]> git.proxmox.com Git - mirror_ubuntu-hirsute-kernel.git/blame - net/ipv4/devinet.c
net: igmp: Reduce Unsolicited report interval to 1s when using IGMPv3
[mirror_ubuntu-hirsute-kernel.git] / net / ipv4 / devinet.c
CommitLineData
1da177e4
LT
1/*
2 * NET3 IP device support routines.
3 *
1da177e4
LT
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
02c30a84 10 * Authors: Ross Biro
1da177e4
LT
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
1da177e4
LT
28
29#include <asm/uaccess.h>
1da177e4 30#include <linux/bitops.h>
4fc268d2 31#include <linux/capability.h>
1da177e4
LT
32#include <linux/module.h>
33#include <linux/types.h>
34#include <linux/kernel.h>
1da177e4
LT
35#include <linux/string.h>
36#include <linux/mm.h>
37#include <linux/socket.h>
38#include <linux/sockios.h>
39#include <linux/in.h>
40#include <linux/errno.h>
41#include <linux/interrupt.h>
1823730f 42#include <linux/if_addr.h>
1da177e4
LT
43#include <linux/if_ether.h>
44#include <linux/inet.h>
45#include <linux/netdevice.h>
46#include <linux/etherdevice.h>
47#include <linux/skbuff.h>
1da177e4
LT
48#include <linux/init.h>
49#include <linux/notifier.h>
50#include <linux/inetdevice.h>
51#include <linux/igmp.h>
5a0e3ad6 52#include <linux/slab.h>
fd23c3b3 53#include <linux/hash.h>
1da177e4
LT
54#ifdef CONFIG_SYSCTL
55#include <linux/sysctl.h>
56#endif
57#include <linux/kmod.h>
edc9e748 58#include <linux/netconf.h>
1da177e4 59
14c85021 60#include <net/arp.h>
1da177e4
LT
61#include <net/ip.h>
62#include <net/route.h>
63#include <net/ip_fib.h>
63f3444f 64#include <net/rtnetlink.h>
752d14dc 65#include <net/net_namespace.h>
5c766d64 66#include <net/addrconf.h>
1da177e4 67
406b6f97
DM
68#include "fib_lookup.h"
69
0027ba84 70static struct ipv4_devconf ipv4_devconf = {
42f811b8 71 .data = {
02291680
EB
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,
42f811b8 76 },
1da177e4
LT
77};
78
79static struct ipv4_devconf ipv4_devconf_dflt = {
42f811b8 80 .data = {
02291680
EB
81 [IPV4_DEVCONF_ACCEPT_REDIRECTS - 1] = 1,
82 [IPV4_DEVCONF_SEND_REDIRECTS - 1] = 1,
83 [IPV4_DEVCONF_SECURE_REDIRECTS - 1] = 1,
84 [IPV4_DEVCONF_SHARED_MEDIA - 1] = 1,
85 [IPV4_DEVCONF_ACCEPT_SOURCE_ROUTE - 1] = 1,
42f811b8 86 },
1da177e4
LT
87};
88
9355bbd6
PE
89#define IPV4_DEVCONF_DFLT(net, attr) \
90 IPV4_DEVCONF((*net->ipv4.devconf_dflt), attr)
42f811b8 91
ef7c79ed 92static const struct nla_policy ifa_ipv4_policy[IFA_MAX+1] = {
5c753978
TG
93 [IFA_LOCAL] = { .type = NLA_U32 },
94 [IFA_ADDRESS] = { .type = NLA_U32 },
95 [IFA_BROADCAST] = { .type = NLA_U32 },
5176f91e 96 [IFA_LABEL] = { .type = NLA_STRING, .len = IFNAMSIZ - 1 },
5c766d64 97 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
5c753978
TG
98};
99
40384999
ED
100#define IN4_ADDR_HSIZE_SHIFT 8
101#define IN4_ADDR_HSIZE (1U << IN4_ADDR_HSIZE_SHIFT)
102
fd23c3b3
DM
103static struct hlist_head inet_addr_lst[IN4_ADDR_HSIZE];
104static DEFINE_SPINLOCK(inet_addr_hash_lock);
105
40384999 106static u32 inet_addr_hash(struct net *net, __be32 addr)
fd23c3b3 107{
40384999 108 u32 val = (__force u32) addr ^ net_hash_mix(net);
fd23c3b3 109
40384999 110 return hash_32(val, IN4_ADDR_HSIZE_SHIFT);
fd23c3b3
DM
111}
112
113static void inet_hash_insert(struct net *net, struct in_ifaddr *ifa)
114{
40384999 115 u32 hash = inet_addr_hash(net, ifa->ifa_local);
fd23c3b3
DM
116
117 spin_lock(&inet_addr_hash_lock);
118 hlist_add_head_rcu(&ifa->hash, &inet_addr_lst[hash]);
119 spin_unlock(&inet_addr_hash_lock);
120}
121
122static void inet_hash_remove(struct in_ifaddr *ifa)
123{
124 spin_lock(&inet_addr_hash_lock);
125 hlist_del_init_rcu(&ifa->hash);
126 spin_unlock(&inet_addr_hash_lock);
127}
128
9435eb1c
DM
129/**
130 * __ip_dev_find - find the first device with a given source address.
131 * @net: the net namespace
132 * @addr: the source address
133 * @devref: if true, take a reference on the found device
134 *
135 * If a caller uses devref=false, it should be protected by RCU, or RTNL
136 */
137struct net_device *__ip_dev_find(struct net *net, __be32 addr, bool devref)
138{
40384999 139 u32 hash = inet_addr_hash(net, addr);
9435eb1c
DM
140 struct net_device *result = NULL;
141 struct in_ifaddr *ifa;
9435eb1c
DM
142
143 rcu_read_lock();
b67bfe0d 144 hlist_for_each_entry_rcu(ifa, &inet_addr_lst[hash], hash) {
e066008b 145 if (ifa->ifa_local == addr) {
40384999
ED
146 struct net_device *dev = ifa->ifa_dev->dev;
147
148 if (!net_eq(dev_net(dev), net))
149 continue;
9435eb1c
DM
150 result = dev;
151 break;
152 }
153 }
406b6f97
DM
154 if (!result) {
155 struct flowi4 fl4 = { .daddr = addr };
156 struct fib_result res = { 0 };
157 struct fib_table *local;
158
159 /* Fallback to FIB local table so that communication
160 * over loopback subnets work.
161 */
162 local = fib_get_table(net, RT_TABLE_LOCAL);
163 if (local &&
164 !fib_table_lookup(local, &fl4, &res, FIB_LOOKUP_NOREF) &&
165 res.type == RTN_LOCAL)
166 result = FIB_RES_DEV(res);
167 }
9435eb1c
DM
168 if (result && devref)
169 dev_hold(result);
170 rcu_read_unlock();
171 return result;
172}
173EXPORT_SYMBOL(__ip_dev_find);
174
d6062cbb 175static void rtmsg_ifa(int event, struct in_ifaddr *, struct nlmsghdr *, u32);
1da177e4 176
e041c683 177static BLOCKING_NOTIFIER_HEAD(inetaddr_chain);
1da177e4
LT
178static void inet_del_ifa(struct in_device *in_dev, struct in_ifaddr **ifap,
179 int destroy);
180#ifdef CONFIG_SYSCTL
66f27a52 181static void devinet_sysctl_register(struct in_device *idev);
51602b2a
PE
182static void devinet_sysctl_unregister(struct in_device *idev);
183#else
40384999 184static void devinet_sysctl_register(struct in_device *idev)
51602b2a
PE
185{
186}
40384999 187static void devinet_sysctl_unregister(struct in_device *idev)
51602b2a
PE
188{
189}
1da177e4
LT
190#endif
191
192/* Locks all the inet devices. */
193
194static struct in_ifaddr *inet_alloc_ifa(void)
195{
93adcc80 196 return kzalloc(sizeof(struct in_ifaddr), GFP_KERNEL);
1da177e4
LT
197}
198
199static void inet_rcu_free_ifa(struct rcu_head *head)
200{
201 struct in_ifaddr *ifa = container_of(head, struct in_ifaddr, rcu_head);
202 if (ifa->ifa_dev)
203 in_dev_put(ifa->ifa_dev);
204 kfree(ifa);
205}
206
40384999 207static void inet_free_ifa(struct in_ifaddr *ifa)
1da177e4
LT
208{
209 call_rcu(&ifa->rcu_head, inet_rcu_free_ifa);
210}
211
212void in_dev_finish_destroy(struct in_device *idev)
213{
214 struct net_device *dev = idev->dev;
215
547b792c
IJ
216 WARN_ON(idev->ifa_list);
217 WARN_ON(idev->mc_list);
e9897071 218 kfree(rcu_dereference_protected(idev->mc_hash, 1));
1da177e4 219#ifdef NET_REFCNT_DEBUG
91df42be 220 pr_debug("%s: %p=%s\n", __func__, idev, dev ? dev->name : "NIL");
1da177e4
LT
221#endif
222 dev_put(dev);
223 if (!idev->dead)
9f9354b9
ED
224 pr_err("Freeing alive in_device %p\n", idev);
225 else
1da177e4 226 kfree(idev);
1da177e4 227}
9f9354b9 228EXPORT_SYMBOL(in_dev_finish_destroy);
1da177e4 229
71e27da9 230static struct in_device *inetdev_init(struct net_device *dev)
1da177e4
LT
231{
232 struct in_device *in_dev;
233
234 ASSERT_RTNL();
235
0da974f4 236 in_dev = kzalloc(sizeof(*in_dev), GFP_KERNEL);
1da177e4
LT
237 if (!in_dev)
238 goto out;
c346dca1 239 memcpy(&in_dev->cnf, dev_net(dev)->ipv4.devconf_dflt,
9355bbd6 240 sizeof(in_dev->cnf));
1da177e4
LT
241 in_dev->cnf.sysctl = NULL;
242 in_dev->dev = dev;
9f9354b9
ED
243 in_dev->arp_parms = neigh_parms_alloc(dev, &arp_tbl);
244 if (!in_dev->arp_parms)
1da177e4 245 goto out_kfree;
0187bdfb
BH
246 if (IPV4_DEVCONF(in_dev->cnf, FORWARDING))
247 dev_disable_lro(dev);
1da177e4
LT
248 /* Reference in_dev->dev */
249 dev_hold(dev);
30c4cf57 250 /* Account for reference dev->ip_ptr (below) */
1da177e4 251 in_dev_hold(in_dev);
1da177e4 252
66f27a52 253 devinet_sysctl_register(in_dev);
1da177e4
LT
254 ip_mc_init_dev(in_dev);
255 if (dev->flags & IFF_UP)
256 ip_mc_up(in_dev);
483479ec 257
30c4cf57 258 /* we can receive as soon as ip_ptr is set -- do this last */
cf778b00 259 rcu_assign_pointer(dev->ip_ptr, in_dev);
483479ec 260out:
1da177e4
LT
261 return in_dev;
262out_kfree:
263 kfree(in_dev);
264 in_dev = NULL;
265 goto out;
266}
267
268static void in_dev_rcu_put(struct rcu_head *head)
269{
270 struct in_device *idev = container_of(head, struct in_device, rcu_head);
271 in_dev_put(idev);
272}
273
274static void inetdev_destroy(struct in_device *in_dev)
275{
276 struct in_ifaddr *ifa;
277 struct net_device *dev;
278
279 ASSERT_RTNL();
280
281 dev = in_dev->dev;
1da177e4
LT
282
283 in_dev->dead = 1;
284
285 ip_mc_destroy_dev(in_dev);
286
287 while ((ifa = in_dev->ifa_list) != NULL) {
288 inet_del_ifa(in_dev, &in_dev->ifa_list, 0);
289 inet_free_ifa(ifa);
290 }
291
a9b3cd7f 292 RCU_INIT_POINTER(dev->ip_ptr, NULL);
1da177e4 293
51602b2a 294 devinet_sysctl_unregister(in_dev);
1da177e4
LT
295 neigh_parms_release(&arp_tbl, in_dev->arp_parms);
296 arp_ifdown(dev);
297
298 call_rcu(&in_dev->rcu_head, in_dev_rcu_put);
299}
300
ff428d72 301int inet_addr_onlink(struct in_device *in_dev, __be32 a, __be32 b)
1da177e4
LT
302{
303 rcu_read_lock();
304 for_primary_ifa(in_dev) {
305 if (inet_ifa_match(a, ifa)) {
306 if (!b || inet_ifa_match(b, ifa)) {
307 rcu_read_unlock();
308 return 1;
309 }
310 }
311 } endfor_ifa(in_dev);
312 rcu_read_unlock();
313 return 0;
314}
315
d6062cbb 316static void __inet_del_ifa(struct in_device *in_dev, struct in_ifaddr **ifap,
15e47304 317 int destroy, struct nlmsghdr *nlh, u32 portid)
1da177e4 318{
8f937c60 319 struct in_ifaddr *promote = NULL;
0ff60a45
JHS
320 struct in_ifaddr *ifa, *ifa1 = *ifap;
321 struct in_ifaddr *last_prim = in_dev->ifa_list;
322 struct in_ifaddr *prev_prom = NULL;
323 int do_promote = IN_DEV_PROMOTE_SECONDARIES(in_dev);
1da177e4
LT
324
325 ASSERT_RTNL();
326
e905a9ed 327 /* 1. Deleting primary ifaddr forces deletion all secondaries
8f937c60
HW
328 * unless alias promotion is set
329 **/
1da177e4
LT
330
331 if (!(ifa1->ifa_flags & IFA_F_SECONDARY)) {
1da177e4
LT
332 struct in_ifaddr **ifap1 = &ifa1->ifa_next;
333
334 while ((ifa = *ifap1) != NULL) {
e905a9ed 335 if (!(ifa->ifa_flags & IFA_F_SECONDARY) &&
0ff60a45
JHS
336 ifa1->ifa_scope <= ifa->ifa_scope)
337 last_prim = ifa;
338
1da177e4
LT
339 if (!(ifa->ifa_flags & IFA_F_SECONDARY) ||
340 ifa1->ifa_mask != ifa->ifa_mask ||
341 !inet_ifa_match(ifa1->ifa_address, ifa)) {
342 ifap1 = &ifa->ifa_next;
0ff60a45 343 prev_prom = ifa;
1da177e4
LT
344 continue;
345 }
346
0ff60a45 347 if (!do_promote) {
fd23c3b3 348 inet_hash_remove(ifa);
8f937c60 349 *ifap1 = ifa->ifa_next;
1da177e4 350
15e47304 351 rtmsg_ifa(RTM_DELADDR, ifa, nlh, portid);
e041c683
AS
352 blocking_notifier_call_chain(&inetaddr_chain,
353 NETDEV_DOWN, ifa);
8f937c60
HW
354 inet_free_ifa(ifa);
355 } else {
356 promote = ifa;
357 break;
358 }
1da177e4
LT
359 }
360 }
361
2d230e2b
JA
362 /* On promotion all secondaries from subnet are changing
363 * the primary IP, we must remove all their routes silently
364 * and later to add them back with new prefsrc. Do this
365 * while all addresses are on the device list.
366 */
367 for (ifa = promote; ifa; ifa = ifa->ifa_next) {
368 if (ifa1->ifa_mask == ifa->ifa_mask &&
369 inet_ifa_match(ifa1->ifa_address, ifa))
370 fib_del_ifaddr(ifa, ifa1);
371 }
372
1da177e4
LT
373 /* 2. Unlink it */
374
375 *ifap = ifa1->ifa_next;
fd23c3b3 376 inet_hash_remove(ifa1);
1da177e4
LT
377
378 /* 3. Announce address deletion */
379
380 /* Send message first, then call notifier.
381 At first sight, FIB update triggered by notifier
382 will refer to already deleted ifaddr, that could confuse
383 netlink listeners. It is not true: look, gated sees
384 that route deleted and if it still thinks that ifaddr
385 is valid, it will try to restore deleted routes... Grr.
386 So that, this order is correct.
387 */
15e47304 388 rtmsg_ifa(RTM_DELADDR, ifa1, nlh, portid);
e041c683 389 blocking_notifier_call_chain(&inetaddr_chain, NETDEV_DOWN, ifa1);
1da177e4 390
0ff60a45 391 if (promote) {
04024b93 392 struct in_ifaddr *next_sec = promote->ifa_next;
0ff60a45
JHS
393
394 if (prev_prom) {
395 prev_prom->ifa_next = promote->ifa_next;
396 promote->ifa_next = last_prim->ifa_next;
397 last_prim->ifa_next = promote;
398 }
8f937c60 399
8f937c60 400 promote->ifa_flags &= ~IFA_F_SECONDARY;
15e47304 401 rtmsg_ifa(RTM_NEWADDR, promote, nlh, portid);
e041c683
AS
402 blocking_notifier_call_chain(&inetaddr_chain,
403 NETDEV_UP, promote);
04024b93 404 for (ifa = next_sec; ifa; ifa = ifa->ifa_next) {
0ff60a45
JHS
405 if (ifa1->ifa_mask != ifa->ifa_mask ||
406 !inet_ifa_match(ifa1->ifa_address, ifa))
407 continue;
408 fib_add_ifaddr(ifa);
409 }
410
411 }
6363097c 412 if (destroy)
0ff60a45 413 inet_free_ifa(ifa1);
1da177e4
LT
414}
415
d6062cbb
TG
416static void inet_del_ifa(struct in_device *in_dev, struct in_ifaddr **ifap,
417 int destroy)
418{
419 __inet_del_ifa(in_dev, ifap, destroy, NULL, 0);
420}
421
5c766d64
JP
422static void check_lifetime(struct work_struct *work);
423
424static DECLARE_DELAYED_WORK(check_lifetime_work, check_lifetime);
425
d6062cbb 426static int __inet_insert_ifa(struct in_ifaddr *ifa, struct nlmsghdr *nlh,
15e47304 427 u32 portid)
1da177e4
LT
428{
429 struct in_device *in_dev = ifa->ifa_dev;
430 struct in_ifaddr *ifa1, **ifap, **last_primary;
431
432 ASSERT_RTNL();
433
434 if (!ifa->ifa_local) {
435 inet_free_ifa(ifa);
436 return 0;
437 }
438
439 ifa->ifa_flags &= ~IFA_F_SECONDARY;
440 last_primary = &in_dev->ifa_list;
441
442 for (ifap = &in_dev->ifa_list; (ifa1 = *ifap) != NULL;
443 ifap = &ifa1->ifa_next) {
444 if (!(ifa1->ifa_flags & IFA_F_SECONDARY) &&
445 ifa->ifa_scope <= ifa1->ifa_scope)
446 last_primary = &ifa1->ifa_next;
447 if (ifa1->ifa_mask == ifa->ifa_mask &&
448 inet_ifa_match(ifa1->ifa_address, ifa)) {
449 if (ifa1->ifa_local == ifa->ifa_local) {
450 inet_free_ifa(ifa);
451 return -EEXIST;
452 }
453 if (ifa1->ifa_scope != ifa->ifa_scope) {
454 inet_free_ifa(ifa);
455 return -EINVAL;
456 }
457 ifa->ifa_flags |= IFA_F_SECONDARY;
458 }
459 }
460
461 if (!(ifa->ifa_flags & IFA_F_SECONDARY)) {
462 net_srandom(ifa->ifa_local);
463 ifap = last_primary;
464 }
465
466 ifa->ifa_next = *ifap;
467 *ifap = ifa;
468
fd23c3b3
DM
469 inet_hash_insert(dev_net(in_dev->dev), ifa);
470
5c766d64
JP
471 cancel_delayed_work(&check_lifetime_work);
472 schedule_delayed_work(&check_lifetime_work, 0);
473
1da177e4
LT
474 /* Send message first, then call notifier.
475 Notifier will trigger FIB update, so that
476 listeners of netlink will know about new ifaddr */
15e47304 477 rtmsg_ifa(RTM_NEWADDR, ifa, nlh, portid);
e041c683 478 blocking_notifier_call_chain(&inetaddr_chain, NETDEV_UP, ifa);
1da177e4
LT
479
480 return 0;
481}
482
d6062cbb
TG
483static int inet_insert_ifa(struct in_ifaddr *ifa)
484{
485 return __inet_insert_ifa(ifa, NULL, 0);
486}
487
1da177e4
LT
488static int inet_set_ifa(struct net_device *dev, struct in_ifaddr *ifa)
489{
e5ed6399 490 struct in_device *in_dev = __in_dev_get_rtnl(dev);
1da177e4
LT
491
492 ASSERT_RTNL();
493
494 if (!in_dev) {
71e27da9
HX
495 inet_free_ifa(ifa);
496 return -ENOBUFS;
1da177e4 497 }
71e27da9 498 ipv4_devconf_setall(in_dev);
1da177e4 499 if (ifa->ifa_dev != in_dev) {
547b792c 500 WARN_ON(ifa->ifa_dev);
1da177e4
LT
501 in_dev_hold(in_dev);
502 ifa->ifa_dev = in_dev;
503 }
f97c1e0c 504 if (ipv4_is_loopback(ifa->ifa_local))
1da177e4
LT
505 ifa->ifa_scope = RT_SCOPE_HOST;
506 return inet_insert_ifa(ifa);
507}
508
8723e1b4
ED
509/* Caller must hold RCU or RTNL :
510 * We dont take a reference on found in_device
511 */
7fee0ca2 512struct in_device *inetdev_by_index(struct net *net, int ifindex)
1da177e4
LT
513{
514 struct net_device *dev;
515 struct in_device *in_dev = NULL;
c148fc2e
ED
516
517 rcu_read_lock();
518 dev = dev_get_by_index_rcu(net, ifindex);
1da177e4 519 if (dev)
8723e1b4 520 in_dev = rcu_dereference_rtnl(dev->ip_ptr);
c148fc2e 521 rcu_read_unlock();
1da177e4
LT
522 return in_dev;
523}
9f9354b9 524EXPORT_SYMBOL(inetdev_by_index);
1da177e4
LT
525
526/* Called only from RTNL semaphored context. No locks. */
527
60cad5da
AV
528struct in_ifaddr *inet_ifa_byprefix(struct in_device *in_dev, __be32 prefix,
529 __be32 mask)
1da177e4
LT
530{
531 ASSERT_RTNL();
532
533 for_primary_ifa(in_dev) {
534 if (ifa->ifa_mask == mask && inet_ifa_match(prefix, ifa))
535 return ifa;
536 } endfor_ifa(in_dev);
537 return NULL;
538}
539
661d2967 540static int inet_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh)
1da177e4 541{
3b1e0a65 542 struct net *net = sock_net(skb->sk);
dfdd5fd4 543 struct nlattr *tb[IFA_MAX+1];
1da177e4 544 struct in_device *in_dev;
dfdd5fd4 545 struct ifaddrmsg *ifm;
1da177e4 546 struct in_ifaddr *ifa, **ifap;
dfdd5fd4 547 int err = -EINVAL;
1da177e4
LT
548
549 ASSERT_RTNL();
550
dfdd5fd4
TG
551 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv4_policy);
552 if (err < 0)
553 goto errout;
554
555 ifm = nlmsg_data(nlh);
7fee0ca2 556 in_dev = inetdev_by_index(net, ifm->ifa_index);
dfdd5fd4
TG
557 if (in_dev == NULL) {
558 err = -ENODEV;
559 goto errout;
560 }
561
1da177e4
LT
562 for (ifap = &in_dev->ifa_list; (ifa = *ifap) != NULL;
563 ifap = &ifa->ifa_next) {
dfdd5fd4 564 if (tb[IFA_LOCAL] &&
a7a628c4 565 ifa->ifa_local != nla_get_be32(tb[IFA_LOCAL]))
dfdd5fd4
TG
566 continue;
567
568 if (tb[IFA_LABEL] && nla_strcmp(tb[IFA_LABEL], ifa->ifa_label))
1da177e4 569 continue;
dfdd5fd4
TG
570
571 if (tb[IFA_ADDRESS] &&
572 (ifm->ifa_prefixlen != ifa->ifa_prefixlen ||
a7a628c4 573 !inet_ifa_match(nla_get_be32(tb[IFA_ADDRESS]), ifa)))
dfdd5fd4
TG
574 continue;
575
15e47304 576 __inet_del_ifa(in_dev, ifap, 1, nlh, NETLINK_CB(skb).portid);
1da177e4
LT
577 return 0;
578 }
dfdd5fd4
TG
579
580 err = -EADDRNOTAVAIL;
581errout:
582 return err;
1da177e4
LT
583}
584
5c766d64
JP
585#define INFINITY_LIFE_TIME 0xFFFFFFFF
586
587static void check_lifetime(struct work_struct *work)
588{
589 unsigned long now, next, next_sec, next_sched;
590 struct in_ifaddr *ifa;
c988d1e8 591 struct hlist_node *n;
5c766d64
JP
592 int i;
593
594 now = jiffies;
595 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
596
5c766d64 597 for (i = 0; i < IN4_ADDR_HSIZE; i++) {
c988d1e8
JP
598 bool change_needed = false;
599
600 rcu_read_lock();
b67bfe0d 601 hlist_for_each_entry_rcu(ifa, &inet_addr_lst[i], hash) {
5c766d64
JP
602 unsigned long age;
603
604 if (ifa->ifa_flags & IFA_F_PERMANENT)
605 continue;
606
607 /* We try to batch several events at once. */
608 age = (now - ifa->ifa_tstamp +
609 ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
610
611 if (ifa->ifa_valid_lft != INFINITY_LIFE_TIME &&
612 age >= ifa->ifa_valid_lft) {
c988d1e8 613 change_needed = true;
5c766d64
JP
614 } else if (ifa->ifa_preferred_lft ==
615 INFINITY_LIFE_TIME) {
616 continue;
617 } else if (age >= ifa->ifa_preferred_lft) {
618 if (time_before(ifa->ifa_tstamp +
619 ifa->ifa_valid_lft * HZ, next))
620 next = ifa->ifa_tstamp +
621 ifa->ifa_valid_lft * HZ;
622
c988d1e8
JP
623 if (!(ifa->ifa_flags & IFA_F_DEPRECATED))
624 change_needed = true;
5c766d64
JP
625 } else if (time_before(ifa->ifa_tstamp +
626 ifa->ifa_preferred_lft * HZ,
627 next)) {
628 next = ifa->ifa_tstamp +
629 ifa->ifa_preferred_lft * HZ;
630 }
631 }
c988d1e8
JP
632 rcu_read_unlock();
633 if (!change_needed)
634 continue;
635 rtnl_lock();
636 hlist_for_each_entry_safe(ifa, n, &inet_addr_lst[i], hash) {
637 unsigned long age;
638
639 if (ifa->ifa_flags & IFA_F_PERMANENT)
640 continue;
641
642 /* We try to batch several events at once. */
643 age = (now - ifa->ifa_tstamp +
644 ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
645
646 if (ifa->ifa_valid_lft != INFINITY_LIFE_TIME &&
647 age >= ifa->ifa_valid_lft) {
648 struct in_ifaddr **ifap;
649
650 for (ifap = &ifa->ifa_dev->ifa_list;
651 *ifap != NULL; ifap = &(*ifap)->ifa_next) {
652 if (*ifap == ifa) {
653 inet_del_ifa(ifa->ifa_dev,
654 ifap, 1);
655 break;
656 }
657 }
658 } else if (ifa->ifa_preferred_lft !=
659 INFINITY_LIFE_TIME &&
660 age >= ifa->ifa_preferred_lft &&
661 !(ifa->ifa_flags & IFA_F_DEPRECATED)) {
662 ifa->ifa_flags |= IFA_F_DEPRECATED;
663 rtmsg_ifa(RTM_NEWADDR, ifa, NULL, 0);
664 }
665 }
666 rtnl_unlock();
5c766d64 667 }
5c766d64
JP
668
669 next_sec = round_jiffies_up(next);
670 next_sched = next;
671
672 /* If rounded timeout is accurate enough, accept it. */
673 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
674 next_sched = next_sec;
675
676 now = jiffies;
677 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
678 if (time_before(next_sched, now + ADDRCONF_TIMER_FUZZ_MAX))
679 next_sched = now + ADDRCONF_TIMER_FUZZ_MAX;
680
681 schedule_delayed_work(&check_lifetime_work, next_sched - now);
682}
683
684static void set_ifa_lifetime(struct in_ifaddr *ifa, __u32 valid_lft,
685 __u32 prefered_lft)
686{
687 unsigned long timeout;
688
689 ifa->ifa_flags &= ~(IFA_F_PERMANENT | IFA_F_DEPRECATED);
690
691 timeout = addrconf_timeout_fixup(valid_lft, HZ);
692 if (addrconf_finite_timeout(timeout))
693 ifa->ifa_valid_lft = timeout;
694 else
695 ifa->ifa_flags |= IFA_F_PERMANENT;
696
697 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
698 if (addrconf_finite_timeout(timeout)) {
699 if (timeout == 0)
700 ifa->ifa_flags |= IFA_F_DEPRECATED;
701 ifa->ifa_preferred_lft = timeout;
702 }
703 ifa->ifa_tstamp = jiffies;
704 if (!ifa->ifa_cstamp)
705 ifa->ifa_cstamp = ifa->ifa_tstamp;
706}
707
708static struct in_ifaddr *rtm_to_ifaddr(struct net *net, struct nlmsghdr *nlh,
709 __u32 *pvalid_lft, __u32 *pprefered_lft)
1da177e4 710{
5c753978
TG
711 struct nlattr *tb[IFA_MAX+1];
712 struct in_ifaddr *ifa;
713 struct ifaddrmsg *ifm;
1da177e4
LT
714 struct net_device *dev;
715 struct in_device *in_dev;
7b218574 716 int err;
1da177e4 717
5c753978
TG
718 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv4_policy);
719 if (err < 0)
720 goto errout;
1da177e4 721
5c753978 722 ifm = nlmsg_data(nlh);
7b218574
DL
723 err = -EINVAL;
724 if (ifm->ifa_prefixlen > 32 || tb[IFA_LOCAL] == NULL)
5c753978 725 goto errout;
1da177e4 726
4b8aa9ab 727 dev = __dev_get_by_index(net, ifm->ifa_index);
7b218574
DL
728 err = -ENODEV;
729 if (dev == NULL)
5c753978 730 goto errout;
1da177e4 731
5c753978 732 in_dev = __in_dev_get_rtnl(dev);
7b218574
DL
733 err = -ENOBUFS;
734 if (in_dev == NULL)
71e27da9 735 goto errout;
1da177e4 736
5c753978 737 ifa = inet_alloc_ifa();
7b218574 738 if (ifa == NULL)
5c753978
TG
739 /*
740 * A potential indev allocation can be left alive, it stays
741 * assigned to its device and is destroy with it.
742 */
5c753978 743 goto errout;
5c753978 744
a4e65d36 745 ipv4_devconf_setall(in_dev);
5c753978
TG
746 in_dev_hold(in_dev);
747
748 if (tb[IFA_ADDRESS] == NULL)
749 tb[IFA_ADDRESS] = tb[IFA_LOCAL];
1da177e4 750
fd23c3b3 751 INIT_HLIST_NODE(&ifa->hash);
1da177e4
LT
752 ifa->ifa_prefixlen = ifm->ifa_prefixlen;
753 ifa->ifa_mask = inet_make_mask(ifm->ifa_prefixlen);
1da177e4
LT
754 ifa->ifa_flags = ifm->ifa_flags;
755 ifa->ifa_scope = ifm->ifa_scope;
5c753978
TG
756 ifa->ifa_dev = in_dev;
757
a7a628c4
AV
758 ifa->ifa_local = nla_get_be32(tb[IFA_LOCAL]);
759 ifa->ifa_address = nla_get_be32(tb[IFA_ADDRESS]);
5c753978
TG
760
761 if (tb[IFA_BROADCAST])
a7a628c4 762 ifa->ifa_broadcast = nla_get_be32(tb[IFA_BROADCAST]);
5c753978 763
5c753978
TG
764 if (tb[IFA_LABEL])
765 nla_strlcpy(ifa->ifa_label, tb[IFA_LABEL], IFNAMSIZ);
1da177e4
LT
766 else
767 memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
768
5c766d64
JP
769 if (tb[IFA_CACHEINFO]) {
770 struct ifa_cacheinfo *ci;
771
772 ci = nla_data(tb[IFA_CACHEINFO]);
773 if (!ci->ifa_valid || ci->ifa_prefered > ci->ifa_valid) {
774 err = -EINVAL;
446266b0 775 goto errout_free;
5c766d64
JP
776 }
777 *pvalid_lft = ci->ifa_valid;
778 *pprefered_lft = ci->ifa_prefered;
779 }
780
5c753978
TG
781 return ifa;
782
446266b0
DB
783errout_free:
784 inet_free_ifa(ifa);
5c753978
TG
785errout:
786 return ERR_PTR(err);
787}
788
5c766d64
JP
789static struct in_ifaddr *find_matching_ifa(struct in_ifaddr *ifa)
790{
791 struct in_device *in_dev = ifa->ifa_dev;
792 struct in_ifaddr *ifa1, **ifap;
793
794 if (!ifa->ifa_local)
795 return NULL;
796
797 for (ifap = &in_dev->ifa_list; (ifa1 = *ifap) != NULL;
798 ifap = &ifa1->ifa_next) {
799 if (ifa1->ifa_mask == ifa->ifa_mask &&
800 inet_ifa_match(ifa1->ifa_address, ifa) &&
801 ifa1->ifa_local == ifa->ifa_local)
802 return ifa1;
803 }
804 return NULL;
805}
806
661d2967 807static int inet_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh)
5c753978 808{
3b1e0a65 809 struct net *net = sock_net(skb->sk);
5c753978 810 struct in_ifaddr *ifa;
5c766d64
JP
811 struct in_ifaddr *ifa_existing;
812 __u32 valid_lft = INFINITY_LIFE_TIME;
813 __u32 prefered_lft = INFINITY_LIFE_TIME;
5c753978
TG
814
815 ASSERT_RTNL();
816
5c766d64 817 ifa = rtm_to_ifaddr(net, nlh, &valid_lft, &prefered_lft);
5c753978
TG
818 if (IS_ERR(ifa))
819 return PTR_ERR(ifa);
820
5c766d64
JP
821 ifa_existing = find_matching_ifa(ifa);
822 if (!ifa_existing) {
823 /* It would be best to check for !NLM_F_CREATE here but
824 * userspace alreay relies on not having to provide this.
825 */
826 set_ifa_lifetime(ifa, valid_lft, prefered_lft);
827 return __inet_insert_ifa(ifa, nlh, NETLINK_CB(skb).portid);
828 } else {
829 inet_free_ifa(ifa);
830
831 if (nlh->nlmsg_flags & NLM_F_EXCL ||
832 !(nlh->nlmsg_flags & NLM_F_REPLACE))
833 return -EEXIST;
34e2ed34
JP
834 ifa = ifa_existing;
835 set_ifa_lifetime(ifa, valid_lft, prefered_lft);
05a324b9
JP
836 cancel_delayed_work(&check_lifetime_work);
837 schedule_delayed_work(&check_lifetime_work, 0);
34e2ed34
JP
838 rtmsg_ifa(RTM_NEWADDR, ifa, nlh, NETLINK_CB(skb).portid);
839 blocking_notifier_call_chain(&inetaddr_chain, NETDEV_UP, ifa);
5c766d64
JP
840 }
841 return 0;
1da177e4
LT
842}
843
844/*
845 * Determine a default network mask, based on the IP address.
846 */
847
40384999 848static int inet_abc_len(__be32 addr)
1da177e4
LT
849{
850 int rc = -1; /* Something else, probably a multicast. */
851
f97c1e0c 852 if (ipv4_is_zeronet(addr))
e905a9ed 853 rc = 0;
1da177e4 854 else {
714e85be 855 __u32 haddr = ntohl(addr);
1da177e4 856
714e85be 857 if (IN_CLASSA(haddr))
1da177e4 858 rc = 8;
714e85be 859 else if (IN_CLASSB(haddr))
1da177e4 860 rc = 16;
714e85be 861 else if (IN_CLASSC(haddr))
1da177e4
LT
862 rc = 24;
863 }
864
e905a9ed 865 return rc;
1da177e4
LT
866}
867
868
e5b13cb1 869int devinet_ioctl(struct net *net, unsigned int cmd, void __user *arg)
1da177e4
LT
870{
871 struct ifreq ifr;
872 struct sockaddr_in sin_orig;
873 struct sockaddr_in *sin = (struct sockaddr_in *)&ifr.ifr_addr;
874 struct in_device *in_dev;
875 struct in_ifaddr **ifap = NULL;
876 struct in_ifaddr *ifa = NULL;
877 struct net_device *dev;
878 char *colon;
879 int ret = -EFAULT;
880 int tryaddrmatch = 0;
881
882 /*
883 * Fetch the caller's info block into kernel space
884 */
885
886 if (copy_from_user(&ifr, arg, sizeof(struct ifreq)))
887 goto out;
888 ifr.ifr_name[IFNAMSIZ - 1] = 0;
889
890 /* save original address for comparison */
891 memcpy(&sin_orig, sin, sizeof(*sin));
892
893 colon = strchr(ifr.ifr_name, ':');
894 if (colon)
895 *colon = 0;
896
e5b13cb1 897 dev_load(net, ifr.ifr_name);
1da177e4 898
132adf54 899 switch (cmd) {
1da177e4
LT
900 case SIOCGIFADDR: /* Get interface address */
901 case SIOCGIFBRDADDR: /* Get the broadcast address */
902 case SIOCGIFDSTADDR: /* Get the destination address */
903 case SIOCGIFNETMASK: /* Get the netmask for the interface */
904 /* Note that these ioctls will not sleep,
905 so that we do not impose a lock.
906 One day we will be forced to put shlock here (I mean SMP)
907 */
908 tryaddrmatch = (sin_orig.sin_family == AF_INET);
909 memset(sin, 0, sizeof(*sin));
910 sin->sin_family = AF_INET;
911 break;
912
913 case SIOCSIFFLAGS:
bf5b30b8 914 ret = -EPERM;
52e804c6 915 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1da177e4
LT
916 goto out;
917 break;
918 case SIOCSIFADDR: /* Set interface address (and family) */
919 case SIOCSIFBRDADDR: /* Set the broadcast address */
920 case SIOCSIFDSTADDR: /* Set the destination address */
921 case SIOCSIFNETMASK: /* Set the netmask for the interface */
bf5b30b8 922 ret = -EPERM;
52e804c6 923 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1da177e4
LT
924 goto out;
925 ret = -EINVAL;
926 if (sin->sin_family != AF_INET)
927 goto out;
928 break;
929 default:
930 ret = -EINVAL;
931 goto out;
932 }
933
934 rtnl_lock();
935
936 ret = -ENODEV;
9f9354b9
ED
937 dev = __dev_get_by_name(net, ifr.ifr_name);
938 if (!dev)
1da177e4
LT
939 goto done;
940
941 if (colon)
942 *colon = ':';
943
9f9354b9
ED
944 in_dev = __in_dev_get_rtnl(dev);
945 if (in_dev) {
1da177e4
LT
946 if (tryaddrmatch) {
947 /* Matthias Andree */
948 /* compare label and address (4.4BSD style) */
949 /* note: we only do this for a limited set of ioctls
950 and only if the original address family was AF_INET.
951 This is checked above. */
952 for (ifap = &in_dev->ifa_list; (ifa = *ifap) != NULL;
953 ifap = &ifa->ifa_next) {
954 if (!strcmp(ifr.ifr_name, ifa->ifa_label) &&
955 sin_orig.sin_addr.s_addr ==
6c91afe1 956 ifa->ifa_local) {
1da177e4
LT
957 break; /* found */
958 }
959 }
960 }
961 /* we didn't get a match, maybe the application is
962 4.3BSD-style and passed in junk so we fall back to
963 comparing just the label */
964 if (!ifa) {
965 for (ifap = &in_dev->ifa_list; (ifa = *ifap) != NULL;
966 ifap = &ifa->ifa_next)
967 if (!strcmp(ifr.ifr_name, ifa->ifa_label))
968 break;
969 }
970 }
971
972 ret = -EADDRNOTAVAIL;
973 if (!ifa && cmd != SIOCSIFADDR && cmd != SIOCSIFFLAGS)
974 goto done;
975
132adf54 976 switch (cmd) {
1da177e4
LT
977 case SIOCGIFADDR: /* Get interface address */
978 sin->sin_addr.s_addr = ifa->ifa_local;
979 goto rarok;
980
981 case SIOCGIFBRDADDR: /* Get the broadcast address */
982 sin->sin_addr.s_addr = ifa->ifa_broadcast;
983 goto rarok;
984
985 case SIOCGIFDSTADDR: /* Get the destination address */
986 sin->sin_addr.s_addr = ifa->ifa_address;
987 goto rarok;
988
989 case SIOCGIFNETMASK: /* Get the netmask for the interface */
990 sin->sin_addr.s_addr = ifa->ifa_mask;
991 goto rarok;
992
993 case SIOCSIFFLAGS:
994 if (colon) {
995 ret = -EADDRNOTAVAIL;
996 if (!ifa)
997 break;
998 ret = 0;
999 if (!(ifr.ifr_flags & IFF_UP))
1000 inet_del_ifa(in_dev, ifap, 1);
1001 break;
1002 }
1003 ret = dev_change_flags(dev, ifr.ifr_flags);
1004 break;
1005
1006 case SIOCSIFADDR: /* Set interface address (and family) */
1007 ret = -EINVAL;
1008 if (inet_abc_len(sin->sin_addr.s_addr) < 0)
1009 break;
1010
1011 if (!ifa) {
1012 ret = -ENOBUFS;
9f9354b9
ED
1013 ifa = inet_alloc_ifa();
1014 if (!ifa)
1da177e4 1015 break;
c7e2e1d7 1016 INIT_HLIST_NODE(&ifa->hash);
1da177e4
LT
1017 if (colon)
1018 memcpy(ifa->ifa_label, ifr.ifr_name, IFNAMSIZ);
1019 else
1020 memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
1021 } else {
1022 ret = 0;
1023 if (ifa->ifa_local == sin->sin_addr.s_addr)
1024 break;
1025 inet_del_ifa(in_dev, ifap, 0);
1026 ifa->ifa_broadcast = 0;
148f9729 1027 ifa->ifa_scope = 0;
1da177e4
LT
1028 }
1029
1030 ifa->ifa_address = ifa->ifa_local = sin->sin_addr.s_addr;
1031
1032 if (!(dev->flags & IFF_POINTOPOINT)) {
1033 ifa->ifa_prefixlen = inet_abc_len(ifa->ifa_address);
1034 ifa->ifa_mask = inet_make_mask(ifa->ifa_prefixlen);
1035 if ((dev->flags & IFF_BROADCAST) &&
1036 ifa->ifa_prefixlen < 31)
1037 ifa->ifa_broadcast = ifa->ifa_address |
1038 ~ifa->ifa_mask;
1039 } else {
1040 ifa->ifa_prefixlen = 32;
1041 ifa->ifa_mask = inet_make_mask(32);
1042 }
5c766d64 1043 set_ifa_lifetime(ifa, INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
1da177e4
LT
1044 ret = inet_set_ifa(dev, ifa);
1045 break;
1046
1047 case SIOCSIFBRDADDR: /* Set the broadcast address */
1048 ret = 0;
1049 if (ifa->ifa_broadcast != sin->sin_addr.s_addr) {
1050 inet_del_ifa(in_dev, ifap, 0);
1051 ifa->ifa_broadcast = sin->sin_addr.s_addr;
1052 inet_insert_ifa(ifa);
1053 }
1054 break;
1055
1056 case SIOCSIFDSTADDR: /* Set the destination address */
1057 ret = 0;
1058 if (ifa->ifa_address == sin->sin_addr.s_addr)
1059 break;
1060 ret = -EINVAL;
1061 if (inet_abc_len(sin->sin_addr.s_addr) < 0)
1062 break;
1063 ret = 0;
1064 inet_del_ifa(in_dev, ifap, 0);
1065 ifa->ifa_address = sin->sin_addr.s_addr;
1066 inet_insert_ifa(ifa);
1067 break;
1068
1069 case SIOCSIFNETMASK: /* Set the netmask for the interface */
1070
1071 /*
1072 * The mask we set must be legal.
1073 */
1074 ret = -EINVAL;
1075 if (bad_mask(sin->sin_addr.s_addr, 0))
1076 break;
1077 ret = 0;
1078 if (ifa->ifa_mask != sin->sin_addr.s_addr) {
a144ea4b 1079 __be32 old_mask = ifa->ifa_mask;
1da177e4
LT
1080 inet_del_ifa(in_dev, ifap, 0);
1081 ifa->ifa_mask = sin->sin_addr.s_addr;
1082 ifa->ifa_prefixlen = inet_mask_len(ifa->ifa_mask);
1083
1084 /* See if current broadcast address matches
1085 * with current netmask, then recalculate
1086 * the broadcast address. Otherwise it's a
1087 * funny address, so don't touch it since
1088 * the user seems to know what (s)he's doing...
1089 */
1090 if ((dev->flags & IFF_BROADCAST) &&
1091 (ifa->ifa_prefixlen < 31) &&
1092 (ifa->ifa_broadcast ==
dcab5e1e 1093 (ifa->ifa_local|~old_mask))) {
1da177e4
LT
1094 ifa->ifa_broadcast = (ifa->ifa_local |
1095 ~sin->sin_addr.s_addr);
1096 }
1097 inet_insert_ifa(ifa);
1098 }
1099 break;
1100 }
1101done:
1102 rtnl_unlock();
1103out:
1104 return ret;
1105rarok:
1106 rtnl_unlock();
1107 ret = copy_to_user(arg, &ifr, sizeof(struct ifreq)) ? -EFAULT : 0;
1108 goto out;
1109}
1110
1111static int inet_gifconf(struct net_device *dev, char __user *buf, int len)
1112{
e5ed6399 1113 struct in_device *in_dev = __in_dev_get_rtnl(dev);
1da177e4
LT
1114 struct in_ifaddr *ifa;
1115 struct ifreq ifr;
1116 int done = 0;
1117
9f9354b9 1118 if (!in_dev)
1da177e4
LT
1119 goto out;
1120
9f9354b9 1121 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
1da177e4
LT
1122 if (!buf) {
1123 done += sizeof(ifr);
1124 continue;
1125 }
1126 if (len < (int) sizeof(ifr))
1127 break;
1128 memset(&ifr, 0, sizeof(struct ifreq));
4299c8a9 1129 strcpy(ifr.ifr_name, ifa->ifa_label);
1da177e4
LT
1130
1131 (*(struct sockaddr_in *)&ifr.ifr_addr).sin_family = AF_INET;
1132 (*(struct sockaddr_in *)&ifr.ifr_addr).sin_addr.s_addr =
1133 ifa->ifa_local;
1134
1135 if (copy_to_user(buf, &ifr, sizeof(struct ifreq))) {
1136 done = -EFAULT;
1137 break;
1138 }
1139 buf += sizeof(struct ifreq);
1140 len -= sizeof(struct ifreq);
1141 done += sizeof(struct ifreq);
1142 }
1143out:
1144 return done;
1145}
1146
a61ced5d 1147__be32 inet_select_addr(const struct net_device *dev, __be32 dst, int scope)
1da177e4 1148{
a61ced5d 1149 __be32 addr = 0;
1da177e4 1150 struct in_device *in_dev;
c346dca1 1151 struct net *net = dev_net(dev);
1da177e4
LT
1152
1153 rcu_read_lock();
e5ed6399 1154 in_dev = __in_dev_get_rcu(dev);
1da177e4
LT
1155 if (!in_dev)
1156 goto no_in_dev;
1157
1158 for_primary_ifa(in_dev) {
1159 if (ifa->ifa_scope > scope)
1160 continue;
1161 if (!dst || inet_ifa_match(dst, ifa)) {
1162 addr = ifa->ifa_local;
1163 break;
1164 }
1165 if (!addr)
1166 addr = ifa->ifa_local;
1167 } endfor_ifa(in_dev);
1da177e4
LT
1168
1169 if (addr)
c6d14c84 1170 goto out_unlock;
9f9354b9 1171no_in_dev:
1da177e4
LT
1172
1173 /* Not loopback addresses on loopback should be preferred
1174 in this case. It is importnat that lo is the first interface
1175 in dev_base list.
1176 */
c6d14c84 1177 for_each_netdev_rcu(net, dev) {
9f9354b9
ED
1178 in_dev = __in_dev_get_rcu(dev);
1179 if (!in_dev)
1da177e4
LT
1180 continue;
1181
1182 for_primary_ifa(in_dev) {
1183 if (ifa->ifa_scope != RT_SCOPE_LINK &&
1184 ifa->ifa_scope <= scope) {
1185 addr = ifa->ifa_local;
c6d14c84 1186 goto out_unlock;
1da177e4
LT
1187 }
1188 } endfor_ifa(in_dev);
1189 }
c6d14c84 1190out_unlock:
1da177e4 1191 rcu_read_unlock();
1da177e4
LT
1192 return addr;
1193}
9f9354b9 1194EXPORT_SYMBOL(inet_select_addr);
1da177e4 1195
60cad5da
AV
1196static __be32 confirm_addr_indev(struct in_device *in_dev, __be32 dst,
1197 __be32 local, int scope)
1da177e4
LT
1198{
1199 int same = 0;
a144ea4b 1200 __be32 addr = 0;
1da177e4
LT
1201
1202 for_ifa(in_dev) {
1203 if (!addr &&
1204 (local == ifa->ifa_local || !local) &&
1205 ifa->ifa_scope <= scope) {
1206 addr = ifa->ifa_local;
1207 if (same)
1208 break;
1209 }
1210 if (!same) {
1211 same = (!local || inet_ifa_match(local, ifa)) &&
1212 (!dst || inet_ifa_match(dst, ifa));
1213 if (same && addr) {
1214 if (local || !dst)
1215 break;
1216 /* Is the selected addr into dst subnet? */
1217 if (inet_ifa_match(addr, ifa))
1218 break;
1219 /* No, then can we use new local src? */
1220 if (ifa->ifa_scope <= scope) {
1221 addr = ifa->ifa_local;
1222 break;
1223 }
1224 /* search for large dst subnet for addr */
1225 same = 0;
1226 }
1227 }
1228 } endfor_ifa(in_dev);
1229
9f9354b9 1230 return same ? addr : 0;
1da177e4
LT
1231}
1232
1233/*
1234 * Confirm that local IP address exists using wildcards:
9bd85e32 1235 * - in_dev: only on this interface, 0=any interface
1da177e4
LT
1236 * - dst: only in the same subnet as dst, 0=any dst
1237 * - local: address, 0=autoselect the local address
1238 * - scope: maximum allowed scope value for the local address
1239 */
9bd85e32
DL
1240__be32 inet_confirm_addr(struct in_device *in_dev,
1241 __be32 dst, __be32 local, int scope)
1da177e4 1242{
60cad5da 1243 __be32 addr = 0;
9bd85e32 1244 struct net_device *dev;
39a6d063 1245 struct net *net;
1da177e4 1246
39a6d063 1247 if (scope != RT_SCOPE_LINK)
9bd85e32 1248 return confirm_addr_indev(in_dev, dst, local, scope);
1da177e4 1249
c346dca1 1250 net = dev_net(in_dev->dev);
1da177e4 1251 rcu_read_lock();
c6d14c84 1252 for_each_netdev_rcu(net, dev) {
9f9354b9
ED
1253 in_dev = __in_dev_get_rcu(dev);
1254 if (in_dev) {
1da177e4
LT
1255 addr = confirm_addr_indev(in_dev, dst, local, scope);
1256 if (addr)
1257 break;
1258 }
1259 }
1260 rcu_read_unlock();
1da177e4
LT
1261
1262 return addr;
1263}
eaddcd76 1264EXPORT_SYMBOL(inet_confirm_addr);
1da177e4
LT
1265
1266/*
1267 * Device notifier
1268 */
1269
1270int register_inetaddr_notifier(struct notifier_block *nb)
1271{
e041c683 1272 return blocking_notifier_chain_register(&inetaddr_chain, nb);
1da177e4 1273}
9f9354b9 1274EXPORT_SYMBOL(register_inetaddr_notifier);
1da177e4
LT
1275
1276int unregister_inetaddr_notifier(struct notifier_block *nb)
1277{
e041c683 1278 return blocking_notifier_chain_unregister(&inetaddr_chain, nb);
1da177e4 1279}
9f9354b9 1280EXPORT_SYMBOL(unregister_inetaddr_notifier);
1da177e4 1281
9f9354b9
ED
1282/* Rename ifa_labels for a device name change. Make some effort to preserve
1283 * existing alias numbering and to create unique labels if possible.
1da177e4
LT
1284*/
1285static void inetdev_changename(struct net_device *dev, struct in_device *in_dev)
e905a9ed 1286{
1da177e4
LT
1287 struct in_ifaddr *ifa;
1288 int named = 0;
1289
e905a9ed
YH
1290 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
1291 char old[IFNAMSIZ], *dot;
1da177e4
LT
1292
1293 memcpy(old, ifa->ifa_label, IFNAMSIZ);
e905a9ed 1294 memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
1da177e4 1295 if (named++ == 0)
573bf470 1296 goto skip;
44344b2a 1297 dot = strchr(old, ':');
e905a9ed
YH
1298 if (dot == NULL) {
1299 sprintf(old, ":%d", named);
1da177e4
LT
1300 dot = old;
1301 }
9f9354b9 1302 if (strlen(dot) + strlen(dev->name) < IFNAMSIZ)
e905a9ed 1303 strcat(ifa->ifa_label, dot);
9f9354b9 1304 else
e905a9ed 1305 strcpy(ifa->ifa_label + (IFNAMSIZ - strlen(dot) - 1), dot);
573bf470
TG
1306skip:
1307 rtmsg_ifa(RTM_NEWADDR, ifa, NULL, 0);
e905a9ed
YH
1308 }
1309}
1da177e4 1310
40384999 1311static bool inetdev_valid_mtu(unsigned int mtu)
06770843
BL
1312{
1313 return mtu >= 68;
1314}
1315
d11327ad
IC
1316static void inetdev_send_gratuitous_arp(struct net_device *dev,
1317 struct in_device *in_dev)
1318
1319{
b76d0789 1320 struct in_ifaddr *ifa;
d11327ad 1321
b76d0789
ZK
1322 for (ifa = in_dev->ifa_list; ifa;
1323 ifa = ifa->ifa_next) {
1324 arp_send(ARPOP_REQUEST, ETH_P_ARP,
1325 ifa->ifa_local, dev,
1326 ifa->ifa_local, NULL,
1327 dev->dev_addr, NULL);
1328 }
d11327ad
IC
1329}
1330
1da177e4
LT
1331/* Called only under RTNL semaphore */
1332
1333static int inetdev_event(struct notifier_block *this, unsigned long event,
1334 void *ptr)
1335{
351638e7 1336 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
748e2d93 1337 struct in_device *in_dev = __in_dev_get_rtnl(dev);
0115e8e3 1338
1da177e4
LT
1339 ASSERT_RTNL();
1340
1341 if (!in_dev) {
8030f544 1342 if (event == NETDEV_REGISTER) {
1da177e4 1343 in_dev = inetdev_init(dev);
b217d616
HX
1344 if (!in_dev)
1345 return notifier_from_errno(-ENOMEM);
0cc217e1 1346 if (dev->flags & IFF_LOOPBACK) {
42f811b8
HX
1347 IN_DEV_CONF_SET(in_dev, NOXFRM, 1);
1348 IN_DEV_CONF_SET(in_dev, NOPOLICY, 1);
8030f544 1349 }
06770843
BL
1350 } else if (event == NETDEV_CHANGEMTU) {
1351 /* Re-enabling IP */
1352 if (inetdev_valid_mtu(dev->mtu))
1353 in_dev = inetdev_init(dev);
1da177e4
LT
1354 }
1355 goto out;
1356 }
1357
1358 switch (event) {
1359 case NETDEV_REGISTER:
91df42be 1360 pr_debug("%s: bug\n", __func__);
a9b3cd7f 1361 RCU_INIT_POINTER(dev->ip_ptr, NULL);
1da177e4
LT
1362 break;
1363 case NETDEV_UP:
06770843 1364 if (!inetdev_valid_mtu(dev->mtu))
1da177e4 1365 break;
0cc217e1 1366 if (dev->flags & IFF_LOOPBACK) {
9f9354b9
ED
1367 struct in_ifaddr *ifa = inet_alloc_ifa();
1368
1369 if (ifa) {
fd23c3b3 1370 INIT_HLIST_NODE(&ifa->hash);
1da177e4
LT
1371 ifa->ifa_local =
1372 ifa->ifa_address = htonl(INADDR_LOOPBACK);
1373 ifa->ifa_prefixlen = 8;
1374 ifa->ifa_mask = inet_make_mask(8);
1375 in_dev_hold(in_dev);
1376 ifa->ifa_dev = in_dev;
1377 ifa->ifa_scope = RT_SCOPE_HOST;
1378 memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
5c766d64
JP
1379 set_ifa_lifetime(ifa, INFINITY_LIFE_TIME,
1380 INFINITY_LIFE_TIME);
1da177e4
LT
1381 inet_insert_ifa(ifa);
1382 }
1383 }
1384 ip_mc_up(in_dev);
eefef1cf
SH
1385 /* fall through */
1386 case NETDEV_CHANGEADDR:
d11327ad
IC
1387 if (!IN_DEV_ARP_NOTIFY(in_dev))
1388 break;
1389 /* fall through */
1390 case NETDEV_NOTIFY_PEERS:
a21090cf 1391 /* Send gratuitous ARP to notify of link change */
d11327ad 1392 inetdev_send_gratuitous_arp(dev, in_dev);
1da177e4
LT
1393 break;
1394 case NETDEV_DOWN:
1395 ip_mc_down(in_dev);
1396 break;
93d9b7d7 1397 case NETDEV_PRE_TYPE_CHANGE:
75c78500
MS
1398 ip_mc_unmap(in_dev);
1399 break;
93d9b7d7 1400 case NETDEV_POST_TYPE_CHANGE:
75c78500
MS
1401 ip_mc_remap(in_dev);
1402 break;
1da177e4 1403 case NETDEV_CHANGEMTU:
06770843 1404 if (inetdev_valid_mtu(dev->mtu))
1da177e4 1405 break;
06770843 1406 /* disable IP when MTU is not enough */
1da177e4
LT
1407 case NETDEV_UNREGISTER:
1408 inetdev_destroy(in_dev);
1409 break;
1410 case NETDEV_CHANGENAME:
1411 /* Do not notify about label change, this event is
1412 * not interesting to applications using netlink.
1413 */
1414 inetdev_changename(dev, in_dev);
1415
51602b2a 1416 devinet_sysctl_unregister(in_dev);
66f27a52 1417 devinet_sysctl_register(in_dev);
1da177e4
LT
1418 break;
1419 }
1420out:
1421 return NOTIFY_DONE;
1422}
1423
1424static struct notifier_block ip_netdev_notifier = {
539afedf 1425 .notifier_call = inetdev_event,
1da177e4
LT
1426};
1427
40384999 1428static size_t inet_nlmsg_size(void)
339bf98f
TG
1429{
1430 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
1431 + nla_total_size(4) /* IFA_ADDRESS */
1432 + nla_total_size(4) /* IFA_LOCAL */
1433 + nla_total_size(4) /* IFA_BROADCAST */
339bf98f
TG
1434 + nla_total_size(IFNAMSIZ); /* IFA_LABEL */
1435}
1436
5c766d64
JP
1437static inline u32 cstamp_delta(unsigned long cstamp)
1438{
1439 return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
1440}
1441
1442static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
1443 unsigned long tstamp, u32 preferred, u32 valid)
1444{
1445 struct ifa_cacheinfo ci;
1446
1447 ci.cstamp = cstamp_delta(cstamp);
1448 ci.tstamp = cstamp_delta(tstamp);
1449 ci.ifa_prefered = preferred;
1450 ci.ifa_valid = valid;
1451
1452 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
1453}
1454
1da177e4 1455static int inet_fill_ifaddr(struct sk_buff *skb, struct in_ifaddr *ifa,
15e47304 1456 u32 portid, u32 seq, int event, unsigned int flags)
1da177e4
LT
1457{
1458 struct ifaddrmsg *ifm;
1459 struct nlmsghdr *nlh;
5c766d64 1460 u32 preferred, valid;
1da177e4 1461
15e47304 1462 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*ifm), flags);
47f68512 1463 if (nlh == NULL)
26932566 1464 return -EMSGSIZE;
47f68512
TG
1465
1466 ifm = nlmsg_data(nlh);
1da177e4
LT
1467 ifm->ifa_family = AF_INET;
1468 ifm->ifa_prefixlen = ifa->ifa_prefixlen;
5c766d64 1469 ifm->ifa_flags = ifa->ifa_flags;
1da177e4
LT
1470 ifm->ifa_scope = ifa->ifa_scope;
1471 ifm->ifa_index = ifa->ifa_dev->dev->ifindex;
47f68512 1472
5c766d64
JP
1473 if (!(ifm->ifa_flags & IFA_F_PERMANENT)) {
1474 preferred = ifa->ifa_preferred_lft;
1475 valid = ifa->ifa_valid_lft;
1476 if (preferred != INFINITY_LIFE_TIME) {
1477 long tval = (jiffies - ifa->ifa_tstamp) / HZ;
1478
1479 if (preferred > tval)
1480 preferred -= tval;
1481 else
1482 preferred = 0;
1483 if (valid != INFINITY_LIFE_TIME) {
1484 if (valid > tval)
1485 valid -= tval;
1486 else
1487 valid = 0;
1488 }
1489 }
1490 } else {
1491 preferred = INFINITY_LIFE_TIME;
1492 valid = INFINITY_LIFE_TIME;
1493 }
f3756b79
DM
1494 if ((ifa->ifa_address &&
1495 nla_put_be32(skb, IFA_ADDRESS, ifa->ifa_address)) ||
1496 (ifa->ifa_local &&
1497 nla_put_be32(skb, IFA_LOCAL, ifa->ifa_local)) ||
1498 (ifa->ifa_broadcast &&
1499 nla_put_be32(skb, IFA_BROADCAST, ifa->ifa_broadcast)) ||
1500 (ifa->ifa_label[0] &&
5c766d64
JP
1501 nla_put_string(skb, IFA_LABEL, ifa->ifa_label)) ||
1502 put_cacheinfo(skb, ifa->ifa_cstamp, ifa->ifa_tstamp,
1503 preferred, valid))
f3756b79 1504 goto nla_put_failure;
1da177e4 1505
47f68512
TG
1506 return nlmsg_end(skb, nlh);
1507
1508nla_put_failure:
26932566
PM
1509 nlmsg_cancel(skb, nlh);
1510 return -EMSGSIZE;
1da177e4
LT
1511}
1512
1513static int inet_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
1514{
3b1e0a65 1515 struct net *net = sock_net(skb->sk);
eec4df98
ED
1516 int h, s_h;
1517 int idx, s_idx;
1518 int ip_idx, s_ip_idx;
1da177e4
LT
1519 struct net_device *dev;
1520 struct in_device *in_dev;
1521 struct in_ifaddr *ifa;
eec4df98 1522 struct hlist_head *head;
1da177e4 1523
eec4df98
ED
1524 s_h = cb->args[0];
1525 s_idx = idx = cb->args[1];
1526 s_ip_idx = ip_idx = cb->args[2];
1527
1528 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1529 idx = 0;
1530 head = &net->dev_index_head[h];
1531 rcu_read_lock();
0465277f
ND
1532 cb->seq = atomic_read(&net->ipv4.dev_addr_genid) ^
1533 net->dev_base_seq;
b67bfe0d 1534 hlist_for_each_entry_rcu(dev, head, index_hlist) {
eec4df98
ED
1535 if (idx < s_idx)
1536 goto cont;
4b97efdf 1537 if (h > s_h || idx > s_idx)
eec4df98
ED
1538 s_ip_idx = 0;
1539 in_dev = __in_dev_get_rcu(dev);
1540 if (!in_dev)
1541 goto cont;
1da177e4 1542
eec4df98
ED
1543 for (ifa = in_dev->ifa_list, ip_idx = 0; ifa;
1544 ifa = ifa->ifa_next, ip_idx++) {
1545 if (ip_idx < s_ip_idx)
1546 continue;
1547 if (inet_fill_ifaddr(skb, ifa,
15e47304 1548 NETLINK_CB(cb->skb).portid,
1da177e4 1549 cb->nlh->nlmsg_seq,
eec4df98
ED
1550 RTM_NEWADDR, NLM_F_MULTI) <= 0) {
1551 rcu_read_unlock();
1552 goto done;
1553 }
0465277f 1554 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
eec4df98 1555 }
7562f876 1556cont:
eec4df98
ED
1557 idx++;
1558 }
1559 rcu_read_unlock();
1da177e4
LT
1560 }
1561
1562done:
eec4df98
ED
1563 cb->args[0] = h;
1564 cb->args[1] = idx;
1565 cb->args[2] = ip_idx;
1da177e4
LT
1566
1567 return skb->len;
1568}
1569
539afedf 1570static void rtmsg_ifa(int event, struct in_ifaddr *ifa, struct nlmsghdr *nlh,
15e47304 1571 u32 portid)
1da177e4 1572{
47f68512 1573 struct sk_buff *skb;
d6062cbb
TG
1574 u32 seq = nlh ? nlh->nlmsg_seq : 0;
1575 int err = -ENOBUFS;
4b8aa9ab 1576 struct net *net;
1da177e4 1577
c346dca1 1578 net = dev_net(ifa->ifa_dev->dev);
339bf98f 1579 skb = nlmsg_new(inet_nlmsg_size(), GFP_KERNEL);
47f68512 1580 if (skb == NULL)
d6062cbb
TG
1581 goto errout;
1582
15e47304 1583 err = inet_fill_ifaddr(skb, ifa, portid, seq, event, 0);
26932566
PM
1584 if (err < 0) {
1585 /* -EMSGSIZE implies BUG in inet_nlmsg_size() */
1586 WARN_ON(err == -EMSGSIZE);
1587 kfree_skb(skb);
1588 goto errout;
1589 }
15e47304 1590 rtnl_notify(skb, net, portid, RTNLGRP_IPV4_IFADDR, nlh, GFP_KERNEL);
1ce85fe4 1591 return;
d6062cbb
TG
1592errout:
1593 if (err < 0)
4b8aa9ab 1594 rtnl_set_sk_err(net, RTNLGRP_IPV4_IFADDR, err);
1da177e4
LT
1595}
1596
9f0f7272
TG
1597static size_t inet_get_link_af_size(const struct net_device *dev)
1598{
1fc19aff 1599 struct in_device *in_dev = rcu_dereference_rtnl(dev->ip_ptr);
9f0f7272
TG
1600
1601 if (!in_dev)
1602 return 0;
1603
1604 return nla_total_size(IPV4_DEVCONF_MAX * 4); /* IFLA_INET_CONF */
1605}
1606
1607static int inet_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
1608{
1fc19aff 1609 struct in_device *in_dev = rcu_dereference_rtnl(dev->ip_ptr);
9f0f7272
TG
1610 struct nlattr *nla;
1611 int i;
1612
1613 if (!in_dev)
1614 return -ENODATA;
1615
1616 nla = nla_reserve(skb, IFLA_INET_CONF, IPV4_DEVCONF_MAX * 4);
1617 if (nla == NULL)
1618 return -EMSGSIZE;
1619
1620 for (i = 0; i < IPV4_DEVCONF_MAX; i++)
1621 ((u32 *) nla_data(nla))[i] = in_dev->cnf.data[i];
1622
1623 return 0;
1624}
1625
1626static const struct nla_policy inet_af_policy[IFLA_INET_MAX+1] = {
1627 [IFLA_INET_CONF] = { .type = NLA_NESTED },
1628};
1629
cf7afbfe
TG
1630static int inet_validate_link_af(const struct net_device *dev,
1631 const struct nlattr *nla)
9f0f7272 1632{
9f0f7272
TG
1633 struct nlattr *a, *tb[IFLA_INET_MAX+1];
1634 int err, rem;
1635
f7fce74e 1636 if (dev && !__in_dev_get_rtnl(dev))
cf7afbfe 1637 return -EAFNOSUPPORT;
9f0f7272
TG
1638
1639 err = nla_parse_nested(tb, IFLA_INET_MAX, nla, inet_af_policy);
1640 if (err < 0)
1641 return err;
1642
1643 if (tb[IFLA_INET_CONF]) {
1644 nla_for_each_nested(a, tb[IFLA_INET_CONF], rem) {
1645 int cfgid = nla_type(a);
1646
1647 if (nla_len(a) < 4)
1648 return -EINVAL;
1649
1650 if (cfgid <= 0 || cfgid > IPV4_DEVCONF_MAX)
1651 return -EINVAL;
1652 }
1653 }
1654
cf7afbfe
TG
1655 return 0;
1656}
1657
1658static int inet_set_link_af(struct net_device *dev, const struct nlattr *nla)
1659{
f7fce74e 1660 struct in_device *in_dev = __in_dev_get_rtnl(dev);
cf7afbfe
TG
1661 struct nlattr *a, *tb[IFLA_INET_MAX+1];
1662 int rem;
1663
1664 if (!in_dev)
1665 return -EAFNOSUPPORT;
1666
1667 if (nla_parse_nested(tb, IFLA_INET_MAX, nla, NULL) < 0)
1668 BUG();
1669
9f0f7272
TG
1670 if (tb[IFLA_INET_CONF]) {
1671 nla_for_each_nested(a, tb[IFLA_INET_CONF], rem)
1672 ipv4_devconf_set(in_dev, nla_type(a), nla_get_u32(a));
1673 }
1674
1675 return 0;
1676}
1677
edc9e748
ND
1678static int inet_netconf_msgsize_devconf(int type)
1679{
1680 int size = NLMSG_ALIGN(sizeof(struct netconfmsg))
1681 + nla_total_size(4); /* NETCONFA_IFINDEX */
1682
9e551110
ND
1683 /* type -1 is used for ALL */
1684 if (type == -1 || type == NETCONFA_FORWARDING)
edc9e748 1685 size += nla_total_size(4);
cc535dfb
ND
1686 if (type == -1 || type == NETCONFA_RP_FILTER)
1687 size += nla_total_size(4);
d67b8c61
ND
1688 if (type == -1 || type == NETCONFA_MC_FORWARDING)
1689 size += nla_total_size(4);
edc9e748
ND
1690
1691 return size;
1692}
1693
1694static int inet_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
1695 struct ipv4_devconf *devconf, u32 portid,
1696 u32 seq, int event, unsigned int flags,
1697 int type)
1698{
1699 struct nlmsghdr *nlh;
1700 struct netconfmsg *ncm;
1701
1702 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
1703 flags);
1704 if (nlh == NULL)
1705 return -EMSGSIZE;
1706
1707 ncm = nlmsg_data(nlh);
1708 ncm->ncm_family = AF_INET;
1709
1710 if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
1711 goto nla_put_failure;
1712
9e551110
ND
1713 /* type -1 is used for ALL */
1714 if ((type == -1 || type == NETCONFA_FORWARDING) &&
edc9e748
ND
1715 nla_put_s32(skb, NETCONFA_FORWARDING,
1716 IPV4_DEVCONF(*devconf, FORWARDING)) < 0)
1717 goto nla_put_failure;
cc535dfb
ND
1718 if ((type == -1 || type == NETCONFA_RP_FILTER) &&
1719 nla_put_s32(skb, NETCONFA_RP_FILTER,
1720 IPV4_DEVCONF(*devconf, RP_FILTER)) < 0)
1721 goto nla_put_failure;
d67b8c61
ND
1722 if ((type == -1 || type == NETCONFA_MC_FORWARDING) &&
1723 nla_put_s32(skb, NETCONFA_MC_FORWARDING,
1724 IPV4_DEVCONF(*devconf, MC_FORWARDING)) < 0)
1725 goto nla_put_failure;
edc9e748
ND
1726
1727 return nlmsg_end(skb, nlh);
1728
1729nla_put_failure:
1730 nlmsg_cancel(skb, nlh);
1731 return -EMSGSIZE;
1732}
1733
d67b8c61
ND
1734void inet_netconf_notify_devconf(struct net *net, int type, int ifindex,
1735 struct ipv4_devconf *devconf)
edc9e748
ND
1736{
1737 struct sk_buff *skb;
1738 int err = -ENOBUFS;
1739
1740 skb = nlmsg_new(inet_netconf_msgsize_devconf(type), GFP_ATOMIC);
1741 if (skb == NULL)
1742 goto errout;
1743
1744 err = inet_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
1745 RTM_NEWNETCONF, 0, type);
1746 if (err < 0) {
1747 /* -EMSGSIZE implies BUG in inet_netconf_msgsize_devconf() */
1748 WARN_ON(err == -EMSGSIZE);
1749 kfree_skb(skb);
1750 goto errout;
1751 }
1752 rtnl_notify(skb, net, 0, RTNLGRP_IPV4_NETCONF, NULL, GFP_ATOMIC);
1753 return;
1754errout:
1755 if (err < 0)
1756 rtnl_set_sk_err(net, RTNLGRP_IPV4_NETCONF, err);
1757}
1758
9e551110
ND
1759static const struct nla_policy devconf_ipv4_policy[NETCONFA_MAX+1] = {
1760 [NETCONFA_IFINDEX] = { .len = sizeof(int) },
1761 [NETCONFA_FORWARDING] = { .len = sizeof(int) },
cc535dfb 1762 [NETCONFA_RP_FILTER] = { .len = sizeof(int) },
9e551110
ND
1763};
1764
1765static int inet_netconf_get_devconf(struct sk_buff *in_skb,
661d2967 1766 struct nlmsghdr *nlh)
9e551110
ND
1767{
1768 struct net *net = sock_net(in_skb->sk);
1769 struct nlattr *tb[NETCONFA_MAX+1];
1770 struct netconfmsg *ncm;
1771 struct sk_buff *skb;
1772 struct ipv4_devconf *devconf;
1773 struct in_device *in_dev;
1774 struct net_device *dev;
1775 int ifindex;
1776 int err;
1777
1778 err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
1779 devconf_ipv4_policy);
1780 if (err < 0)
1781 goto errout;
1782
1783 err = EINVAL;
1784 if (!tb[NETCONFA_IFINDEX])
1785 goto errout;
1786
1787 ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
1788 switch (ifindex) {
1789 case NETCONFA_IFINDEX_ALL:
1790 devconf = net->ipv4.devconf_all;
1791 break;
1792 case NETCONFA_IFINDEX_DEFAULT:
1793 devconf = net->ipv4.devconf_dflt;
1794 break;
1795 default:
1796 dev = __dev_get_by_index(net, ifindex);
1797 if (dev == NULL)
1798 goto errout;
1799 in_dev = __in_dev_get_rtnl(dev);
1800 if (in_dev == NULL)
1801 goto errout;
1802 devconf = &in_dev->cnf;
1803 break;
1804 }
1805
1806 err = -ENOBUFS;
1807 skb = nlmsg_new(inet_netconf_msgsize_devconf(-1), GFP_ATOMIC);
1808 if (skb == NULL)
1809 goto errout;
1810
1811 err = inet_netconf_fill_devconf(skb, ifindex, devconf,
1812 NETLINK_CB(in_skb).portid,
1813 nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
1814 -1);
1815 if (err < 0) {
1816 /* -EMSGSIZE implies BUG in inet_netconf_msgsize_devconf() */
1817 WARN_ON(err == -EMSGSIZE);
1818 kfree_skb(skb);
1819 goto errout;
1820 }
1821 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
1822errout:
1823 return err;
1824}
1825
7a674200
ND
1826static int inet_netconf_dump_devconf(struct sk_buff *skb,
1827 struct netlink_callback *cb)
1828{
1829 struct net *net = sock_net(skb->sk);
1830 int h, s_h;
1831 int idx, s_idx;
1832 struct net_device *dev;
1833 struct in_device *in_dev;
1834 struct hlist_head *head;
1835
1836 s_h = cb->args[0];
1837 s_idx = idx = cb->args[1];
1838
1839 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1840 idx = 0;
1841 head = &net->dev_index_head[h];
1842 rcu_read_lock();
0465277f
ND
1843 cb->seq = atomic_read(&net->ipv4.dev_addr_genid) ^
1844 net->dev_base_seq;
7a674200
ND
1845 hlist_for_each_entry_rcu(dev, head, index_hlist) {
1846 if (idx < s_idx)
1847 goto cont;
1848 in_dev = __in_dev_get_rcu(dev);
1849 if (!in_dev)
1850 goto cont;
1851
1852 if (inet_netconf_fill_devconf(skb, dev->ifindex,
1853 &in_dev->cnf,
1854 NETLINK_CB(cb->skb).portid,
1855 cb->nlh->nlmsg_seq,
1856 RTM_NEWNETCONF,
1857 NLM_F_MULTI,
1858 -1) <= 0) {
1859 rcu_read_unlock();
1860 goto done;
1861 }
0465277f 1862 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
7a674200
ND
1863cont:
1864 idx++;
1865 }
1866 rcu_read_unlock();
1867 }
1868 if (h == NETDEV_HASHENTRIES) {
1869 if (inet_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
1870 net->ipv4.devconf_all,
1871 NETLINK_CB(cb->skb).portid,
1872 cb->nlh->nlmsg_seq,
1873 RTM_NEWNETCONF, NLM_F_MULTI,
1874 -1) <= 0)
1875 goto done;
1876 else
1877 h++;
1878 }
1879 if (h == NETDEV_HASHENTRIES + 1) {
1880 if (inet_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
1881 net->ipv4.devconf_dflt,
1882 NETLINK_CB(cb->skb).portid,
1883 cb->nlh->nlmsg_seq,
1884 RTM_NEWNETCONF, NLM_F_MULTI,
1885 -1) <= 0)
1886 goto done;
1887 else
1888 h++;
1889 }
1890done:
1891 cb->args[0] = h;
1892 cb->args[1] = idx;
1893
1894 return skb->len;
1895}
1896
1da177e4
LT
1897#ifdef CONFIG_SYSCTL
1898
c0ce9fb3 1899static void devinet_copy_dflt_conf(struct net *net, int i)
31be3085
HX
1900{
1901 struct net_device *dev;
1902
c6d14c84
ED
1903 rcu_read_lock();
1904 for_each_netdev_rcu(net, dev) {
31be3085 1905 struct in_device *in_dev;
c6d14c84 1906
31be3085
HX
1907 in_dev = __in_dev_get_rcu(dev);
1908 if (in_dev && !test_bit(i, in_dev->cnf.state))
9355bbd6 1909 in_dev->cnf.data[i] = net->ipv4.devconf_dflt->data[i];
31be3085 1910 }
c6d14c84 1911 rcu_read_unlock();
31be3085
HX
1912}
1913
c6d14c84 1914/* called with RTNL locked */
c0ce9fb3 1915static void inet_forward_change(struct net *net)
68dd299b
PE
1916{
1917 struct net_device *dev;
586f1211 1918 int on = IPV4_DEVCONF_ALL(net, FORWARDING);
68dd299b 1919
586f1211 1920 IPV4_DEVCONF_ALL(net, ACCEPT_REDIRECTS) = !on;
9355bbd6 1921 IPV4_DEVCONF_DFLT(net, FORWARDING) = on;
edc9e748
ND
1922 inet_netconf_notify_devconf(net, NETCONFA_FORWARDING,
1923 NETCONFA_IFINDEX_ALL,
1924 net->ipv4.devconf_all);
1925 inet_netconf_notify_devconf(net, NETCONFA_FORWARDING,
1926 NETCONFA_IFINDEX_DEFAULT,
1927 net->ipv4.devconf_dflt);
68dd299b 1928
c0ce9fb3 1929 for_each_netdev(net, dev) {
68dd299b 1930 struct in_device *in_dev;
0187bdfb
BH
1931 if (on)
1932 dev_disable_lro(dev);
68dd299b
PE
1933 rcu_read_lock();
1934 in_dev = __in_dev_get_rcu(dev);
edc9e748 1935 if (in_dev) {
68dd299b 1936 IN_DEV_CONF_SET(in_dev, FORWARDING, on);
edc9e748
ND
1937 inet_netconf_notify_devconf(net, NETCONFA_FORWARDING,
1938 dev->ifindex, &in_dev->cnf);
1939 }
68dd299b
PE
1940 rcu_read_unlock();
1941 }
68dd299b
PE
1942}
1943
fe2c6338 1944static int devinet_conf_proc(struct ctl_table *ctl, int write,
8d65af78 1945 void __user *buffer,
31be3085
HX
1946 size_t *lenp, loff_t *ppos)
1947{
d01ff0a0 1948 int old_value = *(int *)ctl->data;
8d65af78 1949 int ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
d01ff0a0 1950 int new_value = *(int *)ctl->data;
31be3085
HX
1951
1952 if (write) {
1953 struct ipv4_devconf *cnf = ctl->extra1;
c0ce9fb3 1954 struct net *net = ctl->extra2;
31be3085
HX
1955 int i = (int *)ctl->data - cnf->data;
1956
1957 set_bit(i, cnf->state);
1958
9355bbd6 1959 if (cnf == net->ipv4.devconf_dflt)
c0ce9fb3 1960 devinet_copy_dflt_conf(net, i);
d0daebc3
TG
1961 if (i == IPV4_DEVCONF_ACCEPT_LOCAL - 1 ||
1962 i == IPV4_DEVCONF_ROUTE_LOCALNET - 1)
d01ff0a0 1963 if ((new_value == 0) && (old_value != 0))
4ccfe6d4 1964 rt_cache_flush(net);
cc535dfb
ND
1965 if (i == IPV4_DEVCONF_RP_FILTER - 1 &&
1966 new_value != old_value) {
1967 int ifindex;
1968
1969 if (cnf == net->ipv4.devconf_dflt)
1970 ifindex = NETCONFA_IFINDEX_DEFAULT;
1971 else if (cnf == net->ipv4.devconf_all)
1972 ifindex = NETCONFA_IFINDEX_ALL;
1973 else {
1974 struct in_device *idev =
1975 container_of(cnf, struct in_device,
1976 cnf);
1977 ifindex = idev->dev->ifindex;
1978 }
1979 inet_netconf_notify_devconf(net, NETCONFA_RP_FILTER,
1980 ifindex, cnf);
1981 }
31be3085
HX
1982 }
1983
1984 return ret;
1985}
1986
fe2c6338 1987static int devinet_sysctl_forward(struct ctl_table *ctl, int write,
8d65af78 1988 void __user *buffer,
1da177e4
LT
1989 size_t *lenp, loff_t *ppos)
1990{
1991 int *valp = ctl->data;
1992 int val = *valp;
88af182e 1993 loff_t pos = *ppos;
8d65af78 1994 int ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
1da177e4
LT
1995
1996 if (write && *valp != val) {
c0ce9fb3
PE
1997 struct net *net = ctl->extra2;
1998
0187bdfb 1999 if (valp != &IPV4_DEVCONF_DFLT(net, FORWARDING)) {
88af182e
EB
2000 if (!rtnl_trylock()) {
2001 /* Restore the original values before restarting */
2002 *valp = val;
2003 *ppos = pos;
9b8adb5e 2004 return restart_syscall();
88af182e 2005 }
0187bdfb
BH
2006 if (valp == &IPV4_DEVCONF_ALL(net, FORWARDING)) {
2007 inet_forward_change(net);
edc9e748 2008 } else {
0187bdfb
BH
2009 struct ipv4_devconf *cnf = ctl->extra1;
2010 struct in_device *idev =
2011 container_of(cnf, struct in_device, cnf);
edc9e748
ND
2012 if (*valp)
2013 dev_disable_lro(idev->dev);
2014 inet_netconf_notify_devconf(net,
2015 NETCONFA_FORWARDING,
2016 idev->dev->ifindex,
2017 cnf);
0187bdfb
BH
2018 }
2019 rtnl_unlock();
4ccfe6d4 2020 rt_cache_flush(net);
edc9e748
ND
2021 } else
2022 inet_netconf_notify_devconf(net, NETCONFA_FORWARDING,
2023 NETCONFA_IFINDEX_DEFAULT,
2024 net->ipv4.devconf_dflt);
1da177e4
LT
2025 }
2026
2027 return ret;
2028}
2029
fe2c6338 2030static int ipv4_doint_and_flush(struct ctl_table *ctl, int write,
323e126f
DM
2031 void __user *buffer,
2032 size_t *lenp, loff_t *ppos)
1da177e4
LT
2033{
2034 int *valp = ctl->data;
2035 int val = *valp;
8d65af78 2036 int ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
76e6ebfb 2037 struct net *net = ctl->extra2;
1da177e4
LT
2038
2039 if (write && *valp != val)
4ccfe6d4 2040 rt_cache_flush(net);
1da177e4
LT
2041
2042 return ret;
2043}
2044
f8572d8f 2045#define DEVINET_SYSCTL_ENTRY(attr, name, mval, proc) \
42f811b8 2046 { \
42f811b8
HX
2047 .procname = name, \
2048 .data = ipv4_devconf.data + \
02291680 2049 IPV4_DEVCONF_ ## attr - 1, \
42f811b8
HX
2050 .maxlen = sizeof(int), \
2051 .mode = mval, \
2052 .proc_handler = proc, \
31be3085 2053 .extra1 = &ipv4_devconf, \
42f811b8
HX
2054 }
2055
2056#define DEVINET_SYSCTL_RW_ENTRY(attr, name) \
f8572d8f 2057 DEVINET_SYSCTL_ENTRY(attr, name, 0644, devinet_conf_proc)
42f811b8
HX
2058
2059#define DEVINET_SYSCTL_RO_ENTRY(attr, name) \
f8572d8f 2060 DEVINET_SYSCTL_ENTRY(attr, name, 0444, devinet_conf_proc)
42f811b8 2061
f8572d8f
EB
2062#define DEVINET_SYSCTL_COMPLEX_ENTRY(attr, name, proc) \
2063 DEVINET_SYSCTL_ENTRY(attr, name, 0644, proc)
42f811b8
HX
2064
2065#define DEVINET_SYSCTL_FLUSHING_ENTRY(attr, name) \
f8572d8f 2066 DEVINET_SYSCTL_COMPLEX_ENTRY(attr, name, ipv4_doint_and_flush)
42f811b8 2067
1da177e4
LT
2068static struct devinet_sysctl_table {
2069 struct ctl_table_header *sysctl_header;
02291680 2070 struct ctl_table devinet_vars[__IPV4_DEVCONF_MAX];
1da177e4
LT
2071} devinet_sysctl = {
2072 .devinet_vars = {
42f811b8 2073 DEVINET_SYSCTL_COMPLEX_ENTRY(FORWARDING, "forwarding",
f8572d8f 2074 devinet_sysctl_forward),
42f811b8
HX
2075 DEVINET_SYSCTL_RO_ENTRY(MC_FORWARDING, "mc_forwarding"),
2076
2077 DEVINET_SYSCTL_RW_ENTRY(ACCEPT_REDIRECTS, "accept_redirects"),
2078 DEVINET_SYSCTL_RW_ENTRY(SECURE_REDIRECTS, "secure_redirects"),
2079 DEVINET_SYSCTL_RW_ENTRY(SHARED_MEDIA, "shared_media"),
2080 DEVINET_SYSCTL_RW_ENTRY(RP_FILTER, "rp_filter"),
2081 DEVINET_SYSCTL_RW_ENTRY(SEND_REDIRECTS, "send_redirects"),
2082 DEVINET_SYSCTL_RW_ENTRY(ACCEPT_SOURCE_ROUTE,
2083 "accept_source_route"),
8153a10c 2084 DEVINET_SYSCTL_RW_ENTRY(ACCEPT_LOCAL, "accept_local"),
28f6aeea 2085 DEVINET_SYSCTL_RW_ENTRY(SRC_VMARK, "src_valid_mark"),
42f811b8
HX
2086 DEVINET_SYSCTL_RW_ENTRY(PROXY_ARP, "proxy_arp"),
2087 DEVINET_SYSCTL_RW_ENTRY(MEDIUM_ID, "medium_id"),
2088 DEVINET_SYSCTL_RW_ENTRY(BOOTP_RELAY, "bootp_relay"),
2089 DEVINET_SYSCTL_RW_ENTRY(LOG_MARTIANS, "log_martians"),
2090 DEVINET_SYSCTL_RW_ENTRY(TAG, "tag"),
2091 DEVINET_SYSCTL_RW_ENTRY(ARPFILTER, "arp_filter"),
2092 DEVINET_SYSCTL_RW_ENTRY(ARP_ANNOUNCE, "arp_announce"),
2093 DEVINET_SYSCTL_RW_ENTRY(ARP_IGNORE, "arp_ignore"),
2094 DEVINET_SYSCTL_RW_ENTRY(ARP_ACCEPT, "arp_accept"),
eefef1cf 2095 DEVINET_SYSCTL_RW_ENTRY(ARP_NOTIFY, "arp_notify"),
65324144 2096 DEVINET_SYSCTL_RW_ENTRY(PROXY_ARP_PVLAN, "proxy_arp_pvlan"),
42f811b8
HX
2097
2098 DEVINET_SYSCTL_FLUSHING_ENTRY(NOXFRM, "disable_xfrm"),
2099 DEVINET_SYSCTL_FLUSHING_ENTRY(NOPOLICY, "disable_policy"),
2100 DEVINET_SYSCTL_FLUSHING_ENTRY(FORCE_IGMP_VERSION,
2101 "force_igmp_version"),
2102 DEVINET_SYSCTL_FLUSHING_ENTRY(PROMOTE_SECONDARIES,
2103 "promote_secondaries"),
d0daebc3
TG
2104 DEVINET_SYSCTL_FLUSHING_ENTRY(ROUTE_LOCALNET,
2105 "route_localnet"),
1da177e4 2106 },
1da177e4
LT
2107};
2108
ea40b324 2109static int __devinet_sysctl_register(struct net *net, char *dev_name,
f8572d8f 2110 struct ipv4_devconf *p)
1da177e4
LT
2111{
2112 int i;
9fa89642 2113 struct devinet_sysctl_table *t;
8607ddb8 2114 char path[sizeof("net/ipv4/conf/") + IFNAMSIZ];
bfada697 2115
9fa89642 2116 t = kmemdup(&devinet_sysctl, sizeof(*t), GFP_KERNEL);
1da177e4 2117 if (!t)
9fa89642
PE
2118 goto out;
2119
1da177e4
LT
2120 for (i = 0; i < ARRAY_SIZE(t->devinet_vars) - 1; i++) {
2121 t->devinet_vars[i].data += (char *)p - (char *)&ipv4_devconf;
31be3085 2122 t->devinet_vars[i].extra1 = p;
c0ce9fb3 2123 t->devinet_vars[i].extra2 = net;
1da177e4
LT
2124 }
2125
8607ddb8 2126 snprintf(path, sizeof(path), "net/ipv4/conf/%s", dev_name);
1da177e4 2127
8607ddb8 2128 t->sysctl_header = register_net_sysctl(net, path, t->devinet_vars);
1da177e4 2129 if (!t->sysctl_header)
8607ddb8 2130 goto free;
1da177e4
LT
2131
2132 p->sysctl = t;
ea40b324 2133 return 0;
1da177e4 2134
9fa89642 2135free:
1da177e4 2136 kfree(t);
9fa89642 2137out:
ea40b324 2138 return -ENOBUFS;
1da177e4
LT
2139}
2140
51602b2a
PE
2141static void __devinet_sysctl_unregister(struct ipv4_devconf *cnf)
2142{
2143 struct devinet_sysctl_table *t = cnf->sysctl;
2144
2145 if (t == NULL)
2146 return;
2147
2148 cnf->sysctl = NULL;
ff538818 2149 unregister_net_sysctl_table(t->sysctl_header);
51602b2a
PE
2150 kfree(t);
2151}
2152
66f27a52
PE
2153static void devinet_sysctl_register(struct in_device *idev)
2154{
54716e3b 2155 neigh_sysctl_register(idev->dev, idev->arp_parms, "ipv4", NULL);
c346dca1 2156 __devinet_sysctl_register(dev_net(idev->dev), idev->dev->name,
f8572d8f 2157 &idev->cnf);
66f27a52
PE
2158}
2159
51602b2a 2160static void devinet_sysctl_unregister(struct in_device *idev)
1da177e4 2161{
51602b2a
PE
2162 __devinet_sysctl_unregister(&idev->cnf);
2163 neigh_sysctl_unregister(idev->arp_parms);
1da177e4 2164}
1da177e4 2165
68dd299b
PE
2166static struct ctl_table ctl_forward_entry[] = {
2167 {
68dd299b
PE
2168 .procname = "ip_forward",
2169 .data = &ipv4_devconf.data[
02291680 2170 IPV4_DEVCONF_FORWARDING - 1],
68dd299b
PE
2171 .maxlen = sizeof(int),
2172 .mode = 0644,
2173 .proc_handler = devinet_sysctl_forward,
68dd299b 2174 .extra1 = &ipv4_devconf,
c0ce9fb3 2175 .extra2 = &init_net,
68dd299b
PE
2176 },
2177 { },
2178};
2a75de0c 2179#endif
68dd299b 2180
752d14dc
PE
2181static __net_init int devinet_init_net(struct net *net)
2182{
2183 int err;
752d14dc 2184 struct ipv4_devconf *all, *dflt;
2a75de0c
ED
2185#ifdef CONFIG_SYSCTL
2186 struct ctl_table *tbl = ctl_forward_entry;
752d14dc 2187 struct ctl_table_header *forw_hdr;
2a75de0c 2188#endif
752d14dc
PE
2189
2190 err = -ENOMEM;
2191 all = &ipv4_devconf;
2192 dflt = &ipv4_devconf_dflt;
752d14dc 2193
09ad9bc7 2194 if (!net_eq(net, &init_net)) {
752d14dc
PE
2195 all = kmemdup(all, sizeof(ipv4_devconf), GFP_KERNEL);
2196 if (all == NULL)
2197 goto err_alloc_all;
2198
2199 dflt = kmemdup(dflt, sizeof(ipv4_devconf_dflt), GFP_KERNEL);
2200 if (dflt == NULL)
2201 goto err_alloc_dflt;
2202
2a75de0c 2203#ifdef CONFIG_SYSCTL
752d14dc
PE
2204 tbl = kmemdup(tbl, sizeof(ctl_forward_entry), GFP_KERNEL);
2205 if (tbl == NULL)
2206 goto err_alloc_ctl;
2207
02291680 2208 tbl[0].data = &all->data[IPV4_DEVCONF_FORWARDING - 1];
752d14dc
PE
2209 tbl[0].extra1 = all;
2210 tbl[0].extra2 = net;
2a75de0c 2211#endif
752d14dc
PE
2212 }
2213
2214#ifdef CONFIG_SYSCTL
f8572d8f 2215 err = __devinet_sysctl_register(net, "all", all);
752d14dc
PE
2216 if (err < 0)
2217 goto err_reg_all;
2218
f8572d8f 2219 err = __devinet_sysctl_register(net, "default", dflt);
752d14dc
PE
2220 if (err < 0)
2221 goto err_reg_dflt;
2222
2223 err = -ENOMEM;
8607ddb8 2224 forw_hdr = register_net_sysctl(net, "net/ipv4", tbl);
752d14dc
PE
2225 if (forw_hdr == NULL)
2226 goto err_reg_ctl;
2a75de0c 2227 net->ipv4.forw_hdr = forw_hdr;
752d14dc
PE
2228#endif
2229
752d14dc
PE
2230 net->ipv4.devconf_all = all;
2231 net->ipv4.devconf_dflt = dflt;
2232 return 0;
2233
2234#ifdef CONFIG_SYSCTL
2235err_reg_ctl:
2236 __devinet_sysctl_unregister(dflt);
2237err_reg_dflt:
2238 __devinet_sysctl_unregister(all);
2239err_reg_all:
2240 if (tbl != ctl_forward_entry)
2241 kfree(tbl);
752d14dc 2242err_alloc_ctl:
2a75de0c 2243#endif
752d14dc
PE
2244 if (dflt != &ipv4_devconf_dflt)
2245 kfree(dflt);
2246err_alloc_dflt:
2247 if (all != &ipv4_devconf)
2248 kfree(all);
2249err_alloc_all:
2250 return err;
2251}
2252
2253static __net_exit void devinet_exit_net(struct net *net)
2254{
2a75de0c 2255#ifdef CONFIG_SYSCTL
752d14dc
PE
2256 struct ctl_table *tbl;
2257
2258 tbl = net->ipv4.forw_hdr->ctl_table_arg;
752d14dc
PE
2259 unregister_net_sysctl_table(net->ipv4.forw_hdr);
2260 __devinet_sysctl_unregister(net->ipv4.devconf_dflt);
2261 __devinet_sysctl_unregister(net->ipv4.devconf_all);
752d14dc 2262 kfree(tbl);
2a75de0c 2263#endif
752d14dc
PE
2264 kfree(net->ipv4.devconf_dflt);
2265 kfree(net->ipv4.devconf_all);
2266}
2267
2268static __net_initdata struct pernet_operations devinet_ops = {
2269 .init = devinet_init_net,
2270 .exit = devinet_exit_net,
2271};
2272
9f0f7272
TG
2273static struct rtnl_af_ops inet_af_ops = {
2274 .family = AF_INET,
2275 .fill_link_af = inet_fill_link_af,
2276 .get_link_af_size = inet_get_link_af_size,
cf7afbfe
TG
2277 .validate_link_af = inet_validate_link_af,
2278 .set_link_af = inet_set_link_af,
9f0f7272
TG
2279};
2280
1da177e4
LT
2281void __init devinet_init(void)
2282{
fd23c3b3
DM
2283 int i;
2284
2285 for (i = 0; i < IN4_ADDR_HSIZE; i++)
2286 INIT_HLIST_HEAD(&inet_addr_lst[i]);
2287
752d14dc
PE
2288 register_pernet_subsys(&devinet_ops);
2289
1da177e4
LT
2290 register_gifconf(PF_INET, inet_gifconf);
2291 register_netdevice_notifier(&ip_netdev_notifier);
63f3444f 2292
5c766d64
JP
2293 schedule_delayed_work(&check_lifetime_work, 0);
2294
9f0f7272
TG
2295 rtnl_af_register(&inet_af_ops);
2296
c7ac8679
GR
2297 rtnl_register(PF_INET, RTM_NEWADDR, inet_rtm_newaddr, NULL, NULL);
2298 rtnl_register(PF_INET, RTM_DELADDR, inet_rtm_deladdr, NULL, NULL);
2299 rtnl_register(PF_INET, RTM_GETADDR, NULL, inet_dump_ifaddr, NULL);
9e551110 2300 rtnl_register(PF_INET, RTM_GETNETCONF, inet_netconf_get_devconf,
7a674200 2301 inet_netconf_dump_devconf, NULL);
1da177e4
LT
2302}
2303