 register_pernet_subsys(&devinet_ops);
2795 register_netdevice_notifier(&ip_netdev_notifier);
2796
2797 queue_delayed_work(system_power_efficient_wq, &check_lifetime_work, 0);
2798
2799 rtnl_af_register(&inet_af_ops);
2800
2801 rtnl_register(PF_INET, RTM_NEWADDR, inet_rtm_newaddr, NULL, 0);
2802 rtnl_register(PF_INET, RTM_DELADDR, inet_rtm_deladdr, NULL, 0);
2803 rtnl_register(PF_INET, RTM_GETADDR, NULL, inet_dump_ifaddr,
2804 RTNL_FLAG_DUMP_UNLOCKED);
2805 rtnl_register(PF_INET, RTM_GETNETCONF, inet_netconf_get_devconf,
2806 inet_netconf_dump_devconf,
2807 RTNL_FLAG_DOIT_UNLOCKED | RTNL_FLAG_DUMP_UNLOCKED);
2808}