projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
net: l3mdev: address selection should only consider devices in L3 domain
[linux-2.6-block.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,
2690048c
WM		 76 [IPV4_DEVCONF_IGMPV2_UNSOLICITED_REPORT_INTERVAL - 1] = 10000 /*ms*/,
77 [IPV4_DEVCONF_IGMPV3_UNSOLICITED_REPORT_INTERVAL - 1] = 1000 /*ms*/,
42f811b8 78 },
1da177e4
LT		 79};
80
81static struct ipv4_devconf ipv4_devconf_dflt = {
42f811b8 82 .data = {
02291680
EB		 83 [IPV4_DEVCONF_ACCEPT_REDIRECTS - 1] = 1,
84 [IPV4_DEVCONF_SEND_REDIRECTS - 1] = 1,
85 [IPV4_DEVCONF_SECURE_REDIRECTS - 1] = 1,
86 [IPV4_DEVCONF_SHARED_MEDIA - 1] = 1,
87 [IPV4_DEVCONF_ACCEPT_SOURCE_ROUTE - 1] = 1,
2690048c
WM		 88 [IPV4_DEVCONF_IGMPV2_UNSOLICITED_REPORT_INTERVAL - 1] = 10000 /*ms*/,
89 [IPV4_DEVCONF_IGMPV3_UNSOLICITED_REPORT_INTERVAL - 1] = 1000 /*ms*/,
42f811b8 90 },
1da177e4
LT		 91};
92
9355bbd6
PE		 93#define IPV4_DEVCONF_DFLT(net, attr) \
94 IPV4_DEVCONF((*net->ipv4.devconf_dflt), attr)
42f811b8 95
ef7c79ed 96static const struct nla_policy ifa_ipv4_policy[IFA_MAX+1] = {
5c753978
TG		 97 [IFA_LOCAL] = { .type = NLA_U32 },
98 [IFA_ADDRESS] = { .type = NLA_U32 },
99 [IFA_BROADCAST] = { .type = NLA_U32 },
5176f91e 100 [IFA_LABEL] = { .type = NLA_STRING, .len = IFNAMSIZ - 1 },
5c766d64 101 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
ad6c8135 102 [IFA_FLAGS] = { .type = NLA_U32 },
5c753978
TG		 103};
104
40384999
ED		 105#define IN4_ADDR_HSIZE_SHIFT 8
106#define IN4_ADDR_HSIZE (1U << IN4_ADDR_HSIZE_SHIFT)
107
fd23c3b3 108static struct hlist_head inet_addr_lst[IN4_ADDR_HSIZE];
fd23c3b3 109
6eada011 110static u32 inet_addr_hash(const struct net *net, __be32 addr)
fd23c3b3 111{
40384999 112 u32 val = (__force u32) addr ^ net_hash_mix(net);
fd23c3b3 113
40384999 114 return hash_32(val, IN4_ADDR_HSIZE_SHIFT);
fd23c3b3
DM		 115}
116
117static void inet_hash_insert(struct net *net, struct in_ifaddr *ifa)
118{
40384999 119 u32 hash = inet_addr_hash(net, ifa->ifa_local);
fd23c3b3 120
32a4be48 121 ASSERT_RTNL();
fd23c3b3 122 hlist_add_head_rcu(&ifa->hash, &inet_addr_lst[hash]);
fd23c3b3
DM		 123}
124
125static void inet_hash_remove(struct in_ifaddr *ifa)
126{
32a4be48 127 ASSERT_RTNL();
fd23c3b3 128 hlist_del_init_rcu(&ifa->hash);
fd23c3b3
DM		 129}
130
9435eb1c
DM		 131/**
132 * __ip_dev_find - find the first device with a given source address.
133 * @net: the net namespace
134 * @addr: the source address
135 * @devref: if true, take a reference on the found device
136 *
137 * If a caller uses devref=false, it should be protected by RCU, or RTNL
138 */
139struct net_device *__ip_dev_find(struct net *net, __be32 addr, bool devref)
140{
40384999 141 u32 hash = inet_addr_hash(net, addr);
9435eb1c
DM		 142 struct net_device *result = NULL;
143 struct in_ifaddr *ifa;
9435eb1c
DM		 144
145 rcu_read_lock();
b67bfe0d 146 hlist_for_each_entry_rcu(ifa, &inet_addr_lst[hash], hash) {
e066008b 147 if (ifa->ifa_local == addr) {
40384999
ED		 148 struct net_device *dev = ifa->ifa_dev->dev;
149
150 if (!net_eq(dev_net(dev), net))
151 continue;
9435eb1c
DM		 152 result = dev;
153 break;
154 }
155 }
406b6f97
DM		 156 if (!result) {
157 struct flowi4 fl4 = { .daddr = addr };
158 struct fib_result res = { 0 };
159 struct fib_table *local;
160
161 /* Fallback to FIB local table so that communication
162 * over loopback subnets work.
163 */
164 local = fib_get_table(net, RT_TABLE_LOCAL);
165 if (local &&
166 !fib_table_lookup(local, &fl4, &res, FIB_LOOKUP_NOREF) &&
167 res.type == RTN_LOCAL)
168 result = FIB_RES_DEV(res);
169 }
9435eb1c
DM		 170 if (result && devref)
171 dev_hold(result);
172 rcu_read_unlock();
173 return result;
174}
175EXPORT_SYMBOL(__ip_dev_find);
176
d6062cbb 177static void rtmsg_ifa(int event, struct in_ifaddr *, struct nlmsghdr *, u32);
1da177e4 178
e041c683 179static BLOCKING_NOTIFIER_HEAD(inetaddr_chain);
1da177e4
LT		 180static void inet_del_ifa(struct in_device *in_dev, struct in_ifaddr **ifap,
181 int destroy);
182#ifdef CONFIG_SYSCTL
20e61da7 183static int devinet_sysctl_register(struct in_device *idev);
51602b2a
PE		 184static void devinet_sysctl_unregister(struct in_device *idev);
185#else
20e61da7 186static int devinet_sysctl_register(struct in_device *idev)
51602b2a 187{
20e61da7 188 return 0;
51602b2a 189}
40384999 190static void devinet_sysctl_unregister(struct in_device *idev)
51602b2a
PE		 191{
192}
1da177e4
LT		 193#endif
194
195/* Locks all the inet devices. */
196
197static struct in_ifaddr *inet_alloc_ifa(void)
198{
93adcc80 199 return kzalloc(sizeof(struct in_ifaddr), GFP_KERNEL);
1da177e4
LT		 200}
201
202static void inet_rcu_free_ifa(struct rcu_head *head)
203{
204 struct in_ifaddr *ifa = container_of(head, struct in_ifaddr, rcu_head);
205 if (ifa->ifa_dev)
206 in_dev_put(ifa->ifa_dev);
207 kfree(ifa);
208}
209
40384999 210static void inet_free_ifa(struct in_ifaddr *ifa)
1da177e4
LT		 211{
212 call_rcu(&ifa->rcu_head, inet_rcu_free_ifa);
213}
214
215void in_dev_finish_destroy(struct in_device *idev)
216{
217 struct net_device *dev = idev->dev;
218
547b792c
IJ		 219 WARN_ON(idev->ifa_list);
220 WARN_ON(idev->mc_list);
e9897071 221 kfree(rcu_dereference_protected(idev->mc_hash, 1));
1da177e4 222#ifdef NET_REFCNT_DEBUG
91df42be 223 pr_debug("%s: %p=%s\n", __func__, idev, dev ? dev->name : "NIL");
1da177e4
LT		 224#endif
225 dev_put(dev);
226 if (!idev->dead)
9f9354b9
ED		 227 pr_err("Freeing alive in_device %p\n", idev);
228 else
1da177e4 229 kfree(idev);
1da177e4 230}
9f9354b9 231EXPORT_SYMBOL(in_dev_finish_destroy);
1da177e4 232
71e27da9 233static struct in_device *inetdev_init(struct net_device *dev)
1da177e4
LT		 234{
235 struct in_device *in_dev;
20e61da7 236 int err = -ENOMEM;
1da177e4
LT		 237
238 ASSERT_RTNL();
239
0da974f4 240 in_dev = kzalloc(sizeof(*in_dev), GFP_KERNEL);
1da177e4
LT		 241 if (!in_dev)
242 goto out;
c346dca1 243 memcpy(&in_dev->cnf, dev_net(dev)->ipv4.devconf_dflt,
9355bbd6 244 sizeof(in_dev->cnf));
1da177e4
LT		 245 in_dev->cnf.sysctl = NULL;
246 in_dev->dev = dev;
9f9354b9
ED		 247 in_dev->arp_parms = neigh_parms_alloc(dev, &arp_tbl);
248 if (!in_dev->arp_parms)
1da177e4 249 goto out_kfree;
0187bdfb
BH		 250 if (IPV4_DEVCONF(in_dev->cnf, FORWARDING))
251 dev_disable_lro(dev);
1da177e4
LT		 252 /* Reference in_dev->dev */
253 dev_hold(dev);
30c4cf57 254 /* Account for reference dev->ip_ptr (below) */
1da177e4 255 in_dev_hold(in_dev);
1da177e4 256
20e61da7
WC		 257 err = devinet_sysctl_register(in_dev);
258 if (err) {
259 in_dev->dead = 1;
260 in_dev_put(in_dev);
261 in_dev = NULL;
262 goto out;
263 }
1da177e4
LT		 264 ip_mc_init_dev(in_dev);
265 if (dev->flags & IFF_UP)
266 ip_mc_up(in_dev);
483479ec 267
30c4cf57 268 /* we can receive as soon as ip_ptr is set -- do this last */
cf778b00 269 rcu_assign_pointer(dev->ip_ptr, in_dev);
483479ec 270out:
20e61da7 271 return in_dev ?: ERR_PTR(err);
1da177e4
LT		 272out_kfree:
273 kfree(in_dev);
274 in_dev = NULL;
275 goto out;
276}
277
278static void in_dev_rcu_put(struct rcu_head *head)
279{
280 struct in_device *idev = container_of(head, struct in_device, rcu_head);
281 in_dev_put(idev);
282}
283
284static void inetdev_destroy(struct in_device *in_dev)
285{
286 struct in_ifaddr *ifa;
287 struct net_device *dev;
288
289 ASSERT_RTNL();
290
291 dev = in_dev->dev;
1da177e4
LT		 292
293 in_dev->dead = 1;
294
295 ip_mc_destroy_dev(in_dev);
296
297 while ((ifa = in_dev->ifa_list) != NULL) {
298 inet_del_ifa(in_dev, &in_dev->ifa_list, 0);
299 inet_free_ifa(ifa);
300 }
301
a9b3cd7f 302 RCU_INIT_POINTER(dev->ip_ptr, NULL);
1da177e4 303
51602b2a 304 devinet_sysctl_unregister(in_dev);
1da177e4
LT		 305 neigh_parms_release(&arp_tbl, in_dev->arp_parms);
306 arp_ifdown(dev);
307
308 call_rcu(&in_dev->rcu_head, in_dev_rcu_put);
309}
310
ff428d72 311int inet_addr_onlink(struct in_device *in_dev, __be32 a, __be32 b)
1da177e4
LT		 312{
313 rcu_read_lock();
314 for_primary_ifa(in_dev) {
315 if (inet_ifa_match(a, ifa)) {
316 if (!b || inet_ifa_match(b, ifa)) {
317 rcu_read_unlock();
318 return 1;
319 }
320 }
321 } endfor_ifa(in_dev);
322 rcu_read_unlock();
323 return 0;
324}
325
d6062cbb 326static void __inet_del_ifa(struct in_device *in_dev, struct in_ifaddr **ifap,
15e47304 327 int destroy, struct nlmsghdr *nlh, u32 portid)
1da177e4 328{
8f937c60 329 struct in_ifaddr *promote = NULL;
0ff60a45
JHS		 330 struct in_ifaddr *ifa, *ifa1 = *ifap;
331 struct in_ifaddr *last_prim = in_dev->ifa_list;
332 struct in_ifaddr *prev_prom = NULL;
333 int do_promote = IN_DEV_PROMOTE_SECONDARIES(in_dev);
1da177e4
LT		 334
335 ASSERT_RTNL();
336
e905a9ed 337 /* 1. Deleting primary ifaddr forces deletion all secondaries
8f937c60
HW		 338 * unless alias promotion is set
339 **/
1da177e4
LT		 340
341 if (!(ifa1->ifa_flags & IFA_F_SECONDARY)) {
1da177e4
LT		 342 struct in_ifaddr **ifap1 = &ifa1->ifa_next;
343
344 while ((ifa = *ifap1) != NULL) {
e905a9ed 345 if (!(ifa->ifa_flags & IFA_F_SECONDARY) &&
0ff60a45
JHS		 346 ifa1->ifa_scope <= ifa->ifa_scope)
347 last_prim = ifa;
348
1da177e4
LT		 349 if (!(ifa->ifa_flags & IFA_F_SECONDARY) ||
350 ifa1->ifa_mask != ifa->ifa_mask ||
351 !inet_ifa_match(ifa1->ifa_address, ifa)) {
352 ifap1 = &ifa->ifa_next;
0ff60a45 353 prev_prom = ifa;
1da177e4
LT		 354 continue;
355 }
356
0ff60a45 357 if (!do_promote) {
fd23c3b3 358 inet_hash_remove(ifa);
8f937c60 359 *ifap1 = ifa->ifa_next;
1da177e4 360
15e47304 361 rtmsg_ifa(RTM_DELADDR, ifa, nlh, portid);
e041c683
AS		 362 blocking_notifier_call_chain(&inetaddr_chain,
363 NETDEV_DOWN, ifa);
8f937c60
HW		 364 inet_free_ifa(ifa);
365 } else {
366 promote = ifa;
367 break;
368 }
1da177e4
LT		 369 }
370 }
371
2d230e2b
JA		 372 /* On promotion all secondaries from subnet are changing
373 * the primary IP, we must remove all their routes silently
374 * and later to add them back with new prefsrc. Do this
375 * while all addresses are on the device list.
376 */
377 for (ifa = promote; ifa; ifa = ifa->ifa_next) {
378 if (ifa1->ifa_mask == ifa->ifa_mask &&
379 inet_ifa_match(ifa1->ifa_address, ifa))
380 fib_del_ifaddr(ifa, ifa1);
381 }
382
1da177e4
LT		 383 /* 2. Unlink it */
384
385 *ifap = ifa1->ifa_next;
fd23c3b3 386 inet_hash_remove(ifa1);
1da177e4
LT		 387
388 /* 3. Announce address deletion */
389
390 /* Send message first, then call notifier.
391 At first sight, FIB update triggered by notifier
392 will refer to already deleted ifaddr, that could confuse
393 netlink listeners. It is not true: look, gated sees
394 that route deleted and if it still thinks that ifaddr
395 is valid, it will try to restore deleted routes... Grr.
396 So that, this order is correct.
397 */
15e47304 398 rtmsg_ifa(RTM_DELADDR, ifa1, nlh, portid);
e041c683 399 blocking_notifier_call_chain(&inetaddr_chain, NETDEV_DOWN, ifa1);
1da177e4 400
0ff60a45 401 if (promote) {
04024b93 402 struct in_ifaddr *next_sec = promote->ifa_next;
0ff60a45
JHS		 403
404 if (prev_prom) {
405 prev_prom->ifa_next = promote->ifa_next;
406 promote->ifa_next = last_prim->ifa_next;
407 last_prim->ifa_next = promote;
408 }
8f937c60 409
8f937c60 410 promote->ifa_flags &= ~IFA_F_SECONDARY;
15e47304 411 rtmsg_ifa(RTM_NEWADDR, promote, nlh, portid);
e041c683
AS		 412 blocking_notifier_call_chain(&inetaddr_chain,
413 NETDEV_UP, promote);
04024b93 414 for (ifa = next_sec; ifa; ifa = ifa->ifa_next) {
0ff60a45
JHS		 415 if (ifa1->ifa_mask != ifa->ifa_mask ||
416 !inet_ifa_match(ifa1->ifa_address, ifa))
417 continue;
418 fib_add_ifaddr(ifa);
419 }
420
421 }
6363097c 422 if (destroy)
0ff60a45 423 inet_free_ifa(ifa1);
1da177e4
LT		 424}
425
d6062cbb
TG		 426static void inet_del_ifa(struct in_device *in_dev, struct in_ifaddr **ifap,
427 int destroy)
428{
429 __inet_del_ifa(in_dev, ifap, destroy, NULL, 0);
430}
431
5c766d64
JP		 432static void check_lifetime(struct work_struct *work);
433
434static DECLARE_DELAYED_WORK(check_lifetime_work, check_lifetime);
435
d6062cbb 436static int __inet_insert_ifa(struct in_ifaddr *ifa, struct nlmsghdr *nlh,
15e47304 437 u32 portid)
1da177e4
LT		 438{
439 struct in_device *in_dev = ifa->ifa_dev;
440 struct in_ifaddr *ifa1, **ifap, **last_primary;
441
442 ASSERT_RTNL();
443
444 if (!ifa->ifa_local) {
445 inet_free_ifa(ifa);
446 return 0;
447 }
448
449 ifa->ifa_flags &= ~IFA_F_SECONDARY;
450 last_primary = &in_dev->ifa_list;
451
452 for (ifap = &in_dev->ifa_list; (ifa1 = *ifap) != NULL;
453 ifap = &ifa1->ifa_next) {
454 if (!(ifa1->ifa_flags & IFA_F_SECONDARY) &&
455 ifa->ifa_scope <= ifa1->ifa_scope)
456 last_primary = &ifa1->ifa_next;
457 if (ifa1->ifa_mask == ifa->ifa_mask &&
458 inet_ifa_match(ifa1->ifa_address, ifa)) {
459 if (ifa1->ifa_local == ifa->ifa_local) {
460 inet_free_ifa(ifa);
461 return -EEXIST;
462 }
463 if (ifa1->ifa_scope != ifa->ifa_scope) {
464 inet_free_ifa(ifa);
465 return -EINVAL;
466 }
467 ifa->ifa_flags |= IFA_F_SECONDARY;
468 }
469 }
470
471 if (!(ifa->ifa_flags & IFA_F_SECONDARY)) {
63862b5b 472 prandom_seed((__force u32) ifa->ifa_local);
1da177e4
LT		 473 ifap = last_primary;
474 }
475
476 ifa->ifa_next = *ifap;
477 *ifap = ifa;
478
fd23c3b3
DM		 479 inet_hash_insert(dev_net(in_dev->dev), ifa);
480
5c766d64 481 cancel_delayed_work(&check_lifetime_work);
906e073f 482 queue_delayed_work(system_power_efficient_wq, &check_lifetime_work, 0);
5c766d64 483
1da177e4
LT		 484 /* Send message first, then call notifier.
485 Notifier will trigger FIB update, so that
486 listeners of netlink will know about new ifaddr */
15e47304 487 rtmsg_ifa(RTM_NEWADDR, ifa, nlh, portid);
e041c683 488 blocking_notifier_call_chain(&inetaddr_chain, NETDEV_UP, ifa);
1da177e4
LT		 489
490 return 0;
491}
492
d6062cbb
TG		 493static int inet_insert_ifa(struct in_ifaddr *ifa)
494{
495 return __inet_insert_ifa(ifa, NULL, 0);
496}
497
1da177e4
LT		 498static int inet_set_ifa(struct net_device *dev, struct in_ifaddr *ifa)
499{
e5ed6399 500 struct in_device *in_dev = __in_dev_get_rtnl(dev);
1da177e4
LT		 501
502 ASSERT_RTNL();
503
504 if (!in_dev) {
71e27da9
HX		 505 inet_free_ifa(ifa);
506 return -ENOBUFS;
1da177e4 507 }
71e27da9 508 ipv4_devconf_setall(in_dev);
1d4c8c29 509 neigh_parms_data_state_setall(in_dev->arp_parms);
1da177e4 510 if (ifa->ifa_dev != in_dev) {
547b792c 511 WARN_ON(ifa->ifa_dev);
1da177e4
LT		 512 in_dev_hold(in_dev);
513 ifa->ifa_dev = in_dev;
514 }
f97c1e0c 515 if (ipv4_is_loopback(ifa->ifa_local))
1da177e4
LT		 516 ifa->ifa_scope = RT_SCOPE_HOST;
517 return inet_insert_ifa(ifa);
518}
519
8723e1b4
ED		 520/* Caller must hold RCU or RTNL :
521 * We dont take a reference on found in_device
522 */
7fee0ca2 523struct in_device *inetdev_by_index(struct net *net, int ifindex)
1da177e4
LT		 524{
525 struct net_device *dev;
526 struct in_device *in_dev = NULL;
c148fc2e
ED		 527
528 rcu_read_lock();
529 dev = dev_get_by_index_rcu(net, ifindex);
1da177e4 530 if (dev)
8723e1b4 531 in_dev = rcu_dereference_rtnl(dev->ip_ptr);
c148fc2e 532 rcu_read_unlock();
1da177e4
LT		 533 return in_dev;
534}
9f9354b9 535EXPORT_SYMBOL(inetdev_by_index);
1da177e4
LT		 536
537/* Called only from RTNL semaphored context. No locks. */
538
60cad5da
AV		 539struct in_ifaddr *inet_ifa_byprefix(struct in_device *in_dev, __be32 prefix,
540 __be32 mask)
1da177e4
LT		 541{
542 ASSERT_RTNL();
543
544 for_primary_ifa(in_dev) {
545 if (ifa->ifa_mask == mask && inet_ifa_match(prefix, ifa))
546 return ifa;
547 } endfor_ifa(in_dev);
548 return NULL;
549}
550
93a714d6
MC		 551static int ip_mc_config(struct sock *sk, bool join, const struct in_ifaddr *ifa)
552{
553 struct ip_mreqn mreq = {
554 .imr_multiaddr.s_addr = ifa->ifa_address,
555 .imr_ifindex = ifa->ifa_dev->dev->ifindex,
556 };
557 int ret;
558
559 ASSERT_RTNL();
560
561 lock_sock(sk);
562 if (join)
54ff9ef3 563 ret = ip_mc_join_group(sk, &mreq);
93a714d6 564 else
54ff9ef3 565 ret = ip_mc_leave_group(sk, &mreq);
93a714d6
MC		 566 release_sock(sk);
567
568 return ret;
569}
570
661d2967 571static int inet_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh)
1da177e4 572{
3b1e0a65 573 struct net *net = sock_net(skb->sk);
dfdd5fd4 574 struct nlattr *tb[IFA_MAX+1];
1da177e4 575 struct in_device *in_dev;
dfdd5fd4 576 struct ifaddrmsg *ifm;
1da177e4 577 struct in_ifaddr *ifa, **ifap;
dfdd5fd4 578 int err = -EINVAL;
1da177e4
LT		 579
580 ASSERT_RTNL();
581
dfdd5fd4
TG		 582 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv4_policy);
583 if (err < 0)
584 goto errout;
585
586 ifm = nlmsg_data(nlh);
7fee0ca2 587 in_dev = inetdev_by_index(net, ifm->ifa_index);
51456b29 588 if (!in_dev) {
dfdd5fd4
TG		 589 err = -ENODEV;
590 goto errout;
591 }
592
1da177e4
LT		 593 for (ifap = &in_dev->ifa_list; (ifa = *ifap) != NULL;
594 ifap = &ifa->ifa_next) {
dfdd5fd4 595 if (tb[IFA_LOCAL] &&
67b61f6c 596 ifa->ifa_local != nla_get_in_addr(tb[IFA_LOCAL]))
dfdd5fd4
TG		 597 continue;
598
599 if (tb[IFA_LABEL] && nla_strcmp(tb[IFA_LABEL], ifa->ifa_label))
1da177e4 600 continue;
dfdd5fd4
TG		 601
602 if (tb[IFA_ADDRESS] &&
603 (ifm->ifa_prefixlen != ifa->ifa_prefixlen ||
67b61f6c 604 !inet_ifa_match(nla_get_in_addr(tb[IFA_ADDRESS]), ifa)))
dfdd5fd4
TG		 605 continue;
606
93a714d6
MC		 607 if (ipv4_is_multicast(ifa->ifa_address))
608 ip_mc_config(net->ipv4.mc_autojoin_sk, false, ifa);
15e47304 609 __inet_del_ifa(in_dev, ifap, 1, nlh, NETLINK_CB(skb).portid);
1da177e4
LT		 610 return 0;
611 }
dfdd5fd4
TG		 612
613 err = -EADDRNOTAVAIL;
614errout:
615 return err;
1da177e4
LT		 616}
617
5c766d64
JP		 618#define INFINITY_LIFE_TIME 0xFFFFFFFF
619
620static void check_lifetime(struct work_struct *work)
621{
622 unsigned long now, next, next_sec, next_sched;
623 struct in_ifaddr *ifa;
c988d1e8 624 struct hlist_node *n;
5c766d64
JP		 625 int i;
626
627 now = jiffies;
628 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
629
5c766d64 630 for (i = 0; i < IN4_ADDR_HSIZE; i++) {
c988d1e8
JP		 631 bool change_needed = false;
632
633 rcu_read_lock();
b67bfe0d 634 hlist_for_each_entry_rcu(ifa, &inet_addr_lst[i], hash) {
5c766d64
JP		 635 unsigned long age;
636
637 if (ifa->ifa_flags & IFA_F_PERMANENT)
638 continue;
639
640 /* We try to batch several events at once. */
641 age = (now - ifa->ifa_tstamp +
642 ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
643
644 if (ifa->ifa_valid_lft != INFINITY_LIFE_TIME &&
645 age >= ifa->ifa_valid_lft) {
c988d1e8 646 change_needed = true;
5c766d64
JP		 647 } else if (ifa->ifa_preferred_lft ==
648 INFINITY_LIFE_TIME) {
649 continue;
650 } else if (age >= ifa->ifa_preferred_lft) {
651 if (time_before(ifa->ifa_tstamp +
652 ifa->ifa_valid_lft * HZ, next))
653 next = ifa->ifa_tstamp +
654 ifa->ifa_valid_lft * HZ;
655
c988d1e8
JP		 656 if (!(ifa->ifa_flags & IFA_F_DEPRECATED))
657 change_needed = true;
5c766d64
JP		 658 } else if (time_before(ifa->ifa_tstamp +
659 ifa->ifa_preferred_lft * HZ,
660 next)) {
661 next = ifa->ifa_tstamp +
662 ifa->ifa_preferred_lft * HZ;
663 }
664 }
c988d1e8
JP		 665 rcu_read_unlock();
666 if (!change_needed)
667 continue;
668 rtnl_lock();
669 hlist_for_each_entry_safe(ifa, n, &inet_addr_lst[i], hash) {
670 unsigned long age;
671
672 if (ifa->ifa_flags & IFA_F_PERMANENT)
673 continue;
674
675 /* We try to batch several events at once. */
676 age = (now - ifa->ifa_tstamp +
677 ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
678
679 if (ifa->ifa_valid_lft != INFINITY_LIFE_TIME &&
680 age >= ifa->ifa_valid_lft) {
681 struct in_ifaddr **ifap;
682
683 for (ifap = &ifa->ifa_dev->ifa_list;
684 *ifap != NULL; ifap = &(*ifap)->ifa_next) {
685 if (*ifap == ifa) {
686 inet_del_ifa(ifa->ifa_dev,
687 ifap, 1);
688 break;
689 }
690 }
691 } else if (ifa->ifa_preferred_lft !=
692 INFINITY_LIFE_TIME &&
693 age >= ifa->ifa_preferred_lft &&
694 !(ifa->ifa_flags & IFA_F_DEPRECATED)) {
695 ifa->ifa_flags |= IFA_F_DEPRECATED;
696 rtmsg_ifa(RTM_NEWADDR, ifa, NULL, 0);
697 }
698 }
699 rtnl_unlock();
5c766d64 700 }
5c766d64
JP		 701
702 next_sec = round_jiffies_up(next);
703 next_sched = next;
704
705 /* If rounded timeout is accurate enough, accept it. */
706 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
707 next_sched = next_sec;
708
709 now = jiffies;
710 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
711 if (time_before(next_sched, now + ADDRCONF_TIMER_FUZZ_MAX))
712 next_sched = now + ADDRCONF_TIMER_FUZZ_MAX;
713
906e073f 714 queue_delayed_work(system_power_efficient_wq, &check_lifetime_work,
715 next_sched - now);
5c766d64
JP		 716}
717
718static void set_ifa_lifetime(struct in_ifaddr *ifa, __u32 valid_lft,
719 __u32 prefered_lft)
720{
721 unsigned long timeout;
722
723 ifa->ifa_flags &= ~(IFA_F_PERMANENT | IFA_F_DEPRECATED);
724
725 timeout = addrconf_timeout_fixup(valid_lft, HZ);
726 if (addrconf_finite_timeout(timeout))
727 ifa->ifa_valid_lft = timeout;
728 else
729 ifa->ifa_flags |= IFA_F_PERMANENT;
730
731 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
732 if (addrconf_finite_timeout(timeout)) {
733 if (timeout == 0)
734 ifa->ifa_flags |= IFA_F_DEPRECATED;
735 ifa->ifa_preferred_lft = timeout;
736 }
737 ifa->ifa_tstamp = jiffies;
738 if (!ifa->ifa_cstamp)
739 ifa->ifa_cstamp = ifa->ifa_tstamp;
740}
741
742static struct in_ifaddr *rtm_to_ifaddr(struct net *net, struct nlmsghdr *nlh,
743 __u32 *pvalid_lft, __u32 *pprefered_lft)
1da177e4 744{
5c753978
TG		 745 struct nlattr *tb[IFA_MAX+1];
746 struct in_ifaddr *ifa;
747 struct ifaddrmsg *ifm;
1da177e4
LT		 748 struct net_device *dev;
749 struct in_device *in_dev;
7b218574 750 int err;
1da177e4 751
5c753978
TG		 752 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv4_policy);
753 if (err < 0)
754 goto errout;
1da177e4 755
5c753978 756 ifm = nlmsg_data(nlh);
7b218574 757 err = -EINVAL;
51456b29 758 if (ifm->ifa_prefixlen > 32 || !tb[IFA_LOCAL])
5c753978 759 goto errout;
1da177e4 760
4b8aa9ab 761 dev = __dev_get_by_index(net, ifm->ifa_index);
7b218574 762 err = -ENODEV;
51456b29 763 if (!dev)
5c753978 764 goto errout;
1da177e4 765
5c753978 766 in_dev = __in_dev_get_rtnl(dev);
7b218574 767 err = -ENOBUFS;
51456b29 768 if (!in_dev)
71e27da9 769 goto errout;
1da177e4 770
5c753978 771 ifa = inet_alloc_ifa();
51456b29 772 if (!ifa)
5c753978
TG		 773 /*
774 * A potential indev allocation can be left alive, it stays
775 * assigned to its device and is destroy with it.
776 */
5c753978 777 goto errout;
5c753978 778
a4e65d36 779 ipv4_devconf_setall(in_dev);
1d4c8c29 780 neigh_parms_data_state_setall(in_dev->arp_parms);
5c753978
TG		 781 in_dev_hold(in_dev);
782
51456b29 783 if (!tb[IFA_ADDRESS])
5c753978 784 tb[IFA_ADDRESS] = tb[IFA_LOCAL];
1da177e4 785
fd23c3b3 786 INIT_HLIST_NODE(&ifa->hash);
1da177e4
LT		 787 ifa->ifa_prefixlen = ifm->ifa_prefixlen;
788 ifa->ifa_mask = inet_make_mask(ifm->ifa_prefixlen);
ad6c8135
JP		 789 ifa->ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) :
790 ifm->ifa_flags;
1da177e4 791 ifa->ifa_scope = ifm->ifa_scope;
5c753978
TG		 792 ifa->ifa_dev = in_dev;
793
67b61f6c
JB		 794 ifa->ifa_local = nla_get_in_addr(tb[IFA_LOCAL]);
795 ifa->ifa_address = nla_get_in_addr(tb[IFA_ADDRESS]);
5c753978
TG		 796
797 if (tb[IFA_BROADCAST])
67b61f6c 798 ifa->ifa_broadcast = nla_get_in_addr(tb[IFA_BROADCAST]);
5c753978 799
5c753978
TG		 800 if (tb[IFA_LABEL])
801 nla_strlcpy(ifa->ifa_label, tb[IFA_LABEL], IFNAMSIZ);
1da177e4
LT		 802 else
803 memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
804
5c766d64
JP		 805 if (tb[IFA_CACHEINFO]) {
806 struct ifa_cacheinfo *ci;
807
808 ci = nla_data(tb[IFA_CACHEINFO]);
809 if (!ci->ifa_valid || ci->ifa_prefered > ci->ifa_valid) {
810 err = -EINVAL;
446266b0 811 goto errout_free;
5c766d64
JP		 812 }
813 *pvalid_lft = ci->ifa_valid;
814 *pprefered_lft = ci->ifa_prefered;
815 }
816
5c753978
TG		 817 return ifa;
818
446266b0
DB		 819errout_free:
820 inet_free_ifa(ifa);
5c753978
TG		 821errout:
822 return ERR_PTR(err);
823}
824
5c766d64
JP		 825static struct in_ifaddr *find_matching_ifa(struct in_ifaddr *ifa)
826{
827 struct in_device *in_dev = ifa->ifa_dev;
828 struct in_ifaddr *ifa1, **ifap;
829
830 if (!ifa->ifa_local)
831 return NULL;
832
833 for (ifap = &in_dev->ifa_list; (ifa1 = *ifap) != NULL;
834 ifap = &ifa1->ifa_next) {
835 if (ifa1->ifa_mask == ifa->ifa_mask &&
836 inet_ifa_match(ifa1->ifa_address, ifa) &&
837 ifa1->ifa_local == ifa->ifa_local)
838 return ifa1;
839 }
840 return NULL;
841}
842
661d2967 843static int inet_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh)
5c753978 844{
3b1e0a65 845 struct net *net = sock_net(skb->sk);
5c753978 846 struct in_ifaddr *ifa;
5c766d64
JP		 847 struct in_ifaddr *ifa_existing;
848 __u32 valid_lft = INFINITY_LIFE_TIME;
849 __u32 prefered_lft = INFINITY_LIFE_TIME;
5c753978
TG		 850
851 ASSERT_RTNL();
852
5c766d64 853 ifa = rtm_to_ifaddr(net, nlh, &valid_lft, &prefered_lft);
5c753978
TG		 854 if (IS_ERR(ifa))
855 return PTR_ERR(ifa);
856
5c766d64
JP		 857 ifa_existing = find_matching_ifa(ifa);
858 if (!ifa_existing) {
859 /* It would be best to check for !NLM_F_CREATE here but
614d056c 860 * userspace already relies on not having to provide this.
5c766d64
JP		 861 */
862 set_ifa_lifetime(ifa, valid_lft, prefered_lft);
93a714d6
MC		 863 if (ifa->ifa_flags & IFA_F_MCAUTOJOIN) {
864 int ret = ip_mc_config(net->ipv4.mc_autojoin_sk,
865 true, ifa);
866
867 if (ret < 0) {
868 inet_free_ifa(ifa);
869 return ret;
870 }
871 }
5c766d64
JP		 872 return __inet_insert_ifa(ifa, nlh, NETLINK_CB(skb).portid);
873 } else {
874 inet_free_ifa(ifa);
875
876 if (nlh->nlmsg_flags & NLM_F_EXCL ||
877 !(nlh->nlmsg_flags & NLM_F_REPLACE))
878 return -EEXIST;
34e2ed34
JP		 879 ifa = ifa_existing;
880 set_ifa_lifetime(ifa, valid_lft, prefered_lft);
05a324b9 881 cancel_delayed_work(&check_lifetime_work);
906e073f 882 queue_delayed_work(system_power_efficient_wq,
883 &check_lifetime_work, 0);
34e2ed34 884 rtmsg_ifa(RTM_NEWADDR, ifa, nlh, NETLINK_CB(skb).portid);
5c766d64
JP		 885 }
886 return 0;
1da177e4
LT		 887}
888
889/*
890 * Determine a default network mask, based on the IP address.
891 */
892
40384999 893static int inet_abc_len(__be32 addr)
1da177e4
LT		 894{
895 int rc = -1; /* Something else, probably a multicast. */
896
f97c1e0c 897 if (ipv4_is_zeronet(addr))
e905a9ed 898 rc = 0;
1da177e4 899 else {
714e85be 900 __u32 haddr = ntohl(addr);
1da177e4 901
714e85be 902 if (IN_CLASSA(haddr))
1da177e4 903 rc = 8;
714e85be 904 else if (IN_CLASSB(haddr))
1da177e4 905 rc = 16;
714e85be 906 else if (IN_CLASSC(haddr))
1da177e4
LT		 907 rc = 24;
908 }
909
e905a9ed 910 return rc;
1da177e4
LT		 911}
912
913
e5b13cb1 914int devinet_ioctl(struct net *net, unsigned int cmd, void __user *arg)
1da177e4
LT		 915{
916 struct ifreq ifr;
917 struct sockaddr_in sin_orig;
918 struct sockaddr_in *sin = (struct sockaddr_in *)&ifr.ifr_addr;
919 struct in_device *in_dev;
920 struct in_ifaddr **ifap = NULL;
921 struct in_ifaddr *ifa = NULL;
922 struct net_device *dev;
923 char *colon;
924 int ret = -EFAULT;
925 int tryaddrmatch = 0;
926
927 /*
928 * Fetch the caller's info block into kernel space
929 */
930
931 if (copy_from_user(&ifr, arg, sizeof(struct ifreq)))
932 goto out;
933 ifr.ifr_name[IFNAMSIZ - 1] = 0;
934
935 /* save original address for comparison */
936 memcpy(&sin_orig, sin, sizeof(*sin));
937
938 colon = strchr(ifr.ifr_name, ':');
939 if (colon)
940 *colon = 0;
941
e5b13cb1 942 dev_load(net, ifr.ifr_name);
1da177e4 943
132adf54 944 switch (cmd) {
1da177e4
LT		 945 case SIOCGIFADDR: /* Get interface address */
946 case SIOCGIFBRDADDR: /* Get the broadcast address */
947 case SIOCGIFDSTADDR: /* Get the destination address */
948 case SIOCGIFNETMASK: /* Get the netmask for the interface */
949 /* Note that these ioctls will not sleep,
950 so that we do not impose a lock.
951 One day we will be forced to put shlock here (I mean SMP)
952 */
953 tryaddrmatch = (sin_orig.sin_family == AF_INET);
954 memset(sin, 0, sizeof(*sin));
955 sin->sin_family = AF_INET;
956 break;
957
958 case SIOCSIFFLAGS:
bf5b30b8 959 ret = -EPERM;
52e804c6 960 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1da177e4
LT		 961 goto out;
962 break;
963 case SIOCSIFADDR: /* Set interface address (and family) */
964 case SIOCSIFBRDADDR: /* Set the broadcast address */
965 case SIOCSIFDSTADDR: /* Set the destination address */
966 case SIOCSIFNETMASK: /* Set the netmask for the interface */
bf5b30b8 967 ret = -EPERM;
52e804c6 968 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1da177e4
LT		 969 goto out;
970 ret = -EINVAL;
971 if (sin->sin_family != AF_INET)
972 goto out;
973 break;
974 default:
975 ret = -EINVAL;
976 goto out;
977 }
978
979 rtnl_lock();
980
981 ret = -ENODEV;
9f9354b9
ED		 982 dev = __dev_get_by_name(net, ifr.ifr_name);
983 if (!dev)
1da177e4
LT		 984 goto done;
985
986 if (colon)
987 *colon = ':';
988
9f9354b9
ED		 989 in_dev = __in_dev_get_rtnl(dev);
990 if (in_dev) {
1da177e4
LT		 991 if (tryaddrmatch) {
992 /* Matthias Andree */
993 /* compare label and address (4.4BSD style) */
994 /* note: we only do this for a limited set of ioctls
995 and only if the original address family was AF_INET.
996 This is checked above. */
997 for (ifap = &in_dev->ifa_list; (ifa = *ifap) != NULL;
998 ifap = &ifa->ifa_next) {
999 if (!strcmp(ifr.ifr_name, ifa->ifa_label) &&
1000 sin_orig.sin_addr.s_addr ==
6c91afe1 1001 ifa->ifa_local) {
1da177e4
LT		 1002 break; /* found */
1003 }
1004 }
1005 }
1006 /* we didn't get a match, maybe the application is
1007 4.3BSD-style and passed in junk so we fall back to
1008 comparing just the label */
1009 if (!ifa) {
1010 for (ifap = &in_dev->ifa_list; (ifa = *ifap) != NULL;
1011 ifap = &ifa->ifa_next)
1012 if (!strcmp(ifr.ifr_name, ifa->ifa_label))
1013 break;
1014 }
1015 }
1016
1017 ret = -EADDRNOTAVAIL;
1018 if (!ifa && cmd != SIOCSIFADDR && cmd != SIOCSIFFLAGS)
1019 goto done;
1020
132adf54 1021 switch (cmd) {
1da177e4
LT		 1022 case SIOCGIFADDR: /* Get interface address */
1023 sin->sin_addr.s_addr = ifa->ifa_local;
1024 goto rarok;
1025
1026 case SIOCGIFBRDADDR: /* Get the broadcast address */
1027 sin->sin_addr.s_addr = ifa->ifa_broadcast;
1028 goto rarok;
1029
1030 case SIOCGIFDSTADDR: /* Get the destination address */
1031 sin->sin_addr.s_addr = ifa->ifa_address;
1032 goto rarok;
1033
1034 case SIOCGIFNETMASK: /* Get the netmask for the interface */
1035 sin->sin_addr.s_addr = ifa->ifa_mask;
1036 goto rarok;
1037
1038 case SIOCSIFFLAGS:
1039 if (colon) {
1040 ret = -EADDRNOTAVAIL;
1041 if (!ifa)
1042 break;
1043 ret = 0;
1044 if (!(ifr.ifr_flags & IFF_UP))
1045 inet_del_ifa(in_dev, ifap, 1);
1046 break;
1047 }
1048 ret = dev_change_flags(dev, ifr.ifr_flags);
1049 break;
1050
1051 case SIOCSIFADDR: /* Set interface address (and family) */
1052 ret = -EINVAL;
1053 if (inet_abc_len(sin->sin_addr.s_addr) < 0)
1054 break;
1055
1056 if (!ifa) {
1057 ret = -ENOBUFS;
9f9354b9
ED		 1058 ifa = inet_alloc_ifa();
1059 if (!ifa)
1da177e4 1060 break;
c7e2e1d7 1061 INIT_HLIST_NODE(&ifa->hash);
1da177e4
LT		 1062 if (colon)
1063 memcpy(ifa->ifa_label, ifr.ifr_name, IFNAMSIZ);
1064 else
1065 memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
1066 } else {
1067 ret = 0;
1068 if (ifa->ifa_local == sin->sin_addr.s_addr)
1069 break;
1070 inet_del_ifa(in_dev, ifap, 0);
1071 ifa->ifa_broadcast = 0;
148f9729 1072 ifa->ifa_scope = 0;
1da177e4
LT		 1073 }
1074
1075 ifa->ifa_address = ifa->ifa_local = sin->sin_addr.s_addr;
1076
1077 if (!(dev->flags & IFF_POINTOPOINT)) {
1078 ifa->ifa_prefixlen = inet_abc_len(ifa->ifa_address);
1079 ifa->ifa_mask = inet_make_mask(ifa->ifa_prefixlen);
1080 if ((dev->flags & IFF_BROADCAST) &&
1081 ifa->ifa_prefixlen < 31)
1082 ifa->ifa_broadcast = ifa->ifa_address |
1083 ~ifa->ifa_mask;
1084 } else {
1085 ifa->ifa_prefixlen = 32;
1086 ifa->ifa_mask = inet_make_mask(32);
1087 }
5c766d64 1088 set_ifa_lifetime(ifa, INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
1da177e4
LT		 1089 ret = inet_set_ifa(dev, ifa);
1090 break;
1091
1092 case SIOCSIFBRDADDR: /* Set the broadcast address */
1093 ret = 0;
1094 if (ifa->ifa_broadcast != sin->sin_addr.s_addr) {
1095 inet_del_ifa(in_dev, ifap, 0);
1096 ifa->ifa_broadcast = sin->sin_addr.s_addr;
1097 inet_insert_ifa(ifa);
1098 }
1099 break;
1100
1101 case SIOCSIFDSTADDR: /* Set the destination address */
1102 ret = 0;
1103 if (ifa->ifa_address == sin->sin_addr.s_addr)
1104 break;
1105 ret = -EINVAL;
1106 if (inet_abc_len(sin->sin_addr.s_addr) < 0)
1107 break;
1108 ret = 0;
1109 inet_del_ifa(in_dev, ifap, 0);
1110 ifa->ifa_address = sin->sin_addr.s_addr;
1111 inet_insert_ifa(ifa);
1112 break;
1113
1114 case SIOCSIFNETMASK: /* Set the netmask for the interface */
1115
1116 /*
1117 * The mask we set must be legal.
1118 */
1119 ret = -EINVAL;
1120 if (bad_mask(sin->sin_addr.s_addr, 0))
1121 break;
1122 ret = 0;
1123 if (ifa->ifa_mask != sin->sin_addr.s_addr) {
a144ea4b 1124 __be32 old_mask = ifa->ifa_mask;
1da177e4
LT		 1125 inet_del_ifa(in_dev, ifap, 0);
1126 ifa->ifa_mask = sin->sin_addr.s_addr;
1127 ifa->ifa_prefixlen = inet_mask_len(ifa->ifa_mask);
1128
1129 /* See if current broadcast address matches
1130 * with current netmask, then recalculate
1131 * the broadcast address. Otherwise it's a
1132 * funny address, so don't touch it since
1133 * the user seems to know what (s)he's doing...
1134 */
1135 if ((dev->flags & IFF_BROADCAST) &&
1136 (ifa->ifa_prefixlen < 31) &&
1137 (ifa->ifa_broadcast ==
dcab5e1e 1138 (ifa->ifa_local|~old_mask))) {
1da177e4
LT		 1139 ifa->ifa_broadcast = (ifa->ifa_local |
1140 ~sin->sin_addr.s_addr);
1141 }
1142 inet_insert_ifa(ifa);
1143 }
1144 break;
1145 }
1146done:
1147 rtnl_unlock();
1148out:
1149 return ret;
1150rarok:
1151 rtnl_unlock();
1152 ret = copy_to_user(arg, &ifr, sizeof(struct ifreq)) ? -EFAULT : 0;
1153 goto out;
1154}
1155
1156static int inet_gifconf(struct net_device *dev, char __user *buf, int len)
1157{
e5ed6399 1158 struct in_device *in_dev = __in_dev_get_rtnl(dev);
1da177e4
LT		 1159 struct in_ifaddr *ifa;
1160 struct ifreq ifr;
1161 int done = 0;
1162
9f9354b9 1163 if (!in_dev)
1da177e4
LT		 1164 goto out;
1165
9f9354b9 1166 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
1da177e4
LT		 1167 if (!buf) {
1168 done += sizeof(ifr);
1169 continue;
1170 }
1171 if (len < (int) sizeof(ifr))
1172 break;
1173 memset(&ifr, 0, sizeof(struct ifreq));
4299c8a9 1174 strcpy(ifr.ifr_name, ifa->ifa_label);
1da177e4
LT		 1175
1176 (*(struct sockaddr_in *)&ifr.ifr_addr).sin_family = AF_INET;
1177 (*(struct sockaddr_in *)&ifr.ifr_addr).sin_addr.s_addr =
1178 ifa->ifa_local;
1179
1180 if (copy_to_user(buf, &ifr, sizeof(struct ifreq))) {
1181 done = -EFAULT;
1182 break;
1183 }
1184 buf += sizeof(struct ifreq);
1185 len -= sizeof(struct ifreq);
1186 done += sizeof(struct ifreq);
1187 }
1188out:
1189 return done;
1190}
1191
a61ced5d 1192__be32 inet_select_addr(const struct net_device *dev, __be32 dst, int scope)
1da177e4 1193{
a61ced5d 1194 __be32 addr = 0;
1da177e4 1195 struct in_device *in_dev;
c346dca1 1196 struct net *net = dev_net(dev);
3f2fb9a8 1197 int master_idx;
1da177e4
LT		 1198
1199 rcu_read_lock();
e5ed6399 1200 in_dev = __in_dev_get_rcu(dev);
1da177e4
LT		 1201 if (!in_dev)
1202 goto no_in_dev;
1203
1204 for_primary_ifa(in_dev) {
1205 if (ifa->ifa_scope > scope)
1206 continue;
1207 if (!dst || inet_ifa_match(dst, ifa)) {
1208 addr = ifa->ifa_local;
1209 break;
1210 }
1211 if (!addr)
1212 addr = ifa->ifa_local;
1213 } endfor_ifa(in_dev);
1da177e4
LT		 1214
1215 if (addr)
c6d14c84 1216 goto out_unlock;
9f9354b9 1217no_in_dev:
3f2fb9a8 1218 master_idx = l3mdev_master_ifindex_rcu(dev);
1da177e4
LT		 1219
1220 /* Not loopback addresses on loopback should be preferred
ca9f1fd2 1221 in this case. It is important that lo is the first interface
1da177e4
LT		 1222 in dev_base list.
1223 */
c6d14c84 1224 for_each_netdev_rcu(net, dev) {
3f2fb9a8
DA		 1225 if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1226 continue;
1227
9f9354b9
ED		 1228 in_dev = __in_dev_get_rcu(dev);
1229 if (!in_dev)
1da177e4
LT		 1230 continue;
1231
1232 for_primary_ifa(in_dev) {
1233 if (ifa->ifa_scope != RT_SCOPE_LINK &&
1234 ifa->ifa_scope <= scope) {
1235 addr = ifa->ifa_local;
c6d14c84 1236 goto out_unlock;
1da177e4
LT		 1237 }
1238 } endfor_ifa(in_dev);
1239 }
c6d14c84 1240out_unlock:
1da177e4 1241 rcu_read_unlock();
1da177e4
LT		 1242 return addr;
1243}
9f9354b9 1244EXPORT_SYMBOL(inet_select_addr);
1da177e4 1245
60cad5da
AV		 1246static __be32 confirm_addr_indev(struct in_device *in_dev, __be32 dst,
1247 __be32 local, int scope)
1da177e4
LT		 1248{
1249 int same = 0;
a144ea4b 1250 __be32 addr = 0;
1da177e4
LT		 1251
1252 for_ifa(in_dev) {
1253 if (!addr &&
1254 (local == ifa->ifa_local || !local) &&
1255 ifa->ifa_scope <= scope) {
1256 addr = ifa->ifa_local;
1257 if (same)
1258 break;
1259 }
1260 if (!same) {
1261 same = (!local || inet_ifa_match(local, ifa)) &&
1262 (!dst || inet_ifa_match(dst, ifa));
1263 if (same && addr) {
1264 if (local || !dst)
1265 break;
1266 /* Is the selected addr into dst subnet? */
1267 if (inet_ifa_match(addr, ifa))
1268 break;
1269 /* No, then can we use new local src? */
1270 if (ifa->ifa_scope <= scope) {
1271 addr = ifa->ifa_local;
1272 break;
1273 }
1274 /* search for large dst subnet for addr */
1275 same = 0;
1276 }
1277 }
1278 } endfor_ifa(in_dev);
1279
9f9354b9 1280 return same ? addr : 0;
1da177e4
LT		 1281}
1282
1283/*
1284 * Confirm that local IP address exists using wildcards:
b601fa19
ND		 1285 * - net: netns to check, cannot be NULL
1286 * - in_dev: only on this interface, NULL=any interface
1da177e4
LT		 1287 * - dst: only in the same subnet as dst, 0=any dst
1288 * - local: address, 0=autoselect the local address
1289 * - scope: maximum allowed scope value for the local address
1290 */
b601fa19 1291__be32 inet_confirm_addr(struct net *net, struct in_device *in_dev,
9bd85e32 1292 __be32 dst, __be32 local, int scope)
1da177e4 1293{
60cad5da 1294 __be32 addr = 0;
9bd85e32 1295 struct net_device *dev;
1da177e4 1296
00db4124 1297 if (in_dev)
9bd85e32 1298 return confirm_addr_indev(in_dev, dst, local, scope);
1da177e4 1299
1da177e4 1300 rcu_read_lock();
c6d14c84 1301 for_each_netdev_rcu(net, dev) {
9f9354b9
ED		 1302 in_dev = __in_dev_get_rcu(dev);
1303 if (in_dev) {
1da177e4
LT		 1304 addr = confirm_addr_indev(in_dev, dst, local, scope);
1305 if (addr)
1306 break;
1307 }
1308 }
1309 rcu_read_unlock();
1da177e4
LT		 1310
1311 return addr;
1312}
eaddcd76 1313EXPORT_SYMBOL(inet_confirm_addr);
1da177e4
LT		 1314
1315/*
1316 * Device notifier
1317 */
1318
1319int register_inetaddr_notifier(struct notifier_block *nb)
1320{
e041c683 1321 return blocking_notifier_chain_register(&inetaddr_chain, nb);
1da177e4 1322}
9f9354b9 1323EXPORT_SYMBOL(register_inetaddr_notifier);
1da177e4
LT		 1324
1325int unregister_inetaddr_notifier(struct notifier_block *nb)
1326{
e041c683 1327 return blocking_notifier_chain_unregister(&inetaddr_chain, nb);
1da177e4 1328}
9f9354b9 1329EXPORT_SYMBOL(unregister_inetaddr_notifier);
1da177e4 1330
9f9354b9
ED		 1331/* Rename ifa_labels for a device name change. Make some effort to preserve
1332 * existing alias numbering and to create unique labels if possible.
1da177e4
LT		 1333*/
1334static void inetdev_changename(struct net_device *dev, struct in_device *in_dev)
e905a9ed 1335{
1da177e4
LT		 1336 struct in_ifaddr *ifa;
1337 int named = 0;
1338
e905a9ed
YH		 1339 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
1340 char old[IFNAMSIZ], *dot;
1da177e4
LT		 1341
1342 memcpy(old, ifa->ifa_label, IFNAMSIZ);
e905a9ed 1343 memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
1da177e4 1344 if (named++ == 0)
573bf470 1345 goto skip;
44344b2a 1346 dot = strchr(old, ':');
51456b29 1347 if (!dot) {
e905a9ed 1348 sprintf(old, ":%d", named);
1da177e4
LT		 1349 dot = old;
1350 }
9f9354b9 1351 if (strlen(dot) + strlen(dev->name) < IFNAMSIZ)
e905a9ed 1352 strcat(ifa->ifa_label, dot);
9f9354b9 1353 else
e905a9ed 1354 strcpy(ifa->ifa_label + (IFNAMSIZ - strlen(dot) - 1), dot);
573bf470
TG		 1355skip:
1356 rtmsg_ifa(RTM_NEWADDR, ifa, NULL, 0);
e905a9ed
YH		 1357 }
1358}
1da177e4 1359
40384999 1360static bool inetdev_valid_mtu(unsigned int mtu)
06770843
BL		 1361{
1362 return mtu >= 68;
1363}
1364
d11327ad
IC		 1365static void inetdev_send_gratuitous_arp(struct net_device *dev,
1366 struct in_device *in_dev)
1367
1368{
b76d0789 1369 struct in_ifaddr *ifa;
d11327ad 1370
b76d0789
ZK		 1371 for (ifa = in_dev->ifa_list; ifa;
1372 ifa = ifa->ifa_next) {
1373 arp_send(ARPOP_REQUEST, ETH_P_ARP,
1374 ifa->ifa_local, dev,
1375 ifa->ifa_local, NULL,
1376 dev->dev_addr, NULL);
1377 }
d11327ad
IC		 1378}
1379
1da177e4
LT		 1380/* Called only under RTNL semaphore */
1381
1382static int inetdev_event(struct notifier_block *this, unsigned long event,
1383 void *ptr)
1384{
351638e7 1385 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
748e2d93 1386 struct in_device *in_dev = __in_dev_get_rtnl(dev);
0115e8e3 1387
1da177e4
LT		 1388 ASSERT_RTNL();
1389
1390 if (!in_dev) {
8030f544 1391 if (event == NETDEV_REGISTER) {
1da177e4 1392 in_dev = inetdev_init(dev);
20e61da7
WC		 1393 if (IS_ERR(in_dev))
1394 return notifier_from_errno(PTR_ERR(in_dev));
0cc217e1 1395 if (dev->flags & IFF_LOOPBACK) {
42f811b8
HX		 1396 IN_DEV_CONF_SET(in_dev, NOXFRM, 1);
1397 IN_DEV_CONF_SET(in_dev, NOPOLICY, 1);
8030f544 1398 }
06770843
BL		 1399 } else if (event == NETDEV_CHANGEMTU) {
1400 /* Re-enabling IP */
1401 if (inetdev_valid_mtu(dev->mtu))
1402 in_dev = inetdev_init(dev);
1da177e4
LT		 1403 }
1404 goto out;
1405 }
1406
1407 switch (event) {
1408 case NETDEV_REGISTER:
91df42be 1409 pr_debug("%s: bug\n", __func__);
a9b3cd7f 1410 RCU_INIT_POINTER(dev->ip_ptr, NULL);
1da177e4
LT		 1411 break;
1412 case NETDEV_UP:
06770843 1413 if (!inetdev_valid_mtu(dev->mtu))
1da177e4 1414 break;
0cc217e1 1415 if (dev->flags & IFF_LOOPBACK) {
9f9354b9
ED		 1416 struct in_ifaddr *ifa = inet_alloc_ifa();
1417
1418 if (ifa) {
fd23c3b3 1419 INIT_HLIST_NODE(&ifa->hash);
1da177e4
LT		 1420 ifa->ifa_local =
1421 ifa->ifa_address = htonl(INADDR_LOOPBACK);
1422 ifa->ifa_prefixlen = 8;
1423 ifa->ifa_mask = inet_make_mask(8);
1424 in_dev_hold(in_dev);
1425 ifa->ifa_dev = in_dev;
1426 ifa->ifa_scope = RT_SCOPE_HOST;
1427 memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
5c766d64
JP		 1428 set_ifa_lifetime(ifa, INFINITY_LIFE_TIME,
1429 INFINITY_LIFE_TIME);
dfd1582d
JP		 1430 ipv4_devconf_setall(in_dev);
1431 neigh_parms_data_state_setall(in_dev->arp_parms);
1da177e4
LT		 1432 inet_insert_ifa(ifa);
1433 }
1434 }
1435 ip_mc_up(in_dev);
eefef1cf
SH		 1436 /* fall through */
1437 case NETDEV_CHANGEADDR:
d11327ad
IC		 1438 if (!IN_DEV_ARP_NOTIFY(in_dev))
1439 break;
1440 /* fall through */
1441 case NETDEV_NOTIFY_PEERS:
a21090cf 1442 /* Send gratuitous ARP to notify of link change */
d11327ad 1443 inetdev_send_gratuitous_arp(dev, in_dev);
1da177e4
LT		 1444 break;
1445 case NETDEV_DOWN:
1446 ip_mc_down(in_dev);
1447 break;
93d9b7d7 1448 case NETDEV_PRE_TYPE_CHANGE:
75c78500
MS		 1449 ip_mc_unmap(in_dev);
1450 break;
93d9b7d7 1451 case NETDEV_POST_TYPE_CHANGE:
75c78500
MS		 1452 ip_mc_remap(in_dev);
1453 break;
1da177e4 1454 case NETDEV_CHANGEMTU:
06770843 1455 if (inetdev_valid_mtu(dev->mtu))
1da177e4 1456 break;
06770843 1457 /* disable IP when MTU is not enough */
1da177e4
LT		 1458 case NETDEV_UNREGISTER:
1459 inetdev_destroy(in_dev);
1460 break;
1461 case NETDEV_CHANGENAME:
1462 /* Do not notify about label change, this event is
1463 * not interesting to applications using netlink.
1464 */
1465 inetdev_changename(dev, in_dev);
1466
51602b2a 1467 devinet_sysctl_unregister(in_dev);
66f27a52 1468 devinet_sysctl_register(in_dev);
1da177e4
LT		 1469 break;
1470 }
1471out:
1472 return NOTIFY_DONE;
1473}
1474
1475static struct notifier_block ip_netdev_notifier = {
539afedf 1476 .notifier_call = inetdev_event,
1da177e4
LT		 1477};
1478
40384999 1479static size_t inet_nlmsg_size(void)
339bf98f
TG		 1480{
1481 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
1482 + nla_total_size(4) /* IFA_ADDRESS */
1483 + nla_total_size(4) /* IFA_LOCAL */
1484 + nla_total_size(4) /* IFA_BROADCAST */
ad6c8135 1485 + nla_total_size(IFNAMSIZ) /* IFA_LABEL */
63b5f152
GU		 1486 + nla_total_size(4) /* IFA_FLAGS */
1487 + nla_total_size(sizeof(struct ifa_cacheinfo)); /* IFA_CACHEINFO */
339bf98f
TG		 1488}
1489
5c766d64
JP		 1490static inline u32 cstamp_delta(unsigned long cstamp)
1491{
1492 return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
1493}
1494
1495static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
1496 unsigned long tstamp, u32 preferred, u32 valid)
1497{
1498 struct ifa_cacheinfo ci;
1499
1500 ci.cstamp = cstamp_delta(cstamp);
1501 ci.tstamp = cstamp_delta(tstamp);
1502 ci.ifa_prefered = preferred;
1503 ci.ifa_valid = valid;
1504
1505 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
1506}
1507
1da177e4 1508static int inet_fill_ifaddr(struct sk_buff *skb, struct in_ifaddr *ifa,
15e47304 1509 u32 portid, u32 seq, int event, unsigned int flags)
1da177e4
LT		 1510{
1511 struct ifaddrmsg *ifm;
1512 struct nlmsghdr *nlh;
5c766d64 1513 u32 preferred, valid;
1da177e4 1514
15e47304 1515 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*ifm), flags);
51456b29 1516 if (!nlh)
26932566 1517 return -EMSGSIZE;
47f68512
TG		 1518
1519 ifm = nlmsg_data(nlh);
1da177e4
LT		 1520 ifm->ifa_family = AF_INET;
1521 ifm->ifa_prefixlen = ifa->ifa_prefixlen;
5c766d64 1522 ifm->ifa_flags = ifa->ifa_flags;
1da177e4
LT		 1523 ifm->ifa_scope = ifa->ifa_scope;
1524 ifm->ifa_index = ifa->ifa_dev->dev->ifindex;
47f68512 1525
5c766d64
JP		 1526 if (!(ifm->ifa_flags & IFA_F_PERMANENT)) {
1527 preferred = ifa->ifa_preferred_lft;
1528 valid = ifa->ifa_valid_lft;
1529 if (preferred != INFINITY_LIFE_TIME) {
1530 long tval = (jiffies - ifa->ifa_tstamp) / HZ;
1531
1532 if (preferred > tval)
1533 preferred -= tval;
1534 else
1535 preferred = 0;
1536 if (valid != INFINITY_LIFE_TIME) {
1537 if (valid > tval)
1538 valid -= tval;
1539 else
1540 valid = 0;
1541 }
1542 }
1543 } else {
1544 preferred = INFINITY_LIFE_TIME;
1545 valid = INFINITY_LIFE_TIME;
1546 }
f3756b79 1547 if ((ifa->ifa_address &&
930345ea 1548 nla_put_in_addr(skb, IFA_ADDRESS, ifa->ifa_address)) ||
f3756b79 1549 (ifa->ifa_local &&
930345ea 1550 nla_put_in_addr(skb, IFA_LOCAL, ifa->ifa_local)) ||
f3756b79 1551 (ifa->ifa_broadcast &&
930345ea 1552 nla_put_in_addr(skb, IFA_BROADCAST, ifa->ifa_broadcast)) ||
f3756b79 1553 (ifa->ifa_label[0] &&
5c766d64 1554 nla_put_string(skb, IFA_LABEL, ifa->ifa_label)) ||
ad6c8135 1555 nla_put_u32(skb, IFA_FLAGS, ifa->ifa_flags) ||
5c766d64
JP		 1556 put_cacheinfo(skb, ifa->ifa_cstamp, ifa->ifa_tstamp,
1557 preferred, valid))
f3756b79 1558 goto nla_put_failure;
1da177e4 1559
053c095a
JB		 1560 nlmsg_end(skb, nlh);
1561 return 0;
47f68512
TG		 1562
1563nla_put_failure:
26932566
PM		 1564 nlmsg_cancel(skb, nlh);
1565 return -EMSGSIZE;
1da177e4
LT		 1566}
1567
1568static int inet_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
1569{
3b1e0a65 1570 struct net *net = sock_net(skb->sk);
eec4df98
ED		 1571 int h, s_h;
1572 int idx, s_idx;
1573 int ip_idx, s_ip_idx;
1da177e4
LT		 1574 struct net_device *dev;
1575 struct in_device *in_dev;
1576 struct in_ifaddr *ifa;
eec4df98 1577 struct hlist_head *head;
1da177e4 1578
eec4df98
ED		 1579 s_h = cb->args[0];
1580 s_idx = idx = cb->args[1];
1581 s_ip_idx = ip_idx = cb->args[2];
1582
1583 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1584 idx = 0;
1585 head = &net->dev_index_head[h];
1586 rcu_read_lock();
0465277f
ND		 1587 cb->seq = atomic_read(&net->ipv4.dev_addr_genid) ^
1588 net->dev_base_seq;
b67bfe0d 1589 hlist_for_each_entry_rcu(dev, head, index_hlist) {
eec4df98
ED		 1590 if (idx < s_idx)
1591 goto cont;
4b97efdf 1592 if (h > s_h || idx > s_idx)
eec4df98
ED		 1593 s_ip_idx = 0;
1594 in_dev = __in_dev_get_rcu(dev);
1595 if (!in_dev)
1596 goto cont;
1da177e4 1597
eec4df98
ED		 1598 for (ifa = in_dev->ifa_list, ip_idx = 0; ifa;
1599 ifa = ifa->ifa_next, ip_idx++) {
1600 if (ip_idx < s_ip_idx)
1601 continue;
1602 if (inet_fill_ifaddr(skb, ifa,
15e47304 1603 NETLINK_CB(cb->skb).portid,
1da177e4 1604 cb->nlh->nlmsg_seq,
053c095a 1605 RTM_NEWADDR, NLM_F_MULTI) < 0) {
eec4df98
ED		 1606 rcu_read_unlock();
1607 goto done;
1608 }
0465277f 1609 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
eec4df98 1610 }
7562f876 1611cont:
eec4df98
ED		 1612 idx++;
1613 }
1614 rcu_read_unlock();
1da177e4
LT		 1615 }
1616
1617done:
eec4df98
ED		 1618 cb->args[0] = h;
1619 cb->args[1] = idx;
1620 cb->args[2] = ip_idx;
1da177e4
LT		 1621
1622 return skb->len;
1623}
1624
539afedf 1625static void rtmsg_ifa(int event, struct in_ifaddr *ifa, struct nlmsghdr *nlh,
15e47304 1626 u32 portid)
1da177e4 1627{
47f68512 1628 struct sk_buff *skb;
d6062cbb
TG		 1629 u32 seq = nlh ? nlh->nlmsg_seq : 0;
1630 int err = -ENOBUFS;
4b8aa9ab 1631 struct net *net;
1da177e4 1632
c346dca1 1633 net = dev_net(ifa->ifa_dev->dev);
339bf98f 1634 skb = nlmsg_new(inet_nlmsg_size(), GFP_KERNEL);
51456b29 1635 if (!skb)
d6062cbb
TG		 1636 goto errout;
1637
15e47304 1638 err = inet_fill_ifaddr(skb, ifa, portid, seq, event, 0);
26932566
PM		 1639 if (err < 0) {
1640 /* -EMSGSIZE implies BUG in inet_nlmsg_size() */
1641 WARN_ON(err == -EMSGSIZE);
1642 kfree_skb(skb);
1643 goto errout;
1644 }
15e47304 1645 rtnl_notify(skb, net, portid, RTNLGRP_IPV4_IFADDR, nlh, GFP_KERNEL);
1ce85fe4 1646 return;
d6062cbb
TG		 1647errout:
1648 if (err < 0)
4b8aa9ab 1649 rtnl_set_sk_err(net, RTNLGRP_IPV4_IFADDR, err);
1da177e4
LT		 1650}
1651
b1974ed0
AR		 1652static size_t inet_get_link_af_size(const struct net_device *dev,
1653 u32 ext_filter_mask)
9f0f7272 1654{
1fc19aff 1655 struct in_device *in_dev = rcu_dereference_rtnl(dev->ip_ptr);
9f0f7272
TG		 1656
1657 if (!in_dev)
1658 return 0;
1659
1660 return nla_total_size(IPV4_DEVCONF_MAX * 4); /* IFLA_INET_CONF */
1661}
1662
d5566fd7
SV		 1663static int inet_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
1664 u32 ext_filter_mask)
9f0f7272 1665{
1fc19aff 1666 struct in_device *in_dev = rcu_dereference_rtnl(dev->ip_ptr);
9f0f7272
TG		 1667 struct nlattr *nla;
1668 int i;
1669
1670 if (!in_dev)
1671 return -ENODATA;
1672
1673 nla = nla_reserve(skb, IFLA_INET_CONF, IPV4_DEVCONF_MAX * 4);
51456b29 1674 if (!nla)
9f0f7272
TG		 1675 return -EMSGSIZE;
1676
1677 for (i = 0; i < IPV4_DEVCONF_MAX; i++)
1678 ((u32 *) nla_data(nla))[i] = in_dev->cnf.data[i];
1679
1680 return 0;
1681}
1682
1683static const struct nla_policy inet_af_policy[IFLA_INET_MAX+1] = {
1684 [IFLA_INET_CONF] = { .type = NLA_NESTED },
1685};
1686
cf7afbfe
TG		 1687static int inet_validate_link_af(const struct net_device *dev,
1688 const struct nlattr *nla)
9f0f7272 1689{
9f0f7272
TG		 1690 struct nlattr *a, *tb[IFLA_INET_MAX+1];
1691 int err, rem;
1692
f7fce74e 1693 if (dev && !__in_dev_get_rtnl(dev))
cf7afbfe 1694 return -EAFNOSUPPORT;
9f0f7272
TG		 1695
1696 err = nla_parse_nested(tb, IFLA_INET_MAX, nla, inet_af_policy);
1697 if (err < 0)
1698 return err;
1699
1700 if (tb[IFLA_INET_CONF]) {
1701 nla_for_each_nested(a, tb[IFLA_INET_CONF], rem) {
1702 int cfgid = nla_type(a);
1703
1704 if (nla_len(a) < 4)
1705 return -EINVAL;
1706
1707 if (cfgid <= 0 || cfgid > IPV4_DEVCONF_MAX)
1708 return -EINVAL;
1709 }
1710 }
1711
cf7afbfe
TG		 1712 return 0;
1713}
1714
1715static int inet_set_link_af(struct net_device *dev, const struct nlattr *nla)
1716{
f7fce74e 1717 struct in_device *in_dev = __in_dev_get_rtnl(dev);
cf7afbfe
TG		 1718 struct nlattr *a, *tb[IFLA_INET_MAX+1];
1719 int rem;
1720
1721 if (!in_dev)
1722 return -EAFNOSUPPORT;
1723
1724 if (nla_parse_nested(tb, IFLA_INET_MAX, nla, NULL) < 0)
1725 BUG();
1726
9f0f7272
TG		 1727 if (tb[IFLA_INET_CONF]) {
1728 nla_for_each_nested(a, tb[IFLA_INET_CONF], rem)
1729 ipv4_devconf_set(in_dev, nla_type(a), nla_get_u32(a));
1730 }
1731
1732 return 0;
1733}
1734
edc9e748
ND		 1735static int inet_netconf_msgsize_devconf(int type)
1736{
1737 int size = NLMSG_ALIGN(sizeof(struct netconfmsg))
1738 + nla_total_size(4); /* NETCONFA_IFINDEX */
1739
9e551110
ND		 1740 /* type -1 is used for ALL */
1741 if (type == -1 || type == NETCONFA_FORWARDING)
edc9e748 1742 size += nla_total_size(4);
cc535dfb
ND		 1743 if (type == -1 || type == NETCONFA_RP_FILTER)
1744 size += nla_total_size(4);
d67b8c61
ND		 1745 if (type == -1 || type == NETCONFA_MC_FORWARDING)
1746 size += nla_total_size(4);
09aea5df 1747 if (type == -1 || type == NETCONFA_PROXY_NEIGH)
f085ff1c 1748 size += nla_total_size(4);
974d7af5
AG		 1749 if (type == -1 || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
1750 size += nla_total_size(4);
edc9e748
ND		 1751
1752 return size;
1753}
1754
1755static int inet_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
1756 struct ipv4_devconf *devconf, u32 portid,
1757 u32 seq, int event, unsigned int flags,
1758 int type)
1759{
1760 struct nlmsghdr *nlh;
1761 struct netconfmsg *ncm;
1762
1763 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
1764 flags);
51456b29 1765 if (!nlh)
edc9e748
ND		 1766 return -EMSGSIZE;
1767
1768 ncm = nlmsg_data(nlh);
1769 ncm->ncm_family = AF_INET;
1770
1771 if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
1772 goto nla_put_failure;
1773
9e551110
ND		 1774 /* type -1 is used for ALL */
1775 if ((type == -1 || type == NETCONFA_FORWARDING) &&
edc9e748
ND		 1776 nla_put_s32(skb, NETCONFA_FORWARDING,
1777 IPV4_DEVCONF(*devconf, FORWARDING)) < 0)
1778 goto nla_put_failure;
cc535dfb
ND		 1779 if ((type == -1 || type == NETCONFA_RP_FILTER) &&
1780 nla_put_s32(skb, NETCONFA_RP_FILTER,
1781 IPV4_DEVCONF(*devconf, RP_FILTER)) < 0)
1782 goto nla_put_failure;
d67b8c61
ND		 1783 if ((type == -1 || type == NETCONFA_MC_FORWARDING) &&
1784 nla_put_s32(skb, NETCONFA_MC_FORWARDING,
1785 IPV4_DEVCONF(*devconf, MC_FORWARDING)) < 0)
1786 goto nla_put_failure;
09aea5df 1787 if ((type == -1 || type == NETCONFA_PROXY_NEIGH) &&
1788 nla_put_s32(skb, NETCONFA_PROXY_NEIGH,
f085ff1c 1789 IPV4_DEVCONF(*devconf, PROXY_ARP)) < 0)
1790 goto nla_put_failure;
974d7af5
AG		 1791 if ((type == -1 || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
1792 nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
1793 IPV4_DEVCONF(*devconf, IGNORE_ROUTES_WITH_LINKDOWN)) < 0)
1794 goto nla_put_failure;
edc9e748 1795
053c095a
JB		 1796 nlmsg_end(skb, nlh);
1797 return 0;
edc9e748
ND		 1798
1799nla_put_failure:
1800 nlmsg_cancel(skb, nlh);
1801 return -EMSGSIZE;
1802}
1803
d67b8c61
ND		 1804void inet_netconf_notify_devconf(struct net *net, int type, int ifindex,
1805 struct ipv4_devconf *devconf)
edc9e748
ND		 1806{
1807 struct sk_buff *skb;
1808 int err = -ENOBUFS;
1809
1810 skb = nlmsg_new(inet_netconf_msgsize_devconf(type), GFP_ATOMIC);
51456b29 1811 if (!skb)
edc9e748
ND		 1812 goto errout;
1813
1814 err = inet_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
1815 RTM_NEWNETCONF, 0, type);
1816 if (err < 0) {
1817 /* -EMSGSIZE implies BUG in inet_netconf_msgsize_devconf() */
1818 WARN_ON(err == -EMSGSIZE);
1819 kfree_skb(skb);
1820 goto errout;
1821 }
1822 rtnl_notify(skb, net, 0, RTNLGRP_IPV4_NETCONF, NULL, GFP_ATOMIC);
1823 return;
1824errout:
1825 if (err < 0)
1826 rtnl_set_sk_err(net, RTNLGRP_IPV4_NETCONF, err);
1827}
1828
9e551110
ND		 1829static const struct nla_policy devconf_ipv4_policy[NETCONFA_MAX+1] = {
1830 [NETCONFA_IFINDEX] = { .len = sizeof(int) },
1831 [NETCONFA_FORWARDING] = { .len = sizeof(int) },
cc535dfb 1832 [NETCONFA_RP_FILTER] = { .len = sizeof(int) },
09aea5df 1833 [NETCONFA_PROXY_NEIGH] = { .len = sizeof(int) },
974d7af5 1834 [NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN] = { .len = sizeof(int) },
9e551110
ND		 1835};
1836
1837static int inet_netconf_get_devconf(struct sk_buff *in_skb,
661d2967 1838 struct nlmsghdr *nlh)
9e551110
ND		 1839{
1840 struct net *net = sock_net(in_skb->sk);
1841 struct nlattr *tb[NETCONFA_MAX+1];
1842 struct netconfmsg *ncm;
1843 struct sk_buff *skb;
1844 struct ipv4_devconf *devconf;
1845 struct in_device *in_dev;
1846 struct net_device *dev;
1847 int ifindex;
1848 int err;
1849
1850 err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
1851 devconf_ipv4_policy);
1852 if (err < 0)
1853 goto errout;
1854
a97eb33f 1855 err = -EINVAL;
9e551110
ND		 1856 if (!tb[NETCONFA_IFINDEX])
1857 goto errout;
1858
1859 ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
1860 switch (ifindex) {
1861 case NETCONFA_IFINDEX_ALL:
1862 devconf = net->ipv4.devconf_all;
1863 break;
1864 case NETCONFA_IFINDEX_DEFAULT:
1865 devconf = net->ipv4.devconf_dflt;
1866 break;
1867 default:
1868 dev = __dev_get_by_index(net, ifindex);
51456b29 1869 if (!dev)
9e551110
ND		 1870 goto errout;
1871 in_dev = __in_dev_get_rtnl(dev);
51456b29 1872 if (!in_dev)
9e551110
ND		 1873 goto errout;
1874 devconf = &in_dev->cnf;
1875 break;
1876 }
1877
1878 err = -ENOBUFS;
1879 skb = nlmsg_new(inet_netconf_msgsize_devconf(-1), GFP_ATOMIC);
51456b29 1880 if (!skb)
9e551110
ND		 1881 goto errout;
1882
1883 err = inet_netconf_fill_devconf(skb, ifindex, devconf,
1884 NETLINK_CB(in_skb).portid,
1885 nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
1886 -1);
1887 if (err < 0) {
1888 /* -EMSGSIZE implies BUG in inet_netconf_msgsize_devconf() */
1889 WARN_ON(err == -EMSGSIZE);
1890 kfree_skb(skb);
1891 goto errout;
1892 }
1893 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
1894errout:
1895 return err;
1896}
1897
7a674200
ND		 1898static int inet_netconf_dump_devconf(struct sk_buff *skb,
1899 struct netlink_callback *cb)
1900{
1901 struct net *net = sock_net(skb->sk);
1902 int h, s_h;
1903 int idx, s_idx;
1904 struct net_device *dev;
1905 struct in_device *in_dev;
1906 struct hlist_head *head;
1907
1908 s_h = cb->args[0];
1909 s_idx = idx = cb->args[1];
1910
1911 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1912 idx = 0;
1913 head = &net->dev_index_head[h];
1914 rcu_read_lock();
0465277f
ND		 1915 cb->seq = atomic_read(&net->ipv4.dev_addr_genid) ^
1916 net->dev_base_seq;
7a674200
ND		 1917 hlist_for_each_entry_rcu(dev, head, index_hlist) {
1918 if (idx < s_idx)
1919 goto cont;
1920 in_dev = __in_dev_get_rcu(dev);
1921 if (!in_dev)
1922 goto cont;
1923
1924 if (inet_netconf_fill_devconf(skb, dev->ifindex,
1925 &in_dev->cnf,
1926 NETLINK_CB(cb->skb).portid,
1927 cb->nlh->nlmsg_seq,
1928 RTM_NEWNETCONF,
1929 NLM_F_MULTI,
7b46a644 1930 -1) < 0) {
7a674200
ND		 1931 rcu_read_unlock();
1932 goto done;
1933 }
0465277f 1934 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
7a674200
ND		 1935cont:
1936 idx++;
1937 }
1938 rcu_read_unlock();
1939 }
1940 if (h == NETDEV_HASHENTRIES) {
1941 if (inet_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
1942 net->ipv4.devconf_all,
1943 NETLINK_CB(cb->skb).portid,
1944 cb->nlh->nlmsg_seq,
1945 RTM_NEWNETCONF, NLM_F_MULTI,
7b46a644 1946 -1) < 0)
7a674200
ND		 1947 goto done;
1948 else
1949 h++;
1950 }
1951 if (h == NETDEV_HASHENTRIES + 1) {
1952 if (inet_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
1953 net->ipv4.devconf_dflt,
1954 NETLINK_CB(cb->skb).portid,
1955 cb->nlh->nlmsg_seq,
1956 RTM_NEWNETCONF, NLM_F_MULTI,
7b46a644 1957 -1) < 0)
7a674200
ND		 1958 goto done;
1959 else
1960 h++;
1961 }
1962done:
1963 cb->args[0] = h;
1964 cb->args[1] = idx;
1965
1966 return skb->len;
1967}
1968
1da177e4
LT		 1969#ifdef CONFIG_SYSCTL
1970
c0ce9fb3 1971static void devinet_copy_dflt_conf(struct net *net, int i)
31be3085
HX		 1972{
1973 struct net_device *dev;
1974
c6d14c84
ED		 1975 rcu_read_lock();
1976 for_each_netdev_rcu(net, dev) {
31be3085 1977 struct in_device *in_dev;
c6d14c84 1978
31be3085
HX		 1979 in_dev = __in_dev_get_rcu(dev);
1980 if (in_dev && !test_bit(i, in_dev->cnf.state))
9355bbd6 1981 in_dev->cnf.data[i] = net->ipv4.devconf_dflt->data[i];
31be3085 1982 }
c6d14c84 1983 rcu_read_unlock();
31be3085
HX		 1984}
1985
c6d14c84 1986/* called with RTNL locked */
c0ce9fb3 1987static void inet_forward_change(struct net *net)
68dd299b
PE		 1988{
1989 struct net_device *dev;
586f1211 1990 int on = IPV4_DEVCONF_ALL(net, FORWARDING);
68dd299b 1991
586f1211 1992 IPV4_DEVCONF_ALL(net, ACCEPT_REDIRECTS) = !on;
9355bbd6 1993 IPV4_DEVCONF_DFLT(net, FORWARDING) = on;
edc9e748
ND		 1994 inet_netconf_notify_devconf(net, NETCONFA_FORWARDING,
1995 NETCONFA_IFINDEX_ALL,
1996 net->ipv4.devconf_all);
1997 inet_netconf_notify_devconf(net, NETCONFA_FORWARDING,
1998 NETCONFA_IFINDEX_DEFAULT,
1999 net->ipv4.devconf_dflt);
68dd299b 2000
c0ce9fb3 2001 for_each_netdev(net, dev) {
68dd299b 2002 struct in_device *in_dev;
0187bdfb
BH		 2003 if (on)
2004 dev_disable_lro(dev);
68dd299b
PE		 2005 rcu_read_lock();
2006 in_dev = __in_dev_get_rcu(dev);
edc9e748 2007 if (in_dev) {
68dd299b 2008 IN_DEV_CONF_SET(in_dev, FORWARDING, on);
edc9e748
ND		 2009 inet_netconf_notify_devconf(net, NETCONFA_FORWARDING,
2010 dev->ifindex, &in_dev->cnf);
2011 }
68dd299b
PE		 2012 rcu_read_unlock();
2013 }
68dd299b
PE		 2014}
2015
f085ff1c 2016static int devinet_conf_ifindex(struct net *net, struct ipv4_devconf *cnf)
2017{
2018 if (cnf == net->ipv4.devconf_dflt)
2019 return NETCONFA_IFINDEX_DEFAULT;
2020 else if (cnf == net->ipv4.devconf_all)
2021 return NETCONFA_IFINDEX_ALL;
2022 else {
2023 struct in_device *idev
2024 = container_of(cnf, struct in_device, cnf);
2025 return idev->dev->ifindex;
2026 }
2027}
2028
fe2c6338 2029static int devinet_conf_proc(struct ctl_table *ctl, int write,
8d65af78 2030 void __user *buffer,
31be3085
HX		 2031 size_t *lenp, loff_t *ppos)
2032{
d01ff0a0 2033 int old_value = *(int *)ctl->data;
8d65af78 2034 int ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
d01ff0a0 2035 int new_value = *(int *)ctl->data;
31be3085
HX		 2036
2037 if (write) {
2038 struct ipv4_devconf *cnf = ctl->extra1;
c0ce9fb3 2039 struct net *net = ctl->extra2;
31be3085 2040 int i = (int *)ctl->data - cnf->data;
f085ff1c 2041 int ifindex;
31be3085
HX		 2042
2043 set_bit(i, cnf->state);
2044
9355bbd6 2045 if (cnf == net->ipv4.devconf_dflt)
c0ce9fb3 2046 devinet_copy_dflt_conf(net, i);
d0daebc3
TG		 2047 if (i == IPV4_DEVCONF_ACCEPT_LOCAL - 1 ||
2048 i == IPV4_DEVCONF_ROUTE_LOCALNET - 1)
d01ff0a0 2049 if ((new_value == 0) && (old_value != 0))
4ccfe6d4 2050 rt_cache_flush(net);
f085ff1c 2051
cc535dfb
ND		 2052 if (i == IPV4_DEVCONF_RP_FILTER - 1 &&
2053 new_value != old_value) {
f085ff1c 2054 ifindex = devinet_conf_ifindex(net, cnf);
cc535dfb
ND		 2055 inet_netconf_notify_devconf(net, NETCONFA_RP_FILTER,
2056 ifindex, cnf);
2057 }
f085ff1c 2058 if (i == IPV4_DEVCONF_PROXY_ARP - 1 &&
2059 new_value != old_value) {
2060 ifindex = devinet_conf_ifindex(net, cnf);
09aea5df 2061 inet_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH,
f085ff1c 2062 ifindex, cnf);
2063 }
974d7af5
AG		 2064 if (i == IPV4_DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN - 1 &&
2065 new_value != old_value) {
2066 ifindex = devinet_conf_ifindex(net, cnf);
2067 inet_netconf_notify_devconf(net, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
2068 ifindex, cnf);
2069 }
31be3085
HX		 2070 }
2071
2072 return ret;
2073}
2074
fe2c6338 2075static int devinet_sysctl_forward(struct ctl_table *ctl, int write,
8d65af78 2076 void __user *buffer,
1da177e4
LT		 2077 size_t *lenp, loff_t *ppos)
2078{
2079 int *valp = ctl->data;
2080 int val = *valp;
88af182e 2081 loff_t pos = *ppos;
8d65af78 2082 int ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
1da177e4
LT		 2083
2084 if (write && *valp != val) {
c0ce9fb3
PE		 2085 struct net *net = ctl->extra2;
2086
0187bdfb 2087 if (valp != &IPV4_DEVCONF_DFLT(net, FORWARDING)) {
88af182e
EB		 2088 if (!rtnl_trylock()) {
2089 /* Restore the original values before restarting */
2090 *valp = val;
2091 *ppos = pos;
9b8adb5e 2092 return restart_syscall();
88af182e 2093 }
0187bdfb
BH		 2094 if (valp == &IPV4_DEVCONF_ALL(net, FORWARDING)) {
2095 inet_forward_change(net);
edc9e748 2096 } else {
0187bdfb
BH		 2097 struct ipv4_devconf *cnf = ctl->extra1;
2098 struct in_device *idev =
2099 container_of(cnf, struct in_device, cnf);
edc9e748
ND		 2100 if (*valp)
2101 dev_disable_lro(idev->dev);
2102 inet_netconf_notify_devconf(net,
2103 NETCONFA_FORWARDING,
2104 idev->dev->ifindex,
2105 cnf);
0187bdfb
BH		 2106 }
2107 rtnl_unlock();
4ccfe6d4 2108 rt_cache_flush(net);
edc9e748
ND		 2109 } else
2110 inet_netconf_notify_devconf(net, NETCONFA_FORWARDING,
2111 NETCONFA_IFINDEX_DEFAULT,
2112 net->ipv4.devconf_dflt);
1da177e4
LT		 2113 }
2114
2115 return ret;
2116}
2117
fe2c6338 2118static int ipv4_doint_and_flush(struct ctl_table *ctl, int write,
323e126f
DM		 2119 void __user *buffer,
2120 size_t *lenp, loff_t *ppos)
1da177e4
LT		 2121{
2122 int *valp = ctl->data;
2123 int val = *valp;
8d65af78 2124 int ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
76e6ebfb 2125 struct net *net = ctl->extra2;
1da177e4
LT		 2126
2127 if (write && *valp != val)
4ccfe6d4 2128 rt_cache_flush(net);
1da177e4
LT		 2129
2130 return ret;
2131}
2132
f8572d8f 2133#define DEVINET_SYSCTL_ENTRY(attr, name, mval, proc) \
42f811b8 2134 { \
42f811b8
HX		 2135 .procname = name, \
2136 .data = ipv4_devconf.data + \
02291680 2137 IPV4_DEVCONF_ ## attr - 1, \
42f811b8
HX		 2138 .maxlen = sizeof(int), \
2139 .mode = mval, \
2140 .proc_handler = proc, \
31be3085 2141 .extra1 = &ipv4_devconf, \
42f811b8
HX		 2142 }
2143
2144#define DEVINET_SYSCTL_RW_ENTRY(attr, name) \
f8572d8f 2145 DEVINET_SYSCTL_ENTRY(attr, name, 0644, devinet_conf_proc)
42f811b8
HX		 2146
2147#define DEVINET_SYSCTL_RO_ENTRY(attr, name) \
f8572d8f 2148 DEVINET_SYSCTL_ENTRY(attr, name, 0444, devinet_conf_proc)
42f811b8 2149
f8572d8f
EB		 2150#define DEVINET_SYSCTL_COMPLEX_ENTRY(attr, name, proc) \
2151 DEVINET_SYSCTL_ENTRY(attr, name, 0644, proc)
42f811b8
HX		 2152
2153#define DEVINET_SYSCTL_FLUSHING_ENTRY(attr, name) \
f8572d8f 2154 DEVINET_SYSCTL_COMPLEX_ENTRY(attr, name, ipv4_doint_and_flush)
42f811b8 2155
1da177e4
LT		 2156static struct devinet_sysctl_table {
2157 struct ctl_table_header *sysctl_header;
02291680 2158 struct ctl_table devinet_vars[__IPV4_DEVCONF_MAX];
1da177e4
LT		 2159} devinet_sysctl = {
2160 .devinet_vars = {
42f811b8 2161 DEVINET_SYSCTL_COMPLEX_ENTRY(FORWARDING, "forwarding",
f8572d8f 2162 devinet_sysctl_forward),
42f811b8
HX		 2163 DEVINET_SYSCTL_RO_ENTRY(MC_FORWARDING, "mc_forwarding"),
2164
2165 DEVINET_SYSCTL_RW_ENTRY(ACCEPT_REDIRECTS, "accept_redirects"),
2166 DEVINET_SYSCTL_RW_ENTRY(SECURE_REDIRECTS, "secure_redirects"),
2167 DEVINET_SYSCTL_RW_ENTRY(SHARED_MEDIA, "shared_media"),
2168 DEVINET_SYSCTL_RW_ENTRY(RP_FILTER, "rp_filter"),
2169 DEVINET_SYSCTL_RW_ENTRY(SEND_REDIRECTS, "send_redirects"),
2170 DEVINET_SYSCTL_RW_ENTRY(ACCEPT_SOURCE_ROUTE,
2171 "accept_source_route"),
8153a10c 2172 DEVINET_SYSCTL_RW_ENTRY(ACCEPT_LOCAL, "accept_local"),
28f6aeea 2173 DEVINET_SYSCTL_RW_ENTRY(SRC_VMARK, "src_valid_mark"),
42f811b8
HX		 2174 DEVINET_SYSCTL_RW_ENTRY(PROXY_ARP, "proxy_arp"),
2175 DEVINET_SYSCTL_RW_ENTRY(MEDIUM_ID, "medium_id"),
2176 DEVINET_SYSCTL_RW_ENTRY(BOOTP_RELAY, "bootp_relay"),
2177 DEVINET_SYSCTL_RW_ENTRY(LOG_MARTIANS, "log_martians"),
2178 DEVINET_SYSCTL_RW_ENTRY(TAG, "tag"),
2179 DEVINET_SYSCTL_RW_ENTRY(ARPFILTER, "arp_filter"),
2180 DEVINET_SYSCTL_RW_ENTRY(ARP_ANNOUNCE, "arp_announce"),
2181 DEVINET_SYSCTL_RW_ENTRY(ARP_IGNORE, "arp_ignore"),
2182 DEVINET_SYSCTL_RW_ENTRY(ARP_ACCEPT, "arp_accept"),
eefef1cf 2183 DEVINET_SYSCTL_RW_ENTRY(ARP_NOTIFY, "arp_notify"),
65324144 2184 DEVINET_SYSCTL_RW_ENTRY(PROXY_ARP_PVLAN, "proxy_arp_pvlan"),
5c6fe01c
WM		 2185 DEVINET_SYSCTL_RW_ENTRY(FORCE_IGMP_VERSION,
2186 "force_igmp_version"),
2690048c
WM		 2187 DEVINET_SYSCTL_RW_ENTRY(IGMPV2_UNSOLICITED_REPORT_INTERVAL,
2188 "igmpv2_unsolicited_report_interval"),
2189 DEVINET_SYSCTL_RW_ENTRY(IGMPV3_UNSOLICITED_REPORT_INTERVAL,
2190 "igmpv3_unsolicited_report_interval"),
0eeb075f
AG		 2191 DEVINET_SYSCTL_RW_ENTRY(IGNORE_ROUTES_WITH_LINKDOWN,
2192 "ignore_routes_with_linkdown"),
97daf331
JB		 2193 DEVINET_SYSCTL_RW_ENTRY(DROP_GRATUITOUS_ARP,
2194 "drop_gratuitous_arp"),
42f811b8
HX		 2195
2196 DEVINET_SYSCTL_FLUSHING_ENTRY(NOXFRM, "disable_xfrm"),
2197 DEVINET_SYSCTL_FLUSHING_ENTRY(NOPOLICY, "disable_policy"),
42f811b8
HX		 2198 DEVINET_SYSCTL_FLUSHING_ENTRY(PROMOTE_SECONDARIES,
2199 "promote_secondaries"),
d0daebc3
TG		 2200 DEVINET_SYSCTL_FLUSHING_ENTRY(ROUTE_LOCALNET,
2201 "route_localnet"),
12b74dfa
JB		 2202 DEVINET_SYSCTL_FLUSHING_ENTRY(DROP_UNICAST_IN_L2_MULTICAST,
2203 "drop_unicast_in_l2_multicast"),
1da177e4 2204 },
1da177e4
LT		 2205};
2206
ea40b324 2207static int __devinet_sysctl_register(struct net *net, char *dev_name,
f8572d8f 2208 struct ipv4_devconf *p)
1da177e4
LT		 2209{
2210 int i;
9fa89642 2211 struct devinet_sysctl_table *t;
8607ddb8 2212 char path[sizeof("net/ipv4/conf/") + IFNAMSIZ];
bfada697 2213
9fa89642 2214 t = kmemdup(&devinet_sysctl, sizeof(*t), GFP_KERNEL);
1da177e4 2215 if (!t)
9fa89642
PE		 2216 goto out;
2217
1da177e4
LT		 2218 for (i = 0; i < ARRAY_SIZE(t->devinet_vars) - 1; i++) {
2219 t->devinet_vars[i].data += (char *)p - (char *)&ipv4_devconf;
31be3085 2220 t->devinet_vars[i].extra1 = p;
c0ce9fb3 2221 t->devinet_vars[i].extra2 = net;
1da177e4
LT		 2222 }
2223
8607ddb8 2224 snprintf(path, sizeof(path), "net/ipv4/conf/%s", dev_name);
1da177e4 2225
8607ddb8 2226 t->sysctl_header = register_net_sysctl(net, path, t->devinet_vars);
1da177e4 2227 if (!t->sysctl_header)
8607ddb8 2228 goto free;
1da177e4
LT		 2229
2230 p->sysctl = t;
ea40b324 2231 return 0;
1da177e4 2232
9fa89642 2233free:
1da177e4 2234 kfree(t);
9fa89642 2235out:
ea40b324 2236 return -ENOBUFS;
1da177e4
LT		 2237}
2238
51602b2a
PE		 2239static void __devinet_sysctl_unregister(struct ipv4_devconf *cnf)
2240{
2241 struct devinet_sysctl_table *t = cnf->sysctl;
2242
51456b29 2243 if (!t)
51602b2a
PE		 2244 return;
2245
2246 cnf->sysctl = NULL;
ff538818 2247 unregister_net_sysctl_table(t->sysctl_header);
51602b2a
PE		 2248 kfree(t);
2249}
2250
20e61da7 2251static int devinet_sysctl_register(struct in_device *idev)
66f27a52 2252{
20e61da7
WC		 2253 int err;
2254
2255 if (!sysctl_dev_name_is_allowed(idev->dev->name))
2256 return -EINVAL;
2257
2258 err = neigh_sysctl_register(idev->dev, idev->arp_parms, NULL);
2259 if (err)
2260 return err;
2261 err = __devinet_sysctl_register(dev_net(idev->dev), idev->dev->name,
f8572d8f 2262 &idev->cnf);
20e61da7
WC		 2263 if (err)
2264 neigh_sysctl_unregister(idev->arp_parms);
2265 return err;
66f27a52
PE		 2266}
2267
51602b2a 2268static void devinet_sysctl_unregister(struct in_device *idev)
1da177e4 2269{
51602b2a
PE		 2270 __devinet_sysctl_unregister(&idev->cnf);
2271 neigh_sysctl_unregister(idev->arp_parms);
1da177e4 2272}
1da177e4 2273
68dd299b
PE		 2274static struct ctl_table ctl_forward_entry[] = {
2275 {
68dd299b
PE		 2276 .procname = "ip_forward",
2277 .data = &ipv4_devconf.data[
02291680 2278 IPV4_DEVCONF_FORWARDING - 1],
68dd299b
PE		 2279 .maxlen = sizeof(int),
2280 .mode = 0644,
2281 .proc_handler = devinet_sysctl_forward,
68dd299b 2282 .extra1 = &ipv4_devconf,
c0ce9fb3 2283 .extra2 = &init_net,
68dd299b
PE		 2284 },
2285 { },
2286};
2a75de0c 2287#endif
68dd299b 2288
752d14dc
PE		 2289static __net_init int devinet_init_net(struct net *net)
2290{
2291 int err;
752d14dc 2292 struct ipv4_devconf *all, *dflt;
2a75de0c
ED		 2293#ifdef CONFIG_SYSCTL
2294 struct ctl_table *tbl = ctl_forward_entry;
752d14dc 2295 struct ctl_table_header *forw_hdr;
2a75de0c 2296#endif
752d14dc
PE		 2297
2298 err = -ENOMEM;
2299 all = &ipv4_devconf;
2300 dflt = &ipv4_devconf_dflt;
752d14dc 2301
09ad9bc7 2302 if (!net_eq(net, &init_net)) {
752d14dc 2303 all = kmemdup(all, sizeof(ipv4_devconf), GFP_KERNEL);
51456b29 2304 if (!all)
752d14dc
PE		 2305 goto err_alloc_all;
2306
2307 dflt = kmemdup(dflt, sizeof(ipv4_devconf_dflt), GFP_KERNEL);
51456b29 2308 if (!dflt)
752d14dc
PE		 2309 goto err_alloc_dflt;
2310
2a75de0c 2311#ifdef CONFIG_SYSCTL
752d14dc 2312 tbl = kmemdup(tbl, sizeof(ctl_forward_entry), GFP_KERNEL);
51456b29 2313 if (!tbl)
752d14dc
PE		 2314 goto err_alloc_ctl;
2315
02291680 2316 tbl[0].data = &all->data[IPV4_DEVCONF_FORWARDING - 1];
752d14dc
PE		 2317 tbl[0].extra1 = all;
2318 tbl[0].extra2 = net;
2a75de0c 2319#endif
752d14dc
PE		 2320 }
2321
2322#ifdef CONFIG_SYSCTL
f8572d8f 2323 err = __devinet_sysctl_register(net, "all", all);
752d14dc
PE		 2324 if (err < 0)
2325 goto err_reg_all;
2326
f8572d8f 2327 err = __devinet_sysctl_register(net, "default", dflt);
752d14dc
PE		 2328 if (err < 0)
2329 goto err_reg_dflt;
2330
2331 err = -ENOMEM;
8607ddb8 2332 forw_hdr = register_net_sysctl(net, "net/ipv4", tbl);
51456b29 2333 if (!forw_hdr)
752d14dc 2334 goto err_reg_ctl;
2a75de0c 2335 net->ipv4.forw_hdr = forw_hdr;
752d14dc
PE		 2336#endif
2337
752d14dc
PE		 2338 net->ipv4.devconf_all = all;
2339 net->ipv4.devconf_dflt = dflt;
2340 return 0;
2341
2342#ifdef CONFIG_SYSCTL
2343err_reg_ctl:
2344 __devinet_sysctl_unregister(dflt);
2345err_reg_dflt:
2346 __devinet_sysctl_unregister(all);
2347err_reg_all:
2348 if (tbl != ctl_forward_entry)
2349 kfree(tbl);
752d14dc 2350err_alloc_ctl:
2a75de0c 2351#endif
752d14dc
PE		 2352 if (dflt != &ipv4_devconf_dflt)
2353 kfree(dflt);
2354err_alloc_dflt:
2355 if (all != &ipv4_devconf)
2356 kfree(all);
2357err_alloc_all:
2358 return err;
2359}
2360
2361static __net_exit void devinet_exit_net(struct net *net)
2362{
2a75de0c 2363#ifdef CONFIG_SYSCTL
752d14dc
PE		 2364 struct ctl_table *tbl;
2365
2366 tbl = net->ipv4.forw_hdr->ctl_table_arg;
752d14dc
PE		 2367 unregister_net_sysctl_table(net->ipv4.forw_hdr);
2368 __devinet_sysctl_unregister(net->ipv4.devconf_dflt);
2369 __devinet_sysctl_unregister(net->ipv4.devconf_all);
752d14dc 2370 kfree(tbl);
2a75de0c 2371#endif
752d14dc
PE		 2372 kfree(net->ipv4.devconf_dflt);
2373 kfree(net->ipv4.devconf_all);
2374}
2375
2376static __net_initdata struct pernet_operations devinet_ops = {
2377 .init = devinet_init_net,
2378 .exit = devinet_exit_net,
2379};
2380
207895fd 2381static struct rtnl_af_ops inet_af_ops __read_mostly = {
9f0f7272
TG		 2382 .family = AF_INET,
2383 .fill_link_af = inet_fill_link_af,
2384 .get_link_af_size = inet_get_link_af_size,
cf7afbfe
TG		 2385 .validate_link_af = inet_validate_link_af,
2386 .set_link_af = inet_set_link_af,
9f0f7272
TG		 2387};
2388
1da177e4
LT		 2389void __init devinet_init(void)
2390{
fd23c3b3
DM		 2391 int i;
2392
2393 for (i = 0; i < IN4_ADDR_HSIZE; i++)
2394 INIT_HLIST_HEAD(&inet_addr_lst[i]);
2395
752d14dc
PE		 2396 register_pernet_subsys(&devinet_ops);
2397
1da177e4
LT		 2398 register_gifconf(PF_INET, inet_gifconf);
2399 register_netdevice_notifier(&ip_netdev_notifier);
63f3444f 2400
906e073f 2401 queue_delayed_work(system_power_efficient_wq, &check_lifetime_work, 0);
5c766d64 2402
9f0f7272
TG		 2403 rtnl_af_register(&inet_af_ops);
2404
c7ac8679
GR		 2405 rtnl_register(PF_INET, RTM_NEWADDR, inet_rtm_newaddr, NULL, NULL);
2406 rtnl_register(PF_INET, RTM_DELADDR, inet_rtm_deladdr, NULL, NULL);
2407 rtnl_register(PF_INET, RTM_GETADDR, NULL, inet_dump_ifaddr, NULL);
9e551110 2408 rtnl_register(PF_INET, RTM_GETNETCONF, inet_netconf_get_devconf,
7a674200 2409 inet_netconf_dump_devconf, NULL);
1da177e4 2410}
Linux 6.x block layer and io_uring tree(s)