Merge branch 'for-3.8' of git://linux-nfs.org/~bfields/linux
[GitHub/mt8127/android_kernel_alcatel_ttab.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>
1da177e4 66
406b6f97
DM
67#include "fib_lookup.h"
68
0027ba84 69static struct ipv4_devconf ipv4_devconf = {
42f811b8 70 .data = {
02291680
EB
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,
42f811b8 75 },
1da177e4
LT
76};
77
78static struct ipv4_devconf ipv4_devconf_dflt = {
42f811b8 79 .data = {
02291680
EB
80 [IPV4_DEVCONF_ACCEPT_REDIRECTS - 1] = 1,
81 [IPV4_DEVCONF_SEND_REDIRECTS - 1] = 1,
82 [IPV4_DEVCONF_SECURE_REDIRECTS - 1] = 1,
83 [IPV4_DEVCONF_SHARED_MEDIA - 1] = 1,
84 [IPV4_DEVCONF_ACCEPT_SOURCE_ROUTE - 1] = 1,
42f811b8 85 },
1da177e4
LT
86};
87
9355bbd6
PE
88#define IPV4_DEVCONF_DFLT(net, attr) \
89 IPV4_DEVCONF((*net->ipv4.devconf_dflt), attr)
42f811b8 90
ef7c79ed 91static const struct nla_policy ifa_ipv4_policy[IFA_MAX+1] = {
5c753978
TG
92 [IFA_LOCAL] = { .type = NLA_U32 },
93 [IFA_ADDRESS] = { .type = NLA_U32 },
94 [IFA_BROADCAST] = { .type = NLA_U32 },
5176f91e 95 [IFA_LABEL] = { .type = NLA_STRING, .len = IFNAMSIZ - 1 },
5c753978
TG
96};
97
40384999
ED
98#define IN4_ADDR_HSIZE_SHIFT 8
99#define IN4_ADDR_HSIZE (1U << IN4_ADDR_HSIZE_SHIFT)
100
fd23c3b3
DM
101static struct hlist_head inet_addr_lst[IN4_ADDR_HSIZE];
102static DEFINE_SPINLOCK(inet_addr_hash_lock);
103
40384999 104static u32 inet_addr_hash(struct net *net, __be32 addr)
fd23c3b3 105{
40384999 106 u32 val = (__force u32) addr ^ net_hash_mix(net);
fd23c3b3 107
40384999 108 return hash_32(val, IN4_ADDR_HSIZE_SHIFT);
fd23c3b3
DM
109}
110
111static void inet_hash_insert(struct net *net, struct in_ifaddr *ifa)
112{
40384999 113 u32 hash = inet_addr_hash(net, ifa->ifa_local);
fd23c3b3
DM
114
115 spin_lock(&inet_addr_hash_lock);
116 hlist_add_head_rcu(&ifa->hash, &inet_addr_lst[hash]);
117 spin_unlock(&inet_addr_hash_lock);
118}
119
120static void inet_hash_remove(struct in_ifaddr *ifa)
121{
122 spin_lock(&inet_addr_hash_lock);
123 hlist_del_init_rcu(&ifa->hash);
124 spin_unlock(&inet_addr_hash_lock);
125}
126
9435eb1c
DM
127/**
128 * __ip_dev_find - find the first device with a given source address.
129 * @net: the net namespace
130 * @addr: the source address
131 * @devref: if true, take a reference on the found device
132 *
133 * If a caller uses devref=false, it should be protected by RCU, or RTNL
134 */
135struct net_device *__ip_dev_find(struct net *net, __be32 addr, bool devref)
136{
40384999 137 u32 hash = inet_addr_hash(net, addr);
9435eb1c
DM
138 struct net_device *result = NULL;
139 struct in_ifaddr *ifa;
140 struct hlist_node *node;
141
142 rcu_read_lock();
143 hlist_for_each_entry_rcu(ifa, node, &inet_addr_lst[hash], hash) {
e066008b 144 if (ifa->ifa_local == addr) {
40384999
ED
145 struct net_device *dev = ifa->ifa_dev->dev;
146
147 if (!net_eq(dev_net(dev), net))
148 continue;
9435eb1c
DM
149 result = dev;
150 break;
151 }
152 }
406b6f97
DM
153 if (!result) {
154 struct flowi4 fl4 = { .daddr = addr };
155 struct fib_result res = { 0 };
156 struct fib_table *local;
157
158 /* Fallback to FIB local table so that communication
159 * over loopback subnets work.
160 */
161 local = fib_get_table(net, RT_TABLE_LOCAL);
162 if (local &&
163 !fib_table_lookup(local, &fl4, &res, FIB_LOOKUP_NOREF) &&
164 res.type == RTN_LOCAL)
165 result = FIB_RES_DEV(res);
166 }
9435eb1c
DM
167 if (result && devref)
168 dev_hold(result);
169 rcu_read_unlock();
170 return result;
171}
172EXPORT_SYMBOL(__ip_dev_find);
173
d6062cbb 174static void rtmsg_ifa(int event, struct in_ifaddr *, struct nlmsghdr *, u32);
1da177e4 175
e041c683 176static BLOCKING_NOTIFIER_HEAD(inetaddr_chain);
1da177e4
LT
177static void inet_del_ifa(struct in_device *in_dev, struct in_ifaddr **ifap,
178 int destroy);
179#ifdef CONFIG_SYSCTL
66f27a52 180static void devinet_sysctl_register(struct in_device *idev);
51602b2a
PE
181static void devinet_sysctl_unregister(struct in_device *idev);
182#else
40384999 183static void devinet_sysctl_register(struct in_device *idev)
51602b2a
PE
184{
185}
40384999 186static void devinet_sysctl_unregister(struct in_device *idev)
51602b2a
PE
187{
188}
1da177e4
LT
189#endif
190
191/* Locks all the inet devices. */
192
193static struct in_ifaddr *inet_alloc_ifa(void)
194{
93adcc80 195 return kzalloc(sizeof(struct in_ifaddr), GFP_KERNEL);
1da177e4
LT
196}
197
198static void inet_rcu_free_ifa(struct rcu_head *head)
199{
200 struct in_ifaddr *ifa = container_of(head, struct in_ifaddr, rcu_head);
201 if (ifa->ifa_dev)
202 in_dev_put(ifa->ifa_dev);
203 kfree(ifa);
204}
205
40384999 206static void inet_free_ifa(struct in_ifaddr *ifa)
1da177e4
LT
207{
208 call_rcu(&ifa->rcu_head, inet_rcu_free_ifa);
209}
210
211void in_dev_finish_destroy(struct in_device *idev)
212{
213 struct net_device *dev = idev->dev;
214
547b792c
IJ
215 WARN_ON(idev->ifa_list);
216 WARN_ON(idev->mc_list);
1da177e4 217#ifdef NET_REFCNT_DEBUG
91df42be 218 pr_debug("%s: %p=%s\n", __func__, idev, dev ? dev->name : "NIL");
1da177e4
LT
219#endif
220 dev_put(dev);
221 if (!idev->dead)
9f9354b9
ED
222 pr_err("Freeing alive in_device %p\n", idev);
223 else
1da177e4 224 kfree(idev);
1da177e4 225}
9f9354b9 226EXPORT_SYMBOL(in_dev_finish_destroy);
1da177e4 227
71e27da9 228static struct in_device *inetdev_init(struct net_device *dev)
1da177e4
LT
229{
230 struct in_device *in_dev;
231
232 ASSERT_RTNL();
233
0da974f4 234 in_dev = kzalloc(sizeof(*in_dev), GFP_KERNEL);
1da177e4
LT
235 if (!in_dev)
236 goto out;
c346dca1 237 memcpy(&in_dev->cnf, dev_net(dev)->ipv4.devconf_dflt,
9355bbd6 238 sizeof(in_dev->cnf));
1da177e4
LT
239 in_dev->cnf.sysctl = NULL;
240 in_dev->dev = dev;
9f9354b9
ED
241 in_dev->arp_parms = neigh_parms_alloc(dev, &arp_tbl);
242 if (!in_dev->arp_parms)
1da177e4 243 goto out_kfree;
0187bdfb
BH
244 if (IPV4_DEVCONF(in_dev->cnf, FORWARDING))
245 dev_disable_lro(dev);
1da177e4
LT
246 /* Reference in_dev->dev */
247 dev_hold(dev);
30c4cf57 248 /* Account for reference dev->ip_ptr (below) */
1da177e4 249 in_dev_hold(in_dev);
1da177e4 250
66f27a52 251 devinet_sysctl_register(in_dev);
1da177e4
LT
252 ip_mc_init_dev(in_dev);
253 if (dev->flags & IFF_UP)
254 ip_mc_up(in_dev);
483479ec 255
30c4cf57 256 /* we can receive as soon as ip_ptr is set -- do this last */
cf778b00 257 rcu_assign_pointer(dev->ip_ptr, in_dev);
483479ec 258out:
1da177e4
LT
259 return in_dev;
260out_kfree:
261 kfree(in_dev);
262 in_dev = NULL;
263 goto out;
264}
265
266static void in_dev_rcu_put(struct rcu_head *head)
267{
268 struct in_device *idev = container_of(head, struct in_device, rcu_head);
269 in_dev_put(idev);
270}
271
272static void inetdev_destroy(struct in_device *in_dev)
273{
274 struct in_ifaddr *ifa;
275 struct net_device *dev;
276
277 ASSERT_RTNL();
278
279 dev = in_dev->dev;
1da177e4
LT
280
281 in_dev->dead = 1;
282
283 ip_mc_destroy_dev(in_dev);
284
285 while ((ifa = in_dev->ifa_list) != NULL) {
286 inet_del_ifa(in_dev, &in_dev->ifa_list, 0);
287 inet_free_ifa(ifa);
288 }
289
a9b3cd7f 290 RCU_INIT_POINTER(dev->ip_ptr, NULL);
1da177e4 291
51602b2a 292 devinet_sysctl_unregister(in_dev);
1da177e4
LT
293 neigh_parms_release(&arp_tbl, in_dev->arp_parms);
294 arp_ifdown(dev);
295
296 call_rcu(&in_dev->rcu_head, in_dev_rcu_put);
297}
298
ff428d72 299int inet_addr_onlink(struct in_device *in_dev, __be32 a, __be32 b)
1da177e4
LT
300{
301 rcu_read_lock();
302 for_primary_ifa(in_dev) {
303 if (inet_ifa_match(a, ifa)) {
304 if (!b || inet_ifa_match(b, ifa)) {
305 rcu_read_unlock();
306 return 1;
307 }
308 }
309 } endfor_ifa(in_dev);
310 rcu_read_unlock();
311 return 0;
312}
313
d6062cbb 314static void __inet_del_ifa(struct in_device *in_dev, struct in_ifaddr **ifap,
15e47304 315 int destroy, struct nlmsghdr *nlh, u32 portid)
1da177e4 316{
8f937c60 317 struct in_ifaddr *promote = NULL;
0ff60a45
JHS
318 struct in_ifaddr *ifa, *ifa1 = *ifap;
319 struct in_ifaddr *last_prim = in_dev->ifa_list;
320 struct in_ifaddr *prev_prom = NULL;
321 int do_promote = IN_DEV_PROMOTE_SECONDARIES(in_dev);
1da177e4
LT
322
323 ASSERT_RTNL();
324
e905a9ed 325 /* 1. Deleting primary ifaddr forces deletion all secondaries
8f937c60
HW
326 * unless alias promotion is set
327 **/
1da177e4
LT
328
329 if (!(ifa1->ifa_flags & IFA_F_SECONDARY)) {
1da177e4
LT
330 struct in_ifaddr **ifap1 = &ifa1->ifa_next;
331
332 while ((ifa = *ifap1) != NULL) {
e905a9ed 333 if (!(ifa->ifa_flags & IFA_F_SECONDARY) &&
0ff60a45
JHS
334 ifa1->ifa_scope <= ifa->ifa_scope)
335 last_prim = ifa;
336
1da177e4
LT
337 if (!(ifa->ifa_flags & IFA_F_SECONDARY) ||
338 ifa1->ifa_mask != ifa->ifa_mask ||
339 !inet_ifa_match(ifa1->ifa_address, ifa)) {
340 ifap1 = &ifa->ifa_next;
0ff60a45 341 prev_prom = ifa;
1da177e4
LT
342 continue;
343 }
344
0ff60a45 345 if (!do_promote) {
fd23c3b3 346 inet_hash_remove(ifa);
8f937c60 347 *ifap1 = ifa->ifa_next;
1da177e4 348
15e47304 349 rtmsg_ifa(RTM_DELADDR, ifa, nlh, portid);
e041c683
AS
350 blocking_notifier_call_chain(&inetaddr_chain,
351 NETDEV_DOWN, ifa);
8f937c60
HW
352 inet_free_ifa(ifa);
353 } else {
354 promote = ifa;
355 break;
356 }
1da177e4
LT
357 }
358 }
359
2d230e2b
JA
360 /* On promotion all secondaries from subnet are changing
361 * the primary IP, we must remove all their routes silently
362 * and later to add them back with new prefsrc. Do this
363 * while all addresses are on the device list.
364 */
365 for (ifa = promote; ifa; ifa = ifa->ifa_next) {
366 if (ifa1->ifa_mask == ifa->ifa_mask &&
367 inet_ifa_match(ifa1->ifa_address, ifa))
368 fib_del_ifaddr(ifa, ifa1);
369 }
370
1da177e4
LT
371 /* 2. Unlink it */
372
373 *ifap = ifa1->ifa_next;
fd23c3b3 374 inet_hash_remove(ifa1);
1da177e4
LT
375
376 /* 3. Announce address deletion */
377
378 /* Send message first, then call notifier.
379 At first sight, FIB update triggered by notifier
380 will refer to already deleted ifaddr, that could confuse
381 netlink listeners. It is not true: look, gated sees
382 that route deleted and if it still thinks that ifaddr
383 is valid, it will try to restore deleted routes... Grr.
384 So that, this order is correct.
385 */
15e47304 386 rtmsg_ifa(RTM_DELADDR, ifa1, nlh, portid);
e041c683 387 blocking_notifier_call_chain(&inetaddr_chain, NETDEV_DOWN, ifa1);
1da177e4 388
0ff60a45 389 if (promote) {
04024b93 390 struct in_ifaddr *next_sec = promote->ifa_next;
0ff60a45
JHS
391
392 if (prev_prom) {
393 prev_prom->ifa_next = promote->ifa_next;
394 promote->ifa_next = last_prim->ifa_next;
395 last_prim->ifa_next = promote;
396 }
8f937c60 397
8f937c60 398 promote->ifa_flags &= ~IFA_F_SECONDARY;
15e47304 399 rtmsg_ifa(RTM_NEWADDR, promote, nlh, portid);
e041c683
AS
400 blocking_notifier_call_chain(&inetaddr_chain,
401 NETDEV_UP, promote);
04024b93 402 for (ifa = next_sec; ifa; ifa = ifa->ifa_next) {
0ff60a45
JHS
403 if (ifa1->ifa_mask != ifa->ifa_mask ||
404 !inet_ifa_match(ifa1->ifa_address, ifa))
405 continue;
406 fib_add_ifaddr(ifa);
407 }
408
409 }
6363097c 410 if (destroy)
0ff60a45 411 inet_free_ifa(ifa1);
1da177e4
LT
412}
413
d6062cbb
TG
414static void inet_del_ifa(struct in_device *in_dev, struct in_ifaddr **ifap,
415 int destroy)
416{
417 __inet_del_ifa(in_dev, ifap, destroy, NULL, 0);
418}
419
420static int __inet_insert_ifa(struct in_ifaddr *ifa, struct nlmsghdr *nlh,
15e47304 421 u32 portid)
1da177e4
LT
422{
423 struct in_device *in_dev = ifa->ifa_dev;
424 struct in_ifaddr *ifa1, **ifap, **last_primary;
425
426 ASSERT_RTNL();
427
428 if (!ifa->ifa_local) {
429 inet_free_ifa(ifa);
430 return 0;
431 }
432
433 ifa->ifa_flags &= ~IFA_F_SECONDARY;
434 last_primary = &in_dev->ifa_list;
435
436 for (ifap = &in_dev->ifa_list; (ifa1 = *ifap) != NULL;
437 ifap = &ifa1->ifa_next) {
438 if (!(ifa1->ifa_flags & IFA_F_SECONDARY) &&
439 ifa->ifa_scope <= ifa1->ifa_scope)
440 last_primary = &ifa1->ifa_next;
441 if (ifa1->ifa_mask == ifa->ifa_mask &&
442 inet_ifa_match(ifa1->ifa_address, ifa)) {
443 if (ifa1->ifa_local == ifa->ifa_local) {
444 inet_free_ifa(ifa);
445 return -EEXIST;
446 }
447 if (ifa1->ifa_scope != ifa->ifa_scope) {
448 inet_free_ifa(ifa);
449 return -EINVAL;
450 }
451 ifa->ifa_flags |= IFA_F_SECONDARY;
452 }
453 }
454
455 if (!(ifa->ifa_flags & IFA_F_SECONDARY)) {
456 net_srandom(ifa->ifa_local);
457 ifap = last_primary;
458 }
459
460 ifa->ifa_next = *ifap;
461 *ifap = ifa;
462
fd23c3b3
DM
463 inet_hash_insert(dev_net(in_dev->dev), ifa);
464
1da177e4
LT
465 /* Send message first, then call notifier.
466 Notifier will trigger FIB update, so that
467 listeners of netlink will know about new ifaddr */
15e47304 468 rtmsg_ifa(RTM_NEWADDR, ifa, nlh, portid);
e041c683 469 blocking_notifier_call_chain(&inetaddr_chain, NETDEV_UP, ifa);
1da177e4
LT
470
471 return 0;
472}
473
d6062cbb
TG
474static int inet_insert_ifa(struct in_ifaddr *ifa)
475{
476 return __inet_insert_ifa(ifa, NULL, 0);
477}
478
1da177e4
LT
479static int inet_set_ifa(struct net_device *dev, struct in_ifaddr *ifa)
480{
e5ed6399 481 struct in_device *in_dev = __in_dev_get_rtnl(dev);
1da177e4
LT
482
483 ASSERT_RTNL();
484
485 if (!in_dev) {
71e27da9
HX
486 inet_free_ifa(ifa);
487 return -ENOBUFS;
1da177e4 488 }
71e27da9 489 ipv4_devconf_setall(in_dev);
1da177e4 490 if (ifa->ifa_dev != in_dev) {
547b792c 491 WARN_ON(ifa->ifa_dev);
1da177e4
LT
492 in_dev_hold(in_dev);
493 ifa->ifa_dev = in_dev;
494 }
f97c1e0c 495 if (ipv4_is_loopback(ifa->ifa_local))
1da177e4
LT
496 ifa->ifa_scope = RT_SCOPE_HOST;
497 return inet_insert_ifa(ifa);
498}
499
8723e1b4
ED
500/* Caller must hold RCU or RTNL :
501 * We dont take a reference on found in_device
502 */
7fee0ca2 503struct in_device *inetdev_by_index(struct net *net, int ifindex)
1da177e4
LT
504{
505 struct net_device *dev;
506 struct in_device *in_dev = NULL;
c148fc2e
ED
507
508 rcu_read_lock();
509 dev = dev_get_by_index_rcu(net, ifindex);
1da177e4 510 if (dev)
8723e1b4 511 in_dev = rcu_dereference_rtnl(dev->ip_ptr);
c148fc2e 512 rcu_read_unlock();
1da177e4
LT
513 return in_dev;
514}
9f9354b9 515EXPORT_SYMBOL(inetdev_by_index);
1da177e4
LT
516
517/* Called only from RTNL semaphored context. No locks. */
518
60cad5da
AV
519struct in_ifaddr *inet_ifa_byprefix(struct in_device *in_dev, __be32 prefix,
520 __be32 mask)
1da177e4
LT
521{
522 ASSERT_RTNL();
523
524 for_primary_ifa(in_dev) {
525 if (ifa->ifa_mask == mask && inet_ifa_match(prefix, ifa))
526 return ifa;
527 } endfor_ifa(in_dev);
528 return NULL;
529}
530
531static int inet_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
532{
3b1e0a65 533 struct net *net = sock_net(skb->sk);
dfdd5fd4 534 struct nlattr *tb[IFA_MAX+1];
1da177e4 535 struct in_device *in_dev;
dfdd5fd4 536 struct ifaddrmsg *ifm;
1da177e4 537 struct in_ifaddr *ifa, **ifap;
dfdd5fd4 538 int err = -EINVAL;
1da177e4
LT
539
540 ASSERT_RTNL();
541
dfdd5fd4
TG
542 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv4_policy);
543 if (err < 0)
544 goto errout;
545
546 ifm = nlmsg_data(nlh);
7fee0ca2 547 in_dev = inetdev_by_index(net, ifm->ifa_index);
dfdd5fd4
TG
548 if (in_dev == NULL) {
549 err = -ENODEV;
550 goto errout;
551 }
552
1da177e4
LT
553 for (ifap = &in_dev->ifa_list; (ifa = *ifap) != NULL;
554 ifap = &ifa->ifa_next) {
dfdd5fd4 555 if (tb[IFA_LOCAL] &&
a7a628c4 556 ifa->ifa_local != nla_get_be32(tb[IFA_LOCAL]))
dfdd5fd4
TG
557 continue;
558
559 if (tb[IFA_LABEL] && nla_strcmp(tb[IFA_LABEL], ifa->ifa_label))
1da177e4 560 continue;
dfdd5fd4
TG
561
562 if (tb[IFA_ADDRESS] &&
563 (ifm->ifa_prefixlen != ifa->ifa_prefixlen ||
a7a628c4 564 !inet_ifa_match(nla_get_be32(tb[IFA_ADDRESS]), ifa)))
dfdd5fd4
TG
565 continue;
566
15e47304 567 __inet_del_ifa(in_dev, ifap, 1, nlh, NETLINK_CB(skb).portid);
1da177e4
LT
568 return 0;
569 }
dfdd5fd4
TG
570
571 err = -EADDRNOTAVAIL;
572errout:
573 return err;
1da177e4
LT
574}
575
4b8aa9ab 576static struct in_ifaddr *rtm_to_ifaddr(struct net *net, struct nlmsghdr *nlh)
1da177e4 577{
5c753978
TG
578 struct nlattr *tb[IFA_MAX+1];
579 struct in_ifaddr *ifa;
580 struct ifaddrmsg *ifm;
1da177e4
LT
581 struct net_device *dev;
582 struct in_device *in_dev;
7b218574 583 int err;
1da177e4 584
5c753978
TG
585 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv4_policy);
586 if (err < 0)
587 goto errout;
1da177e4 588
5c753978 589 ifm = nlmsg_data(nlh);
7b218574
DL
590 err = -EINVAL;
591 if (ifm->ifa_prefixlen > 32 || tb[IFA_LOCAL] == NULL)
5c753978 592 goto errout;
1da177e4 593
4b8aa9ab 594 dev = __dev_get_by_index(net, ifm->ifa_index);
7b218574
DL
595 err = -ENODEV;
596 if (dev == NULL)
5c753978 597 goto errout;
1da177e4 598
5c753978 599 in_dev = __in_dev_get_rtnl(dev);
7b218574
DL
600 err = -ENOBUFS;
601 if (in_dev == NULL)
71e27da9 602 goto errout;
1da177e4 603
5c753978 604 ifa = inet_alloc_ifa();
7b218574 605 if (ifa == NULL)
5c753978
TG
606 /*
607 * A potential indev allocation can be left alive, it stays
608 * assigned to its device and is destroy with it.
609 */
5c753978 610 goto errout;
5c753978 611
a4e65d36 612 ipv4_devconf_setall(in_dev);
5c753978
TG
613 in_dev_hold(in_dev);
614
615 if (tb[IFA_ADDRESS] == NULL)
616 tb[IFA_ADDRESS] = tb[IFA_LOCAL];
1da177e4 617
fd23c3b3 618 INIT_HLIST_NODE(&ifa->hash);
1da177e4
LT
619 ifa->ifa_prefixlen = ifm->ifa_prefixlen;
620 ifa->ifa_mask = inet_make_mask(ifm->ifa_prefixlen);
1da177e4
LT
621 ifa->ifa_flags = ifm->ifa_flags;
622 ifa->ifa_scope = ifm->ifa_scope;
5c753978
TG
623 ifa->ifa_dev = in_dev;
624
a7a628c4
AV
625 ifa->ifa_local = nla_get_be32(tb[IFA_LOCAL]);
626 ifa->ifa_address = nla_get_be32(tb[IFA_ADDRESS]);
5c753978
TG
627
628 if (tb[IFA_BROADCAST])
a7a628c4 629 ifa->ifa_broadcast = nla_get_be32(tb[IFA_BROADCAST]);
5c753978 630
5c753978
TG
631 if (tb[IFA_LABEL])
632 nla_strlcpy(ifa->ifa_label, tb[IFA_LABEL], IFNAMSIZ);
1da177e4
LT
633 else
634 memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
635
5c753978
TG
636 return ifa;
637
638errout:
639 return ERR_PTR(err);
640}
641
642static int inet_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
643{
3b1e0a65 644 struct net *net = sock_net(skb->sk);
5c753978
TG
645 struct in_ifaddr *ifa;
646
647 ASSERT_RTNL();
648
4b8aa9ab 649 ifa = rtm_to_ifaddr(net, nlh);
5c753978
TG
650 if (IS_ERR(ifa))
651 return PTR_ERR(ifa);
652
15e47304 653 return __inet_insert_ifa(ifa, nlh, NETLINK_CB(skb).portid);
1da177e4
LT
654}
655
656/*
657 * Determine a default network mask, based on the IP address.
658 */
659
40384999 660static int inet_abc_len(__be32 addr)
1da177e4
LT
661{
662 int rc = -1; /* Something else, probably a multicast. */
663
f97c1e0c 664 if (ipv4_is_zeronet(addr))
e905a9ed 665 rc = 0;
1da177e4 666 else {
714e85be 667 __u32 haddr = ntohl(addr);
1da177e4 668
714e85be 669 if (IN_CLASSA(haddr))
1da177e4 670 rc = 8;
714e85be 671 else if (IN_CLASSB(haddr))
1da177e4 672 rc = 16;
714e85be 673 else if (IN_CLASSC(haddr))
1da177e4
LT
674 rc = 24;
675 }
676
e905a9ed 677 return rc;
1da177e4
LT
678}
679
680
e5b13cb1 681int devinet_ioctl(struct net *net, unsigned int cmd, void __user *arg)
1da177e4
LT
682{
683 struct ifreq ifr;
684 struct sockaddr_in sin_orig;
685 struct sockaddr_in *sin = (struct sockaddr_in *)&ifr.ifr_addr;
686 struct in_device *in_dev;
687 struct in_ifaddr **ifap = NULL;
688 struct in_ifaddr *ifa = NULL;
689 struct net_device *dev;
690 char *colon;
691 int ret = -EFAULT;
692 int tryaddrmatch = 0;
693
694 /*
695 * Fetch the caller's info block into kernel space
696 */
697
698 if (copy_from_user(&ifr, arg, sizeof(struct ifreq)))
699 goto out;
700 ifr.ifr_name[IFNAMSIZ - 1] = 0;
701
702 /* save original address for comparison */
703 memcpy(&sin_orig, sin, sizeof(*sin));
704
705 colon = strchr(ifr.ifr_name, ':');
706 if (colon)
707 *colon = 0;
708
e5b13cb1 709 dev_load(net, ifr.ifr_name);
1da177e4 710
132adf54 711 switch (cmd) {
1da177e4
LT
712 case SIOCGIFADDR: /* Get interface address */
713 case SIOCGIFBRDADDR: /* Get the broadcast address */
714 case SIOCGIFDSTADDR: /* Get the destination address */
715 case SIOCGIFNETMASK: /* Get the netmask for the interface */
716 /* Note that these ioctls will not sleep,
717 so that we do not impose a lock.
718 One day we will be forced to put shlock here (I mean SMP)
719 */
720 tryaddrmatch = (sin_orig.sin_family == AF_INET);
721 memset(sin, 0, sizeof(*sin));
722 sin->sin_family = AF_INET;
723 break;
724
725 case SIOCSIFFLAGS:
bf5b30b8 726 ret = -EPERM;
52e804c6 727 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1da177e4
LT
728 goto out;
729 break;
730 case SIOCSIFADDR: /* Set interface address (and family) */
731 case SIOCSIFBRDADDR: /* Set the broadcast address */
732 case SIOCSIFDSTADDR: /* Set the destination address */
733 case SIOCSIFNETMASK: /* Set the netmask for the interface */
bf5b30b8 734 ret = -EPERM;
52e804c6 735 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1da177e4
LT
736 goto out;
737 ret = -EINVAL;
738 if (sin->sin_family != AF_INET)
739 goto out;
740 break;
741 default:
742 ret = -EINVAL;
743 goto out;
744 }
745
746 rtnl_lock();
747
748 ret = -ENODEV;
9f9354b9
ED
749 dev = __dev_get_by_name(net, ifr.ifr_name);
750 if (!dev)
1da177e4
LT
751 goto done;
752
753 if (colon)
754 *colon = ':';
755
9f9354b9
ED
756 in_dev = __in_dev_get_rtnl(dev);
757 if (in_dev) {
1da177e4
LT
758 if (tryaddrmatch) {
759 /* Matthias Andree */
760 /* compare label and address (4.4BSD style) */
761 /* note: we only do this for a limited set of ioctls
762 and only if the original address family was AF_INET.
763 This is checked above. */
764 for (ifap = &in_dev->ifa_list; (ifa = *ifap) != NULL;
765 ifap = &ifa->ifa_next) {
766 if (!strcmp(ifr.ifr_name, ifa->ifa_label) &&
767 sin_orig.sin_addr.s_addr ==
6c91afe1 768 ifa->ifa_local) {
1da177e4
LT
769 break; /* found */
770 }
771 }
772 }
773 /* we didn't get a match, maybe the application is
774 4.3BSD-style and passed in junk so we fall back to
775 comparing just the label */
776 if (!ifa) {
777 for (ifap = &in_dev->ifa_list; (ifa = *ifap) != NULL;
778 ifap = &ifa->ifa_next)
779 if (!strcmp(ifr.ifr_name, ifa->ifa_label))
780 break;
781 }
782 }
783
784 ret = -EADDRNOTAVAIL;
785 if (!ifa && cmd != SIOCSIFADDR && cmd != SIOCSIFFLAGS)
786 goto done;
787
132adf54 788 switch (cmd) {
1da177e4
LT
789 case SIOCGIFADDR: /* Get interface address */
790 sin->sin_addr.s_addr = ifa->ifa_local;
791 goto rarok;
792
793 case SIOCGIFBRDADDR: /* Get the broadcast address */
794 sin->sin_addr.s_addr = ifa->ifa_broadcast;
795 goto rarok;
796
797 case SIOCGIFDSTADDR: /* Get the destination address */
798 sin->sin_addr.s_addr = ifa->ifa_address;
799 goto rarok;
800
801 case SIOCGIFNETMASK: /* Get the netmask for the interface */
802 sin->sin_addr.s_addr = ifa->ifa_mask;
803 goto rarok;
804
805 case SIOCSIFFLAGS:
806 if (colon) {
807 ret = -EADDRNOTAVAIL;
808 if (!ifa)
809 break;
810 ret = 0;
811 if (!(ifr.ifr_flags & IFF_UP))
812 inet_del_ifa(in_dev, ifap, 1);
813 break;
814 }
815 ret = dev_change_flags(dev, ifr.ifr_flags);
816 break;
817
818 case SIOCSIFADDR: /* Set interface address (and family) */
819 ret = -EINVAL;
820 if (inet_abc_len(sin->sin_addr.s_addr) < 0)
821 break;
822
823 if (!ifa) {
824 ret = -ENOBUFS;
9f9354b9 825 ifa = inet_alloc_ifa();
fd23c3b3 826 INIT_HLIST_NODE(&ifa->hash);
9f9354b9 827 if (!ifa)
1da177e4
LT
828 break;
829 if (colon)
830 memcpy(ifa->ifa_label, ifr.ifr_name, IFNAMSIZ);
831 else
832 memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
833 } else {
834 ret = 0;
835 if (ifa->ifa_local == sin->sin_addr.s_addr)
836 break;
837 inet_del_ifa(in_dev, ifap, 0);
838 ifa->ifa_broadcast = 0;
148f9729 839 ifa->ifa_scope = 0;
1da177e4
LT
840 }
841
842 ifa->ifa_address = ifa->ifa_local = sin->sin_addr.s_addr;
843
844 if (!(dev->flags & IFF_POINTOPOINT)) {
845 ifa->ifa_prefixlen = inet_abc_len(ifa->ifa_address);
846 ifa->ifa_mask = inet_make_mask(ifa->ifa_prefixlen);
847 if ((dev->flags & IFF_BROADCAST) &&
848 ifa->ifa_prefixlen < 31)
849 ifa->ifa_broadcast = ifa->ifa_address |
850 ~ifa->ifa_mask;
851 } else {
852 ifa->ifa_prefixlen = 32;
853 ifa->ifa_mask = inet_make_mask(32);
854 }
855 ret = inet_set_ifa(dev, ifa);
856 break;
857
858 case SIOCSIFBRDADDR: /* Set the broadcast address */
859 ret = 0;
860 if (ifa->ifa_broadcast != sin->sin_addr.s_addr) {
861 inet_del_ifa(in_dev, ifap, 0);
862 ifa->ifa_broadcast = sin->sin_addr.s_addr;
863 inet_insert_ifa(ifa);
864 }
865 break;
866
867 case SIOCSIFDSTADDR: /* Set the destination address */
868 ret = 0;
869 if (ifa->ifa_address == sin->sin_addr.s_addr)
870 break;
871 ret = -EINVAL;
872 if (inet_abc_len(sin->sin_addr.s_addr) < 0)
873 break;
874 ret = 0;
875 inet_del_ifa(in_dev, ifap, 0);
876 ifa->ifa_address = sin->sin_addr.s_addr;
877 inet_insert_ifa(ifa);
878 break;
879
880 case SIOCSIFNETMASK: /* Set the netmask for the interface */
881
882 /*
883 * The mask we set must be legal.
884 */
885 ret = -EINVAL;
886 if (bad_mask(sin->sin_addr.s_addr, 0))
887 break;
888 ret = 0;
889 if (ifa->ifa_mask != sin->sin_addr.s_addr) {
a144ea4b 890 __be32 old_mask = ifa->ifa_mask;
1da177e4
LT
891 inet_del_ifa(in_dev, ifap, 0);
892 ifa->ifa_mask = sin->sin_addr.s_addr;
893 ifa->ifa_prefixlen = inet_mask_len(ifa->ifa_mask);
894
895 /* See if current broadcast address matches
896 * with current netmask, then recalculate
897 * the broadcast address. Otherwise it's a
898 * funny address, so don't touch it since
899 * the user seems to know what (s)he's doing...
900 */
901 if ((dev->flags & IFF_BROADCAST) &&
902 (ifa->ifa_prefixlen < 31) &&
903 (ifa->ifa_broadcast ==
dcab5e1e 904 (ifa->ifa_local|~old_mask))) {
1da177e4
LT
905 ifa->ifa_broadcast = (ifa->ifa_local |
906 ~sin->sin_addr.s_addr);
907 }
908 inet_insert_ifa(ifa);
909 }
910 break;
911 }
912done:
913 rtnl_unlock();
914out:
915 return ret;
916rarok:
917 rtnl_unlock();
918 ret = copy_to_user(arg, &ifr, sizeof(struct ifreq)) ? -EFAULT : 0;
919 goto out;
920}
921
922static int inet_gifconf(struct net_device *dev, char __user *buf, int len)
923{
e5ed6399 924 struct in_device *in_dev = __in_dev_get_rtnl(dev);
1da177e4
LT
925 struct in_ifaddr *ifa;
926 struct ifreq ifr;
927 int done = 0;
928
9f9354b9 929 if (!in_dev)
1da177e4
LT
930 goto out;
931
9f9354b9 932 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
1da177e4
LT
933 if (!buf) {
934 done += sizeof(ifr);
935 continue;
936 }
937 if (len < (int) sizeof(ifr))
938 break;
939 memset(&ifr, 0, sizeof(struct ifreq));
940 if (ifa->ifa_label)
941 strcpy(ifr.ifr_name, ifa->ifa_label);
942 else
943 strcpy(ifr.ifr_name, dev->name);
944
945 (*(struct sockaddr_in *)&ifr.ifr_addr).sin_family = AF_INET;
946 (*(struct sockaddr_in *)&ifr.ifr_addr).sin_addr.s_addr =
947 ifa->ifa_local;
948
949 if (copy_to_user(buf, &ifr, sizeof(struct ifreq))) {
950 done = -EFAULT;
951 break;
952 }
953 buf += sizeof(struct ifreq);
954 len -= sizeof(struct ifreq);
955 done += sizeof(struct ifreq);
956 }
957out:
958 return done;
959}
960
a61ced5d 961__be32 inet_select_addr(const struct net_device *dev, __be32 dst, int scope)
1da177e4 962{
a61ced5d 963 __be32 addr = 0;
1da177e4 964 struct in_device *in_dev;
c346dca1 965 struct net *net = dev_net(dev);
1da177e4
LT
966
967 rcu_read_lock();
e5ed6399 968 in_dev = __in_dev_get_rcu(dev);
1da177e4
LT
969 if (!in_dev)
970 goto no_in_dev;
971
972 for_primary_ifa(in_dev) {
973 if (ifa->ifa_scope > scope)
974 continue;
975 if (!dst || inet_ifa_match(dst, ifa)) {
976 addr = ifa->ifa_local;
977 break;
978 }
979 if (!addr)
980 addr = ifa->ifa_local;
981 } endfor_ifa(in_dev);
1da177e4
LT
982
983 if (addr)
c6d14c84 984 goto out_unlock;
9f9354b9 985no_in_dev:
1da177e4
LT
986
987 /* Not loopback addresses on loopback should be preferred
988 in this case. It is importnat that lo is the first interface
989 in dev_base list.
990 */
c6d14c84 991 for_each_netdev_rcu(net, dev) {
9f9354b9
ED
992 in_dev = __in_dev_get_rcu(dev);
993 if (!in_dev)
1da177e4
LT
994 continue;
995
996 for_primary_ifa(in_dev) {
997 if (ifa->ifa_scope != RT_SCOPE_LINK &&
998 ifa->ifa_scope <= scope) {
999 addr = ifa->ifa_local;
c6d14c84 1000 goto out_unlock;
1da177e4
LT
1001 }
1002 } endfor_ifa(in_dev);
1003 }
c6d14c84 1004out_unlock:
1da177e4 1005 rcu_read_unlock();
1da177e4
LT
1006 return addr;
1007}
9f9354b9 1008EXPORT_SYMBOL(inet_select_addr);
1da177e4 1009
60cad5da
AV
1010static __be32 confirm_addr_indev(struct in_device *in_dev, __be32 dst,
1011 __be32 local, int scope)
1da177e4
LT
1012{
1013 int same = 0;
a144ea4b 1014 __be32 addr = 0;
1da177e4
LT
1015
1016 for_ifa(in_dev) {
1017 if (!addr &&
1018 (local == ifa->ifa_local || !local) &&
1019 ifa->ifa_scope <= scope) {
1020 addr = ifa->ifa_local;
1021 if (same)
1022 break;
1023 }
1024 if (!same) {
1025 same = (!local || inet_ifa_match(local, ifa)) &&
1026 (!dst || inet_ifa_match(dst, ifa));
1027 if (same && addr) {
1028 if (local || !dst)
1029 break;
1030 /* Is the selected addr into dst subnet? */
1031 if (inet_ifa_match(addr, ifa))
1032 break;
1033 /* No, then can we use new local src? */
1034 if (ifa->ifa_scope <= scope) {
1035 addr = ifa->ifa_local;
1036 break;
1037 }
1038 /* search for large dst subnet for addr */
1039 same = 0;
1040 }
1041 }
1042 } endfor_ifa(in_dev);
1043
9f9354b9 1044 return same ? addr : 0;
1da177e4
LT
1045}
1046
1047/*
1048 * Confirm that local IP address exists using wildcards:
9bd85e32 1049 * - in_dev: only on this interface, 0=any interface
1da177e4
LT
1050 * - dst: only in the same subnet as dst, 0=any dst
1051 * - local: address, 0=autoselect the local address
1052 * - scope: maximum allowed scope value for the local address
1053 */
9bd85e32
DL
1054__be32 inet_confirm_addr(struct in_device *in_dev,
1055 __be32 dst, __be32 local, int scope)
1da177e4 1056{
60cad5da 1057 __be32 addr = 0;
9bd85e32 1058 struct net_device *dev;
39a6d063 1059 struct net *net;
1da177e4 1060
39a6d063 1061 if (scope != RT_SCOPE_LINK)
9bd85e32 1062 return confirm_addr_indev(in_dev, dst, local, scope);
1da177e4 1063
c346dca1 1064 net = dev_net(in_dev->dev);
1da177e4 1065 rcu_read_lock();
c6d14c84 1066 for_each_netdev_rcu(net, dev) {
9f9354b9
ED
1067 in_dev = __in_dev_get_rcu(dev);
1068 if (in_dev) {
1da177e4
LT
1069 addr = confirm_addr_indev(in_dev, dst, local, scope);
1070 if (addr)
1071 break;
1072 }
1073 }
1074 rcu_read_unlock();
1da177e4
LT
1075
1076 return addr;
1077}
eaddcd76 1078EXPORT_SYMBOL(inet_confirm_addr);
1da177e4
LT
1079
1080/*
1081 * Device notifier
1082 */
1083
1084int register_inetaddr_notifier(struct notifier_block *nb)
1085{
e041c683 1086 return blocking_notifier_chain_register(&inetaddr_chain, nb);
1da177e4 1087}
9f9354b9 1088EXPORT_SYMBOL(register_inetaddr_notifier);
1da177e4
LT
1089
1090int unregister_inetaddr_notifier(struct notifier_block *nb)
1091{
e041c683 1092 return blocking_notifier_chain_unregister(&inetaddr_chain, nb);
1da177e4 1093}
9f9354b9 1094EXPORT_SYMBOL(unregister_inetaddr_notifier);
1da177e4 1095
9f9354b9
ED
1096/* Rename ifa_labels for a device name change. Make some effort to preserve
1097 * existing alias numbering and to create unique labels if possible.
1da177e4
LT
1098*/
1099static void inetdev_changename(struct net_device *dev, struct in_device *in_dev)
e905a9ed 1100{
1da177e4
LT
1101 struct in_ifaddr *ifa;
1102 int named = 0;
1103
e905a9ed
YH
1104 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
1105 char old[IFNAMSIZ], *dot;
1da177e4
LT
1106
1107 memcpy(old, ifa->ifa_label, IFNAMSIZ);
e905a9ed 1108 memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
1da177e4 1109 if (named++ == 0)
573bf470 1110 goto skip;
44344b2a 1111 dot = strchr(old, ':');
e905a9ed
YH
1112 if (dot == NULL) {
1113 sprintf(old, ":%d", named);
1da177e4
LT
1114 dot = old;
1115 }
9f9354b9 1116 if (strlen(dot) + strlen(dev->name) < IFNAMSIZ)
e905a9ed 1117 strcat(ifa->ifa_label, dot);
9f9354b9 1118 else
e905a9ed 1119 strcpy(ifa->ifa_label + (IFNAMSIZ - strlen(dot) - 1), dot);
573bf470
TG
1120skip:
1121 rtmsg_ifa(RTM_NEWADDR, ifa, NULL, 0);
e905a9ed
YH
1122 }
1123}
1da177e4 1124
40384999 1125static bool inetdev_valid_mtu(unsigned int mtu)
06770843
BL
1126{
1127 return mtu >= 68;
1128}
1129
d11327ad
IC
1130static void inetdev_send_gratuitous_arp(struct net_device *dev,
1131 struct in_device *in_dev)
1132
1133{
b76d0789 1134 struct in_ifaddr *ifa;
d11327ad 1135
b76d0789
ZK
1136 for (ifa = in_dev->ifa_list; ifa;
1137 ifa = ifa->ifa_next) {
1138 arp_send(ARPOP_REQUEST, ETH_P_ARP,
1139 ifa->ifa_local, dev,
1140 ifa->ifa_local, NULL,
1141 dev->dev_addr, NULL);
1142 }
d11327ad
IC
1143}
1144
1da177e4
LT
1145/* Called only under RTNL semaphore */
1146
1147static int inetdev_event(struct notifier_block *this, unsigned long event,
1148 void *ptr)
1149{
1150 struct net_device *dev = ptr;
748e2d93 1151 struct in_device *in_dev = __in_dev_get_rtnl(dev);
0115e8e3 1152
1da177e4
LT
1153 ASSERT_RTNL();
1154
1155 if (!in_dev) {
8030f544 1156 if (event == NETDEV_REGISTER) {
1da177e4 1157 in_dev = inetdev_init(dev);
b217d616
HX
1158 if (!in_dev)
1159 return notifier_from_errno(-ENOMEM);
0cc217e1 1160 if (dev->flags & IFF_LOOPBACK) {
42f811b8
HX
1161 IN_DEV_CONF_SET(in_dev, NOXFRM, 1);
1162 IN_DEV_CONF_SET(in_dev, NOPOLICY, 1);
8030f544 1163 }
06770843
BL
1164 } else if (event == NETDEV_CHANGEMTU) {
1165 /* Re-enabling IP */
1166 if (inetdev_valid_mtu(dev->mtu))
1167 in_dev = inetdev_init(dev);
1da177e4
LT
1168 }
1169 goto out;
1170 }
1171
1172 switch (event) {
1173 case NETDEV_REGISTER:
91df42be 1174 pr_debug("%s: bug\n", __func__);
a9b3cd7f 1175 RCU_INIT_POINTER(dev->ip_ptr, NULL);
1da177e4
LT
1176 break;
1177 case NETDEV_UP:
06770843 1178 if (!inetdev_valid_mtu(dev->mtu))
1da177e4 1179 break;
0cc217e1 1180 if (dev->flags & IFF_LOOPBACK) {
9f9354b9
ED
1181 struct in_ifaddr *ifa = inet_alloc_ifa();
1182
1183 if (ifa) {
fd23c3b3 1184 INIT_HLIST_NODE(&ifa->hash);
1da177e4
LT
1185 ifa->ifa_local =
1186 ifa->ifa_address = htonl(INADDR_LOOPBACK);
1187 ifa->ifa_prefixlen = 8;
1188 ifa->ifa_mask = inet_make_mask(8);
1189 in_dev_hold(in_dev);
1190 ifa->ifa_dev = in_dev;
1191 ifa->ifa_scope = RT_SCOPE_HOST;
1192 memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
1193 inet_insert_ifa(ifa);
1194 }
1195 }
1196 ip_mc_up(in_dev);
eefef1cf
SH
1197 /* fall through */
1198 case NETDEV_CHANGEADDR:
d11327ad
IC
1199 if (!IN_DEV_ARP_NOTIFY(in_dev))
1200 break;
1201 /* fall through */
1202 case NETDEV_NOTIFY_PEERS:
a21090cf 1203 /* Send gratuitous ARP to notify of link change */
d11327ad 1204 inetdev_send_gratuitous_arp(dev, in_dev);
1da177e4
LT
1205 break;
1206 case NETDEV_DOWN:
1207 ip_mc_down(in_dev);
1208 break;
93d9b7d7 1209 case NETDEV_PRE_TYPE_CHANGE:
75c78500
MS
1210 ip_mc_unmap(in_dev);
1211 break;
93d9b7d7 1212 case NETDEV_POST_TYPE_CHANGE:
75c78500
MS
1213 ip_mc_remap(in_dev);
1214 break;
1da177e4 1215 case NETDEV_CHANGEMTU:
06770843 1216 if (inetdev_valid_mtu(dev->mtu))
1da177e4 1217 break;
06770843 1218 /* disable IP when MTU is not enough */
1da177e4
LT
1219 case NETDEV_UNREGISTER:
1220 inetdev_destroy(in_dev);
1221 break;
1222 case NETDEV_CHANGENAME:
1223 /* Do not notify about label change, this event is
1224 * not interesting to applications using netlink.
1225 */
1226 inetdev_changename(dev, in_dev);
1227
51602b2a 1228 devinet_sysctl_unregister(in_dev);
66f27a52 1229 devinet_sysctl_register(in_dev);
1da177e4
LT
1230 break;
1231 }
1232out:
1233 return NOTIFY_DONE;
1234}
1235
1236static struct notifier_block ip_netdev_notifier = {
539afedf 1237 .notifier_call = inetdev_event,
1da177e4
LT
1238};
1239
40384999 1240static size_t inet_nlmsg_size(void)
339bf98f
TG
1241{
1242 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
1243 + nla_total_size(4) /* IFA_ADDRESS */
1244 + nla_total_size(4) /* IFA_LOCAL */
1245 + nla_total_size(4) /* IFA_BROADCAST */
339bf98f
TG
1246 + nla_total_size(IFNAMSIZ); /* IFA_LABEL */
1247}
1248
1da177e4 1249static int inet_fill_ifaddr(struct sk_buff *skb, struct in_ifaddr *ifa,
15e47304 1250 u32 portid, u32 seq, int event, unsigned int flags)
1da177e4
LT
1251{
1252 struct ifaddrmsg *ifm;
1253 struct nlmsghdr *nlh;
1da177e4 1254
15e47304 1255 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*ifm), flags);
47f68512 1256 if (nlh == NULL)
26932566 1257 return -EMSGSIZE;
47f68512
TG
1258
1259 ifm = nlmsg_data(nlh);
1da177e4
LT
1260 ifm->ifa_family = AF_INET;
1261 ifm->ifa_prefixlen = ifa->ifa_prefixlen;
1262 ifm->ifa_flags = ifa->ifa_flags|IFA_F_PERMANENT;
1263 ifm->ifa_scope = ifa->ifa_scope;
1264 ifm->ifa_index = ifa->ifa_dev->dev->ifindex;
47f68512 1265
f3756b79
DM
1266 if ((ifa->ifa_address &&
1267 nla_put_be32(skb, IFA_ADDRESS, ifa->ifa_address)) ||
1268 (ifa->ifa_local &&
1269 nla_put_be32(skb, IFA_LOCAL, ifa->ifa_local)) ||
1270 (ifa->ifa_broadcast &&
1271 nla_put_be32(skb, IFA_BROADCAST, ifa->ifa_broadcast)) ||
1272 (ifa->ifa_label[0] &&
1273 nla_put_string(skb, IFA_LABEL, ifa->ifa_label)))
1274 goto nla_put_failure;
1da177e4 1275
47f68512
TG
1276 return nlmsg_end(skb, nlh);
1277
1278nla_put_failure:
26932566
PM
1279 nlmsg_cancel(skb, nlh);
1280 return -EMSGSIZE;
1da177e4
LT
1281}
1282
1283static int inet_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
1284{
3b1e0a65 1285 struct net *net = sock_net(skb->sk);
eec4df98
ED
1286 int h, s_h;
1287 int idx, s_idx;
1288 int ip_idx, s_ip_idx;
1da177e4
LT
1289 struct net_device *dev;
1290 struct in_device *in_dev;
1291 struct in_ifaddr *ifa;
eec4df98
ED
1292 struct hlist_head *head;
1293 struct hlist_node *node;
1da177e4 1294
eec4df98
ED
1295 s_h = cb->args[0];
1296 s_idx = idx = cb->args[1];
1297 s_ip_idx = ip_idx = cb->args[2];
1298
1299 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1300 idx = 0;
1301 head = &net->dev_index_head[h];
1302 rcu_read_lock();
1303 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
1304 if (idx < s_idx)
1305 goto cont;
4b97efdf 1306 if (h > s_h || idx > s_idx)
eec4df98
ED
1307 s_ip_idx = 0;
1308 in_dev = __in_dev_get_rcu(dev);
1309 if (!in_dev)
1310 goto cont;
1da177e4 1311
eec4df98
ED
1312 for (ifa = in_dev->ifa_list, ip_idx = 0; ifa;
1313 ifa = ifa->ifa_next, ip_idx++) {
1314 if (ip_idx < s_ip_idx)
1315 continue;
1316 if (inet_fill_ifaddr(skb, ifa,
15e47304 1317 NETLINK_CB(cb->skb).portid,
1da177e4 1318 cb->nlh->nlmsg_seq,
eec4df98
ED
1319 RTM_NEWADDR, NLM_F_MULTI) <= 0) {
1320 rcu_read_unlock();
1321 goto done;
1322 }
1323 }
7562f876 1324cont:
eec4df98
ED
1325 idx++;
1326 }
1327 rcu_read_unlock();
1da177e4
LT
1328 }
1329
1330done:
eec4df98
ED
1331 cb->args[0] = h;
1332 cb->args[1] = idx;
1333 cb->args[2] = ip_idx;
1da177e4
LT
1334
1335 return skb->len;
1336}
1337
539afedf 1338static void rtmsg_ifa(int event, struct in_ifaddr *ifa, struct nlmsghdr *nlh,
15e47304 1339 u32 portid)
1da177e4 1340{
47f68512 1341 struct sk_buff *skb;
d6062cbb
TG
1342 u32 seq = nlh ? nlh->nlmsg_seq : 0;
1343 int err = -ENOBUFS;
4b8aa9ab 1344 struct net *net;
1da177e4 1345
c346dca1 1346 net = dev_net(ifa->ifa_dev->dev);
339bf98f 1347 skb = nlmsg_new(inet_nlmsg_size(), GFP_KERNEL);
47f68512 1348 if (skb == NULL)
d6062cbb
TG
1349 goto errout;
1350
15e47304 1351 err = inet_fill_ifaddr(skb, ifa, portid, seq, event, 0);
26932566
PM
1352 if (err < 0) {
1353 /* -EMSGSIZE implies BUG in inet_nlmsg_size() */
1354 WARN_ON(err == -EMSGSIZE);
1355 kfree_skb(skb);
1356 goto errout;
1357 }
15e47304 1358 rtnl_notify(skb, net, portid, RTNLGRP_IPV4_IFADDR, nlh, GFP_KERNEL);
1ce85fe4 1359 return;
d6062cbb
TG
1360errout:
1361 if (err < 0)
4b8aa9ab 1362 rtnl_set_sk_err(net, RTNLGRP_IPV4_IFADDR, err);
1da177e4
LT
1363}
1364
9f0f7272
TG
1365static size_t inet_get_link_af_size(const struct net_device *dev)
1366{
1fc19aff 1367 struct in_device *in_dev = rcu_dereference_rtnl(dev->ip_ptr);
9f0f7272
TG
1368
1369 if (!in_dev)
1370 return 0;
1371
1372 return nla_total_size(IPV4_DEVCONF_MAX * 4); /* IFLA_INET_CONF */
1373}
1374
1375static int inet_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
1376{
1fc19aff 1377 struct in_device *in_dev = rcu_dereference_rtnl(dev->ip_ptr);
9f0f7272
TG
1378 struct nlattr *nla;
1379 int i;
1380
1381 if (!in_dev)
1382 return -ENODATA;
1383
1384 nla = nla_reserve(skb, IFLA_INET_CONF, IPV4_DEVCONF_MAX * 4);
1385 if (nla == NULL)
1386 return -EMSGSIZE;
1387
1388 for (i = 0; i < IPV4_DEVCONF_MAX; i++)
1389 ((u32 *) nla_data(nla))[i] = in_dev->cnf.data[i];
1390
1391 return 0;
1392}
1393
1394static const struct nla_policy inet_af_policy[IFLA_INET_MAX+1] = {
1395 [IFLA_INET_CONF] = { .type = NLA_NESTED },
1396};
1397
cf7afbfe
TG
1398static int inet_validate_link_af(const struct net_device *dev,
1399 const struct nlattr *nla)
9f0f7272 1400{
9f0f7272
TG
1401 struct nlattr *a, *tb[IFLA_INET_MAX+1];
1402 int err, rem;
1403
f7fce74e 1404 if (dev && !__in_dev_get_rtnl(dev))
cf7afbfe 1405 return -EAFNOSUPPORT;
9f0f7272
TG
1406
1407 err = nla_parse_nested(tb, IFLA_INET_MAX, nla, inet_af_policy);
1408 if (err < 0)
1409 return err;
1410
1411 if (tb[IFLA_INET_CONF]) {
1412 nla_for_each_nested(a, tb[IFLA_INET_CONF], rem) {
1413 int cfgid = nla_type(a);
1414
1415 if (nla_len(a) < 4)
1416 return -EINVAL;
1417
1418 if (cfgid <= 0 || cfgid > IPV4_DEVCONF_MAX)
1419 return -EINVAL;
1420 }
1421 }
1422
cf7afbfe
TG
1423 return 0;
1424}
1425
1426static int inet_set_link_af(struct net_device *dev, const struct nlattr *nla)
1427{
f7fce74e 1428 struct in_device *in_dev = __in_dev_get_rtnl(dev);
cf7afbfe
TG
1429 struct nlattr *a, *tb[IFLA_INET_MAX+1];
1430 int rem;
1431
1432 if (!in_dev)
1433 return -EAFNOSUPPORT;
1434
1435 if (nla_parse_nested(tb, IFLA_INET_MAX, nla, NULL) < 0)
1436 BUG();
1437
9f0f7272
TG
1438 if (tb[IFLA_INET_CONF]) {
1439 nla_for_each_nested(a, tb[IFLA_INET_CONF], rem)
1440 ipv4_devconf_set(in_dev, nla_type(a), nla_get_u32(a));
1441 }
1442
1443 return 0;
1444}
1445
edc9e748
ND
1446static int inet_netconf_msgsize_devconf(int type)
1447{
1448 int size = NLMSG_ALIGN(sizeof(struct netconfmsg))
1449 + nla_total_size(4); /* NETCONFA_IFINDEX */
1450
9e551110
ND
1451 /* type -1 is used for ALL */
1452 if (type == -1 || type == NETCONFA_FORWARDING)
edc9e748 1453 size += nla_total_size(4);
cc535dfb
ND
1454 if (type == -1 || type == NETCONFA_RP_FILTER)
1455 size += nla_total_size(4);
d67b8c61
ND
1456 if (type == -1 || type == NETCONFA_MC_FORWARDING)
1457 size += nla_total_size(4);
edc9e748
ND
1458
1459 return size;
1460}
1461
1462static int inet_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
1463 struct ipv4_devconf *devconf, u32 portid,
1464 u32 seq, int event, unsigned int flags,
1465 int type)
1466{
1467 struct nlmsghdr *nlh;
1468 struct netconfmsg *ncm;
1469
1470 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
1471 flags);
1472 if (nlh == NULL)
1473 return -EMSGSIZE;
1474
1475 ncm = nlmsg_data(nlh);
1476 ncm->ncm_family = AF_INET;
1477
1478 if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
1479 goto nla_put_failure;
1480
9e551110
ND
1481 /* type -1 is used for ALL */
1482 if ((type == -1 || type == NETCONFA_FORWARDING) &&
edc9e748
ND
1483 nla_put_s32(skb, NETCONFA_FORWARDING,
1484 IPV4_DEVCONF(*devconf, FORWARDING)) < 0)
1485 goto nla_put_failure;
cc535dfb
ND
1486 if ((type == -1 || type == NETCONFA_RP_FILTER) &&
1487 nla_put_s32(skb, NETCONFA_RP_FILTER,
1488 IPV4_DEVCONF(*devconf, RP_FILTER)) < 0)
1489 goto nla_put_failure;
d67b8c61
ND
1490 if ((type == -1 || type == NETCONFA_MC_FORWARDING) &&
1491 nla_put_s32(skb, NETCONFA_MC_FORWARDING,
1492 IPV4_DEVCONF(*devconf, MC_FORWARDING)) < 0)
1493 goto nla_put_failure;
edc9e748
ND
1494
1495 return nlmsg_end(skb, nlh);
1496
1497nla_put_failure:
1498 nlmsg_cancel(skb, nlh);
1499 return -EMSGSIZE;
1500}
1501
d67b8c61
ND
1502void inet_netconf_notify_devconf(struct net *net, int type, int ifindex,
1503 struct ipv4_devconf *devconf)
edc9e748
ND
1504{
1505 struct sk_buff *skb;
1506 int err = -ENOBUFS;
1507
1508 skb = nlmsg_new(inet_netconf_msgsize_devconf(type), GFP_ATOMIC);
1509 if (skb == NULL)
1510 goto errout;
1511
1512 err = inet_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
1513 RTM_NEWNETCONF, 0, type);
1514 if (err < 0) {
1515 /* -EMSGSIZE implies BUG in inet_netconf_msgsize_devconf() */
1516 WARN_ON(err == -EMSGSIZE);
1517 kfree_skb(skb);
1518 goto errout;
1519 }
1520 rtnl_notify(skb, net, 0, RTNLGRP_IPV4_NETCONF, NULL, GFP_ATOMIC);
1521 return;
1522errout:
1523 if (err < 0)
1524 rtnl_set_sk_err(net, RTNLGRP_IPV4_NETCONF, err);
1525}
1526
9e551110
ND
1527static const struct nla_policy devconf_ipv4_policy[NETCONFA_MAX+1] = {
1528 [NETCONFA_IFINDEX] = { .len = sizeof(int) },
1529 [NETCONFA_FORWARDING] = { .len = sizeof(int) },
cc535dfb 1530 [NETCONFA_RP_FILTER] = { .len = sizeof(int) },
9e551110
ND
1531};
1532
1533static int inet_netconf_get_devconf(struct sk_buff *in_skb,
1534 struct nlmsghdr *nlh,
1535 void *arg)
1536{
1537 struct net *net = sock_net(in_skb->sk);
1538 struct nlattr *tb[NETCONFA_MAX+1];
1539 struct netconfmsg *ncm;
1540 struct sk_buff *skb;
1541 struct ipv4_devconf *devconf;
1542 struct in_device *in_dev;
1543 struct net_device *dev;
1544 int ifindex;
1545 int err;
1546
1547 err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
1548 devconf_ipv4_policy);
1549 if (err < 0)
1550 goto errout;
1551
1552 err = EINVAL;
1553 if (!tb[NETCONFA_IFINDEX])
1554 goto errout;
1555
1556 ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
1557 switch (ifindex) {
1558 case NETCONFA_IFINDEX_ALL:
1559 devconf = net->ipv4.devconf_all;
1560 break;
1561 case NETCONFA_IFINDEX_DEFAULT:
1562 devconf = net->ipv4.devconf_dflt;
1563 break;
1564 default:
1565 dev = __dev_get_by_index(net, ifindex);
1566 if (dev == NULL)
1567 goto errout;
1568 in_dev = __in_dev_get_rtnl(dev);
1569 if (in_dev == NULL)
1570 goto errout;
1571 devconf = &in_dev->cnf;
1572 break;
1573 }
1574
1575 err = -ENOBUFS;
1576 skb = nlmsg_new(inet_netconf_msgsize_devconf(-1), GFP_ATOMIC);
1577 if (skb == NULL)
1578 goto errout;
1579
1580 err = inet_netconf_fill_devconf(skb, ifindex, devconf,
1581 NETLINK_CB(in_skb).portid,
1582 nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
1583 -1);
1584 if (err < 0) {
1585 /* -EMSGSIZE implies BUG in inet_netconf_msgsize_devconf() */
1586 WARN_ON(err == -EMSGSIZE);
1587 kfree_skb(skb);
1588 goto errout;
1589 }
1590 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
1591errout:
1592 return err;
1593}
1594
1da177e4
LT
1595#ifdef CONFIG_SYSCTL
1596
c0ce9fb3 1597static void devinet_copy_dflt_conf(struct net *net, int i)
31be3085
HX
1598{
1599 struct net_device *dev;
1600
c6d14c84
ED
1601 rcu_read_lock();
1602 for_each_netdev_rcu(net, dev) {
31be3085 1603 struct in_device *in_dev;
c6d14c84 1604
31be3085
HX
1605 in_dev = __in_dev_get_rcu(dev);
1606 if (in_dev && !test_bit(i, in_dev->cnf.state))
9355bbd6 1607 in_dev->cnf.data[i] = net->ipv4.devconf_dflt->data[i];
31be3085 1608 }
c6d14c84 1609 rcu_read_unlock();
31be3085
HX
1610}
1611
c6d14c84 1612/* called with RTNL locked */
c0ce9fb3 1613static void inet_forward_change(struct net *net)
68dd299b
PE
1614{
1615 struct net_device *dev;
586f1211 1616 int on = IPV4_DEVCONF_ALL(net, FORWARDING);
68dd299b 1617
586f1211 1618 IPV4_DEVCONF_ALL(net, ACCEPT_REDIRECTS) = !on;
9355bbd6 1619 IPV4_DEVCONF_DFLT(net, FORWARDING) = on;
edc9e748
ND
1620 inet_netconf_notify_devconf(net, NETCONFA_FORWARDING,
1621 NETCONFA_IFINDEX_ALL,
1622 net->ipv4.devconf_all);
1623 inet_netconf_notify_devconf(net, NETCONFA_FORWARDING,
1624 NETCONFA_IFINDEX_DEFAULT,
1625 net->ipv4.devconf_dflt);
68dd299b 1626
c0ce9fb3 1627 for_each_netdev(net, dev) {
68dd299b 1628 struct in_device *in_dev;
0187bdfb
BH
1629 if (on)
1630 dev_disable_lro(dev);
68dd299b
PE
1631 rcu_read_lock();
1632 in_dev = __in_dev_get_rcu(dev);
edc9e748 1633 if (in_dev) {
68dd299b 1634 IN_DEV_CONF_SET(in_dev, FORWARDING, on);
edc9e748
ND
1635 inet_netconf_notify_devconf(net, NETCONFA_FORWARDING,
1636 dev->ifindex, &in_dev->cnf);
1637 }
68dd299b
PE
1638 rcu_read_unlock();
1639 }
68dd299b
PE
1640}
1641
31be3085 1642static int devinet_conf_proc(ctl_table *ctl, int write,
8d65af78 1643 void __user *buffer,
31be3085
HX
1644 size_t *lenp, loff_t *ppos)
1645{
d01ff0a0 1646 int old_value = *(int *)ctl->data;
8d65af78 1647 int ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
d01ff0a0 1648 int new_value = *(int *)ctl->data;
31be3085
HX
1649
1650 if (write) {
1651 struct ipv4_devconf *cnf = ctl->extra1;
c0ce9fb3 1652 struct net *net = ctl->extra2;
31be3085
HX
1653 int i = (int *)ctl->data - cnf->data;
1654
1655 set_bit(i, cnf->state);
1656
9355bbd6 1657 if (cnf == net->ipv4.devconf_dflt)
c0ce9fb3 1658 devinet_copy_dflt_conf(net, i);
d0daebc3
TG
1659 if (i == IPV4_DEVCONF_ACCEPT_LOCAL - 1 ||
1660 i == IPV4_DEVCONF_ROUTE_LOCALNET - 1)
d01ff0a0 1661 if ((new_value == 0) && (old_value != 0))
4ccfe6d4 1662 rt_cache_flush(net);
cc535dfb
ND
1663 if (i == IPV4_DEVCONF_RP_FILTER - 1 &&
1664 new_value != old_value) {
1665 int ifindex;
1666
1667 if (cnf == net->ipv4.devconf_dflt)
1668 ifindex = NETCONFA_IFINDEX_DEFAULT;
1669 else if (cnf == net->ipv4.devconf_all)
1670 ifindex = NETCONFA_IFINDEX_ALL;
1671 else {
1672 struct in_device *idev =
1673 container_of(cnf, struct in_device,
1674 cnf);
1675 ifindex = idev->dev->ifindex;
1676 }
1677 inet_netconf_notify_devconf(net, NETCONFA_RP_FILTER,
1678 ifindex, cnf);
1679 }
31be3085
HX
1680 }
1681
1682 return ret;
1683}
1684
1da177e4 1685static int devinet_sysctl_forward(ctl_table *ctl, int write,
8d65af78 1686 void __user *buffer,
1da177e4
LT
1687 size_t *lenp, loff_t *ppos)
1688{
1689 int *valp = ctl->data;
1690 int val = *valp;
88af182e 1691 loff_t pos = *ppos;
8d65af78 1692 int ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
1da177e4
LT
1693
1694 if (write && *valp != val) {
c0ce9fb3
PE
1695 struct net *net = ctl->extra2;
1696
0187bdfb 1697 if (valp != &IPV4_DEVCONF_DFLT(net, FORWARDING)) {
88af182e
EB
1698 if (!rtnl_trylock()) {
1699 /* Restore the original values before restarting */
1700 *valp = val;
1701 *ppos = pos;
9b8adb5e 1702 return restart_syscall();
88af182e 1703 }
0187bdfb
BH
1704 if (valp == &IPV4_DEVCONF_ALL(net, FORWARDING)) {
1705 inet_forward_change(net);
edc9e748 1706 } else {
0187bdfb
BH
1707 struct ipv4_devconf *cnf = ctl->extra1;
1708 struct in_device *idev =
1709 container_of(cnf, struct in_device, cnf);
edc9e748
ND
1710 if (*valp)
1711 dev_disable_lro(idev->dev);
1712 inet_netconf_notify_devconf(net,
1713 NETCONFA_FORWARDING,
1714 idev->dev->ifindex,
1715 cnf);
0187bdfb
BH
1716 }
1717 rtnl_unlock();
4ccfe6d4 1718 rt_cache_flush(net);
edc9e748
ND
1719 } else
1720 inet_netconf_notify_devconf(net, NETCONFA_FORWARDING,
1721 NETCONFA_IFINDEX_DEFAULT,
1722 net->ipv4.devconf_dflt);
1da177e4
LT
1723 }
1724
1725 return ret;
1726}
1727
323e126f
DM
1728static int ipv4_doint_and_flush(ctl_table *ctl, int write,
1729 void __user *buffer,
1730 size_t *lenp, loff_t *ppos)
1da177e4
LT
1731{
1732 int *valp = ctl->data;
1733 int val = *valp;
8d65af78 1734 int ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
76e6ebfb 1735 struct net *net = ctl->extra2;
1da177e4
LT
1736
1737 if (write && *valp != val)
4ccfe6d4 1738 rt_cache_flush(net);
1da177e4
LT
1739
1740 return ret;
1741}
1742
f8572d8f 1743#define DEVINET_SYSCTL_ENTRY(attr, name, mval, proc) \
42f811b8 1744 { \
42f811b8
HX
1745 .procname = name, \
1746 .data = ipv4_devconf.data + \
02291680 1747 IPV4_DEVCONF_ ## attr - 1, \
42f811b8
HX
1748 .maxlen = sizeof(int), \
1749 .mode = mval, \
1750 .proc_handler = proc, \
31be3085 1751 .extra1 = &ipv4_devconf, \
42f811b8
HX
1752 }
1753
1754#define DEVINET_SYSCTL_RW_ENTRY(attr, name) \
f8572d8f 1755 DEVINET_SYSCTL_ENTRY(attr, name, 0644, devinet_conf_proc)
42f811b8
HX
1756
1757#define DEVINET_SYSCTL_RO_ENTRY(attr, name) \
f8572d8f 1758 DEVINET_SYSCTL_ENTRY(attr, name, 0444, devinet_conf_proc)
42f811b8 1759
f8572d8f
EB
1760#define DEVINET_SYSCTL_COMPLEX_ENTRY(attr, name, proc) \
1761 DEVINET_SYSCTL_ENTRY(attr, name, 0644, proc)
42f811b8
HX
1762
1763#define DEVINET_SYSCTL_FLUSHING_ENTRY(attr, name) \
f8572d8f 1764 DEVINET_SYSCTL_COMPLEX_ENTRY(attr, name, ipv4_doint_and_flush)
42f811b8 1765
1da177e4
LT
1766static struct devinet_sysctl_table {
1767 struct ctl_table_header *sysctl_header;
02291680 1768 struct ctl_table devinet_vars[__IPV4_DEVCONF_MAX];
1da177e4
LT
1769} devinet_sysctl = {
1770 .devinet_vars = {
42f811b8 1771 DEVINET_SYSCTL_COMPLEX_ENTRY(FORWARDING, "forwarding",
f8572d8f 1772 devinet_sysctl_forward),
42f811b8
HX
1773 DEVINET_SYSCTL_RO_ENTRY(MC_FORWARDING, "mc_forwarding"),
1774
1775 DEVINET_SYSCTL_RW_ENTRY(ACCEPT_REDIRECTS, "accept_redirects"),
1776 DEVINET_SYSCTL_RW_ENTRY(SECURE_REDIRECTS, "secure_redirects"),
1777 DEVINET_SYSCTL_RW_ENTRY(SHARED_MEDIA, "shared_media"),
1778 DEVINET_SYSCTL_RW_ENTRY(RP_FILTER, "rp_filter"),
1779 DEVINET_SYSCTL_RW_ENTRY(SEND_REDIRECTS, "send_redirects"),
1780 DEVINET_SYSCTL_RW_ENTRY(ACCEPT_SOURCE_ROUTE,
1781 "accept_source_route"),
8153a10c 1782 DEVINET_SYSCTL_RW_ENTRY(ACCEPT_LOCAL, "accept_local"),
28f6aeea 1783 DEVINET_SYSCTL_RW_ENTRY(SRC_VMARK, "src_valid_mark"),
42f811b8
HX
1784 DEVINET_SYSCTL_RW_ENTRY(PROXY_ARP, "proxy_arp"),
1785 DEVINET_SYSCTL_RW_ENTRY(MEDIUM_ID, "medium_id"),
1786 DEVINET_SYSCTL_RW_ENTRY(BOOTP_RELAY, "bootp_relay"),
1787 DEVINET_SYSCTL_RW_ENTRY(LOG_MARTIANS, "log_martians"),
1788 DEVINET_SYSCTL_RW_ENTRY(TAG, "tag"),
1789 DEVINET_SYSCTL_RW_ENTRY(ARPFILTER, "arp_filter"),
1790 DEVINET_SYSCTL_RW_ENTRY(ARP_ANNOUNCE, "arp_announce"),
1791 DEVINET_SYSCTL_RW_ENTRY(ARP_IGNORE, "arp_ignore"),
1792 DEVINET_SYSCTL_RW_ENTRY(ARP_ACCEPT, "arp_accept"),
eefef1cf 1793 DEVINET_SYSCTL_RW_ENTRY(ARP_NOTIFY, "arp_notify"),
65324144 1794 DEVINET_SYSCTL_RW_ENTRY(PROXY_ARP_PVLAN, "proxy_arp_pvlan"),
42f811b8
HX
1795
1796 DEVINET_SYSCTL_FLUSHING_ENTRY(NOXFRM, "disable_xfrm"),
1797 DEVINET_SYSCTL_FLUSHING_ENTRY(NOPOLICY, "disable_policy"),
1798 DEVINET_SYSCTL_FLUSHING_ENTRY(FORCE_IGMP_VERSION,
1799 "force_igmp_version"),
1800 DEVINET_SYSCTL_FLUSHING_ENTRY(PROMOTE_SECONDARIES,
1801 "promote_secondaries"),
d0daebc3
TG
1802 DEVINET_SYSCTL_FLUSHING_ENTRY(ROUTE_LOCALNET,
1803 "route_localnet"),
1da177e4 1804 },
1da177e4
LT
1805};
1806
ea40b324 1807static int __devinet_sysctl_register(struct net *net, char *dev_name,
f8572d8f 1808 struct ipv4_devconf *p)
1da177e4
LT
1809{
1810 int i;
9fa89642 1811 struct devinet_sysctl_table *t;
8607ddb8 1812 char path[sizeof("net/ipv4/conf/") + IFNAMSIZ];
bfada697 1813
9fa89642 1814 t = kmemdup(&devinet_sysctl, sizeof(*t), GFP_KERNEL);
1da177e4 1815 if (!t)
9fa89642
PE
1816 goto out;
1817
1da177e4
LT
1818 for (i = 0; i < ARRAY_SIZE(t->devinet_vars) - 1; i++) {
1819 t->devinet_vars[i].data += (char *)p - (char *)&ipv4_devconf;
31be3085 1820 t->devinet_vars[i].extra1 = p;
c0ce9fb3 1821 t->devinet_vars[i].extra2 = net;
1da177e4
LT
1822 }
1823
8607ddb8 1824 snprintf(path, sizeof(path), "net/ipv4/conf/%s", dev_name);
1da177e4 1825
8607ddb8 1826 t->sysctl_header = register_net_sysctl(net, path, t->devinet_vars);
1da177e4 1827 if (!t->sysctl_header)
8607ddb8 1828 goto free;
1da177e4
LT
1829
1830 p->sysctl = t;
ea40b324 1831 return 0;
1da177e4 1832
9fa89642 1833free:
1da177e4 1834 kfree(t);
9fa89642 1835out:
ea40b324 1836 return -ENOBUFS;
1da177e4
LT
1837}
1838
51602b2a
PE
1839static void __devinet_sysctl_unregister(struct ipv4_devconf *cnf)
1840{
1841 struct devinet_sysctl_table *t = cnf->sysctl;
1842
1843 if (t == NULL)
1844 return;
1845
1846 cnf->sysctl = NULL;
ff538818 1847 unregister_net_sysctl_table(t->sysctl_header);
51602b2a
PE
1848 kfree(t);
1849}
1850
66f27a52
PE
1851static void devinet_sysctl_register(struct in_device *idev)
1852{
54716e3b 1853 neigh_sysctl_register(idev->dev, idev->arp_parms, "ipv4", NULL);
c346dca1 1854 __devinet_sysctl_register(dev_net(idev->dev), idev->dev->name,
f8572d8f 1855 &idev->cnf);
66f27a52
PE
1856}
1857
51602b2a 1858static void devinet_sysctl_unregister(struct in_device *idev)
1da177e4 1859{
51602b2a
PE
1860 __devinet_sysctl_unregister(&idev->cnf);
1861 neigh_sysctl_unregister(idev->arp_parms);
1da177e4 1862}
1da177e4 1863
68dd299b
PE
1864static struct ctl_table ctl_forward_entry[] = {
1865 {
68dd299b
PE
1866 .procname = "ip_forward",
1867 .data = &ipv4_devconf.data[
02291680 1868 IPV4_DEVCONF_FORWARDING - 1],
68dd299b
PE
1869 .maxlen = sizeof(int),
1870 .mode = 0644,
1871 .proc_handler = devinet_sysctl_forward,
68dd299b 1872 .extra1 = &ipv4_devconf,
c0ce9fb3 1873 .extra2 = &init_net,
68dd299b
PE
1874 },
1875 { },
1876};
2a75de0c 1877#endif
68dd299b 1878
752d14dc
PE
1879static __net_init int devinet_init_net(struct net *net)
1880{
1881 int err;
752d14dc 1882 struct ipv4_devconf *all, *dflt;
2a75de0c
ED
1883#ifdef CONFIG_SYSCTL
1884 struct ctl_table *tbl = ctl_forward_entry;
752d14dc 1885 struct ctl_table_header *forw_hdr;
2a75de0c 1886#endif
752d14dc
PE
1887
1888 err = -ENOMEM;
1889 all = &ipv4_devconf;
1890 dflt = &ipv4_devconf_dflt;
752d14dc 1891
09ad9bc7 1892 if (!net_eq(net, &init_net)) {
752d14dc
PE
1893 all = kmemdup(all, sizeof(ipv4_devconf), GFP_KERNEL);
1894 if (all == NULL)
1895 goto err_alloc_all;
1896
1897 dflt = kmemdup(dflt, sizeof(ipv4_devconf_dflt), GFP_KERNEL);
1898 if (dflt == NULL)
1899 goto err_alloc_dflt;
1900
2a75de0c 1901#ifdef CONFIG_SYSCTL
752d14dc
PE
1902 tbl = kmemdup(tbl, sizeof(ctl_forward_entry), GFP_KERNEL);
1903 if (tbl == NULL)
1904 goto err_alloc_ctl;
1905
02291680 1906 tbl[0].data = &all->data[IPV4_DEVCONF_FORWARDING - 1];
752d14dc
PE
1907 tbl[0].extra1 = all;
1908 tbl[0].extra2 = net;
2a75de0c 1909#endif
752d14dc
PE
1910 }
1911
1912#ifdef CONFIG_SYSCTL
f8572d8f 1913 err = __devinet_sysctl_register(net, "all", all);
752d14dc
PE
1914 if (err < 0)
1915 goto err_reg_all;
1916
f8572d8f 1917 err = __devinet_sysctl_register(net, "default", dflt);
752d14dc
PE
1918 if (err < 0)
1919 goto err_reg_dflt;
1920
1921 err = -ENOMEM;
8607ddb8 1922 forw_hdr = register_net_sysctl(net, "net/ipv4", tbl);
752d14dc
PE
1923 if (forw_hdr == NULL)
1924 goto err_reg_ctl;
2a75de0c 1925 net->ipv4.forw_hdr = forw_hdr;
752d14dc
PE
1926#endif
1927
752d14dc
PE
1928 net->ipv4.devconf_all = all;
1929 net->ipv4.devconf_dflt = dflt;
1930 return 0;
1931
1932#ifdef CONFIG_SYSCTL
1933err_reg_ctl:
1934 __devinet_sysctl_unregister(dflt);
1935err_reg_dflt:
1936 __devinet_sysctl_unregister(all);
1937err_reg_all:
1938 if (tbl != ctl_forward_entry)
1939 kfree(tbl);
752d14dc 1940err_alloc_ctl:
2a75de0c 1941#endif
752d14dc
PE
1942 if (dflt != &ipv4_devconf_dflt)
1943 kfree(dflt);
1944err_alloc_dflt:
1945 if (all != &ipv4_devconf)
1946 kfree(all);
1947err_alloc_all:
1948 return err;
1949}
1950
1951static __net_exit void devinet_exit_net(struct net *net)
1952{
2a75de0c 1953#ifdef CONFIG_SYSCTL
752d14dc
PE
1954 struct ctl_table *tbl;
1955
1956 tbl = net->ipv4.forw_hdr->ctl_table_arg;
752d14dc
PE
1957 unregister_net_sysctl_table(net->ipv4.forw_hdr);
1958 __devinet_sysctl_unregister(net->ipv4.devconf_dflt);
1959 __devinet_sysctl_unregister(net->ipv4.devconf_all);
752d14dc 1960 kfree(tbl);
2a75de0c 1961#endif
752d14dc
PE
1962 kfree(net->ipv4.devconf_dflt);
1963 kfree(net->ipv4.devconf_all);
1964}
1965
1966static __net_initdata struct pernet_operations devinet_ops = {
1967 .init = devinet_init_net,
1968 .exit = devinet_exit_net,
1969};
1970
9f0f7272
TG
1971static struct rtnl_af_ops inet_af_ops = {
1972 .family = AF_INET,
1973 .fill_link_af = inet_fill_link_af,
1974 .get_link_af_size = inet_get_link_af_size,
cf7afbfe
TG
1975 .validate_link_af = inet_validate_link_af,
1976 .set_link_af = inet_set_link_af,
9f0f7272
TG
1977};
1978
1da177e4
LT
1979void __init devinet_init(void)
1980{
fd23c3b3
DM
1981 int i;
1982
1983 for (i = 0; i < IN4_ADDR_HSIZE; i++)
1984 INIT_HLIST_HEAD(&inet_addr_lst[i]);
1985
752d14dc
PE
1986 register_pernet_subsys(&devinet_ops);
1987
1da177e4
LT
1988 register_gifconf(PF_INET, inet_gifconf);
1989 register_netdevice_notifier(&ip_netdev_notifier);
63f3444f 1990
9f0f7272
TG
1991 rtnl_af_register(&inet_af_ops);
1992
c7ac8679
GR
1993 rtnl_register(PF_INET, RTM_NEWADDR, inet_rtm_newaddr, NULL, NULL);
1994 rtnl_register(PF_INET, RTM_DELADDR, inet_rtm_deladdr, NULL, NULL);
1995 rtnl_register(PF_INET, RTM_GETADDR, NULL, inet_dump_ifaddr, NULL);
9e551110
ND
1996 rtnl_register(PF_INET, RTM_GETNETCONF, inet_netconf_get_devconf,
1997 NULL, NULL);
1da177e4
LT
1998}
1999