net/ipv4/devinet.c

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