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 <asm/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/string.h>
  36 #include <linux/mm.h>
  37 #include <linux/socket.h>
  38 #include <linux/sockios.h>
  39 #include <linux/in.h>
  40 #include <linux/errno.h>
  41 #include <linux/interrupt.h>
  42 #include <linux/if_addr.h>
  43 #include <linux/if_ether.h>
  44 #include <linux/inet.h>
  45 #include <linux/netdevice.h>
  46 #include <linux/etherdevice.h>
  47 #include <linux/skbuff.h>
  48 #include <linux/init.h>
  49 #include <linux/notifier.h>
  50 #include <linux/inetdevice.h>
  51 #include <linux/igmp.h>
  52 #include <linux/slab.h>
  53 #include <linux/hash.h>
  54 #ifdef CONFIG_SYSCTL
  55 #include <linux/sysctl.h>
  56 #endif
  57 #include <linux/kmod.h>
  58 #include <linux/netconf.h>
  59
  60 #include <net/arp.h>
  61 #include <net/ip.h>
  62 #include <net/route.h>
  63 #include <net/ip_fib.h>
  64 #include <net/rtnetlink.h>
  65 #include <net/net_namespace.h>
  66 #include <net/addrconf.h>
  67
  68 #include "fib_lookup.h"
  69
  70 static struct ipv4_devconf ipv4_devconf = {
  71         .data = {
  72                 [IPV4_DEVCONF_ACCEPT_REDIRECTS - 1] = 1,
  73                 [IPV4_DEVCONF_SEND_REDIRECTS - 1] = 1,
  74                 [IPV4_DEVCONF_SECURE_REDIRECTS - 1] = 1,
  75                 [IPV4_DEVCONF_SHARED_MEDIA - 1] = 1,
  76                 [IPV4_DEVCONF_IGMPV2_UNSOLICITED_REPORT_INTERVAL - 1] = 10000 /*ms*/,
  77                 [IPV4_DEVCONF_IGMPV3_UNSOLICITED_REPORT_INTERVAL - 1] =  1000 /*ms*/,
  78         },
  79 };
  80
  81 static struct ipv4_devconf ipv4_devconf_dflt = {
  82         .data = {
  83                 [IPV4_DEVCONF_ACCEPT_REDIRECTS - 1] = 1,
  84                 [IPV4_DEVCONF_SEND_REDIRECTS - 1] = 1,
  85                 [IPV4_DEVCONF_SECURE_REDIRECTS - 1] = 1,
  86                 [IPV4_DEVCONF_SHARED_MEDIA - 1] = 1,
  87                 [IPV4_DEVCONF_ACCEPT_SOURCE_ROUTE - 1] = 1,
  88                 [IPV4_DEVCONF_IGMPV2_UNSOLICITED_REPORT_INTERVAL - 1] = 10000 /*ms*/,
  89                 [IPV4_DEVCONF_IGMPV3_UNSOLICITED_REPORT_INTERVAL - 1] =  1000 /*ms*/,
  90         },
  91 };
  92
  93 #define IPV4_DEVCONF_DFLT(net, attr) \
  94         IPV4_DEVCONF((*net->ipv4.devconf_dflt), attr)
  95
  96 static const struct nla_policy ifa_ipv4_policy[IFA_MAX+1] = {
  97         [IFA_LOCAL]             = { .type = NLA_U32 },
  98         [IFA_ADDRESS]           = { .type = NLA_U32 },
  99         [IFA_BROADCAST]         = { .type = NLA_U32 },
 100         [IFA_LABEL]             = { .type = NLA_STRING, .len = IFNAMSIZ - 1 },
 101         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
 102         [IFA_FLAGS]             = { .type = NLA_U32 },
 103 };
 104
 105 #define IN4_ADDR_HSIZE_SHIFT    8
 106 #define IN4_ADDR_HSIZE          (1U << IN4_ADDR_HSIZE_SHIFT)
 107
 108 static struct hlist_head inet_addr_lst[IN4_ADDR_HSIZE];
 109 static DEFINE_SPINLOCK(inet_addr_hash_lock);
 110
 111 static u32 inet_addr_hash(struct net *net, __be32 addr)
 112 {
 113         u32 val = (__force u32) addr ^ net_hash_mix(net);
 114
 115         return hash_32(val, IN4_ADDR_HSIZE_SHIFT);
 116 }
 117
 118 static void inet_hash_insert(struct net *net, struct in_ifaddr *ifa)
 119 {
 120         u32 hash = inet_addr_hash(net, ifa->ifa_local);
 121
 122         spin_lock(&inet_addr_hash_lock);
 123         hlist_add_head_rcu(&ifa->hash, &inet_addr_lst[hash]);
 124         spin_unlock(&inet_addr_hash_lock);
 125 }
 126
 127 static void inet_hash_remove(struct in_ifaddr *ifa)
 128 {
 129         spin_lock(&inet_addr_hash_lock);
 130         hlist_del_init_rcu(&ifa->hash);
 131         spin_unlock(&inet_addr_hash_lock);
 132 }
 133
 134 /**
 135  * __ip_dev_find - find the first device with a given source address.
 136  * @net: the net namespace
 137  * @addr: the source address
 138  * @devref: if true, take a reference on the found device
 139  *
 140  * If a caller uses devref=false, it should be protected by RCU, or RTNL
 141  */
 142 struct net_device *__ip_dev_find(struct net *net, __be32 addr, bool devref)
 143 {
 144         u32 hash = inet_addr_hash(net, addr);
 145         struct net_device *result = NULL;
 146         struct in_ifaddr *ifa;
 147
 148         rcu_read_lock();
 149         hlist_for_each_entry_rcu(ifa, &inet_addr_lst[hash], hash) {
 150                 if (ifa->ifa_local == addr) {
 151                         struct net_device *dev = ifa->ifa_dev->dev;
 152
 153                         if (!net_eq(dev_net(dev), net))
 154                                 continue;
 155                         result = dev;
 156                         break;
 157                 }
 158         }
 159         if (!result) {
 160                 struct flowi4 fl4 = { .daddr = addr };
 161                 struct fib_result res = { 0 };
 162                 struct fib_table *local;
 163
 164                 /* Fallback to FIB local table so that communication
 165                  * over loopback subnets work.
 166                  */
 167                 local = fib_get_table(net, RT_TABLE_LOCAL);
 168                 if (local &&
 169                     !fib_table_lookup(local, &fl4, &res, FIB_LOOKUP_NOREF) &&
 170                     res.type == RTN_LOCAL)
 171                         result = FIB_RES_DEV(res);
 172         }
 173         if (result && devref)
 174                 dev_hold(result);
 175         rcu_read_unlock();
 176         return result;
 177 }
 178 EXPORT_SYMBOL(__ip_dev_find);
 179
 180 static void rtmsg_ifa(int event, struct in_ifaddr *, struct nlmsghdr *, u32);
 181
 182 static BLOCKING_NOTIFIER_HEAD(inetaddr_chain);
 183 static void inet_del_ifa(struct in_device *in_dev, struct in_ifaddr **ifap,
 184                          int destroy);
 185 #ifdef CONFIG_SYSCTL
 186 static void devinet_sysctl_register(struct in_device *idev);
 187 static void devinet_sysctl_unregister(struct in_device *idev);
 188 #else
 189 static void devinet_sysctl_register(struct in_device *idev)
 190 {
 191 }
 192 static void devinet_sysctl_unregister(struct in_device *idev)
 193 {
 194 }
 195 #endif
 196
 197 /* Locks all the inet devices. */
 198
 199 static struct in_ifaddr *inet_alloc_ifa(void)
 200 {
 201         return kzalloc(sizeof(struct in_ifaddr), GFP_KERNEL);
 202 }
 203
 204 static void inet_rcu_free_ifa(struct rcu_head *head)
 205 {
 206         struct in_ifaddr *ifa = container_of(head, struct in_ifaddr, rcu_head);
 207         if (ifa->ifa_dev)
 208                 in_dev_put(ifa->ifa_dev);
 209         kfree(ifa);
 210 }
 211
 212 static void inet_free_ifa(struct in_ifaddr *ifa)
 213 {
 214         call_rcu(&ifa->rcu_head, inet_rcu_free_ifa);
 215 }
 216
 217 void in_dev_finish_destroy(struct in_device *idev)
 218 {
 219         struct net_device *dev = idev->dev;
 220
 221         WARN_ON(idev->ifa_list);
 222         WARN_ON(idev->mc_list);
 223         kfree(rcu_dereference_protected(idev->mc_hash, 1));
 224 #ifdef NET_REFCNT_DEBUG
 225         pr_debug("%s: %p=%s\n", __func__, idev, dev ? dev->name : "NIL");
 226 #endif
 227         dev_put(dev);
 228         if (!idev->dead)
 229                 pr_err("Freeing alive in_device %p\n", idev);
 230         else
 231                 kfree(idev);
 232 }
 233 EXPORT_SYMBOL(in_dev_finish_destroy);
 234
 235 static struct in_device *inetdev_init(struct net_device *dev)
 236 {
 237         struct in_device *in_dev;
 238
 239         ASSERT_RTNL();
 240
 241         in_dev = kzalloc(sizeof(*in_dev), GFP_KERNEL);
 242         if (!in_dev)
 243                 goto out;
 244         memcpy(&in_dev->cnf, dev_net(dev)->ipv4.devconf_dflt,
 245                         sizeof(in_dev->cnf));
 246         in_dev->cnf.sysctl = NULL;
 247         in_dev->dev = dev;
 248         in_dev->arp_parms = neigh_parms_alloc(dev, &arp_tbl);
 249         if (!in_dev->arp_parms)
 250                 goto out_kfree;
 251         if (IPV4_DEVCONF(in_dev->cnf, FORWARDING))
 252                 dev_disable_lro(dev);
 253         /* Reference in_dev->dev */
 254         dev_hold(dev);
 255         /* Account for reference dev->ip_ptr (below) */
 256         in_dev_hold(in_dev);
 257
 258         devinet_sysctl_register(in_dev);
 259         ip_mc_init_dev(in_dev);
 260         if (dev->flags & IFF_UP)
 261                 ip_mc_up(in_dev);
 262
 263         /* we can receive as soon as ip_ptr is set -- do this last */
 264         rcu_assign_pointer(dev->ip_ptr, in_dev);
 265 out:
 266         return in_dev;
 267 out_kfree:
 268         kfree(in_dev);
 269         in_dev = NULL;
 270         goto out;
 271 }
 272
 273 static void in_dev_rcu_put(struct rcu_head *head)
 274 {
 275         struct in_device *idev = container_of(head, struct in_device, rcu_head);
 276         in_dev_put(idev);
 277 }
 278
 279 static void inetdev_destroy(struct in_device *in_dev)
 280 {
 281         struct in_ifaddr *ifa;
 282         struct net_device *dev;
 283
 284         ASSERT_RTNL();
 285
 286         dev = in_dev->dev;
 287
 288         in_dev->dead = 1;
 289
 290         ip_mc_destroy_dev(in_dev);
 291
 292         while ((ifa = in_dev->ifa_list) != NULL) {
 293                 inet_del_ifa(in_dev, &in_dev->ifa_list, 0);
 294                 inet_free_ifa(ifa);
 295         }
 296
 297         RCU_INIT_POINTER(dev->ip_ptr, NULL);
 298
 299         devinet_sysctl_unregister(in_dev);
 300         neigh_parms_release(&arp_tbl, in_dev->arp_parms);
 301         arp_ifdown(dev);
 302
 303         call_rcu(&in_dev->rcu_head, in_dev_rcu_put);
 304 }
 305
 306 int inet_addr_onlink(struct in_device *in_dev, __be32 a, __be32 b)
 307 {
 308         rcu_read_lock();
 309         for_primary_ifa(in_dev) {
 310                 if (inet_ifa_match(a, ifa)) {
 311                         if (!b || inet_ifa_match(b, ifa)) {
 312                                 rcu_read_unlock();
 313                                 return 1;
 314                         }
 315                 }
 316         } endfor_ifa(in_dev);
 317         rcu_read_unlock();
 318         return 0;
 319 }
 320
 321 static void __inet_del_ifa(struct in_device *in_dev, struct in_ifaddr **ifap,
 322                          int destroy, struct nlmsghdr *nlh, u32 portid)
 323 {
 324         struct in_ifaddr *promote = NULL;
 325         struct in_ifaddr *ifa, *ifa1 = *ifap;
 326         struct in_ifaddr *last_prim = in_dev->ifa_list;
 327         struct in_ifaddr *prev_prom = NULL;
 328         int do_promote = IN_DEV_PROMOTE_SECONDARIES(in_dev);
 329
 330         ASSERT_RTNL();
 331
 332         /* 1. Deleting primary ifaddr forces deletion all secondaries
 333          * unless alias promotion is set
 334          **/
 335
 336         if (!(ifa1->ifa_flags & IFA_F_SECONDARY)) {
 337                 struct in_ifaddr **ifap1 = &ifa1->ifa_next;
 338
 339                 while ((ifa = *ifap1) != NULL) {
 340                         if (!(ifa->ifa_flags & IFA_F_SECONDARY) &&
 341                             ifa1->ifa_scope <= ifa->ifa_scope)
 342                                 last_prim = ifa;
 343
 344                         if (!(ifa->ifa_flags & IFA_F_SECONDARY) ||
 345                             ifa1->ifa_mask != ifa->ifa_mask ||
 346                             !inet_ifa_match(ifa1->ifa_address, ifa)) {
 347                                 ifap1 = &ifa->ifa_next;
 348                                 prev_prom = ifa;
 349                                 continue;
 350                         }
 351
 352                         if (!do_promote) {
 353                                 inet_hash_remove(ifa);
 354                                 *ifap1 = ifa->ifa_next;
 355
 356                                 rtmsg_ifa(RTM_DELADDR, ifa, nlh, portid);
 357                                 blocking_notifier_call_chain(&inetaddr_chain,
 358                                                 NETDEV_DOWN, ifa);
 359                                 inet_free_ifa(ifa);
 360                         } else {
 361                                 promote = ifa;
 362                                 break;
 363                         }
 364                 }
 365         }
 366
 367         /* On promotion all secondaries from subnet are changing
 368          * the primary IP, we must remove all their routes silently
 369          * and later to add them back with new prefsrc. Do this
 370          * while all addresses are on the device list.
 371          */
 372         for (ifa = promote; ifa; ifa = ifa->ifa_next) {
 373                 if (ifa1->ifa_mask == ifa->ifa_mask &&
 374                     inet_ifa_match(ifa1->ifa_address, ifa))
 375                         fib_del_ifaddr(ifa, ifa1);
 376         }
 377
 378         /* 2. Unlink it */
 379
 380         *ifap = ifa1->ifa_next;
 381         inet_hash_remove(ifa1);
 382
 383         /* 3. Announce address deletion */
 384
 385         /* Send message first, then call notifier.
 386            At first sight, FIB update triggered by notifier
 387            will refer to already deleted ifaddr, that could confuse
 388            netlink listeners. It is not true: look, gated sees
 389            that route deleted and if it still thinks that ifaddr
 390            is valid, it will try to restore deleted routes... Grr.
 391            So that, this order is correct.
 392          */
 393         rtmsg_ifa(RTM_DELADDR, ifa1, nlh, portid);
 394         blocking_notifier_call_chain(&inetaddr_chain, NETDEV_DOWN, ifa1);
 395
 396         if (promote) {
 397                 struct in_ifaddr *next_sec = promote->ifa_next;
 398
 399                 if (prev_prom) {
 400                         prev_prom->ifa_next = promote->ifa_next;
 401                         promote->ifa_next = last_prim->ifa_next;
 402                         last_prim->ifa_next = promote;
 403                 }
 404
 405                 promote->ifa_flags &= ~IFA_F_SECONDARY;
 406                 rtmsg_ifa(RTM_NEWADDR, promote, nlh, portid);
 407                 blocking_notifier_call_chain(&inetaddr_chain,
 408                                 NETDEV_UP, promote);
 409                 for (ifa = next_sec; ifa; ifa = ifa->ifa_next) {
 410                         if (ifa1->ifa_mask != ifa->ifa_mask ||
 411                             !inet_ifa_match(ifa1->ifa_address, ifa))
 412                                         continue;
 413                         fib_add_ifaddr(ifa);
 414                 }
 415
 416         }
 417         if (destroy)
 418                 inet_free_ifa(ifa1);
 419 }
 420
 421 static void inet_del_ifa(struct in_device *in_dev, struct in_ifaddr **ifap,
 422                          int destroy)
 423 {
 424         __inet_del_ifa(in_dev, ifap, destroy, NULL, 0);
 425 }
 426
 427 static void check_lifetime(struct work_struct *work);
 428
 429 static DECLARE_DELAYED_WORK(check_lifetime_work, check_lifetime);
 430
 431 static int __inet_insert_ifa(struct in_ifaddr *ifa, struct nlmsghdr *nlh,
 432                              u32 portid)
 433 {
 434         struct in_device *in_dev = ifa->ifa_dev;
 435         struct in_ifaddr *ifa1, **ifap, **last_primary;
 436
 437         ASSERT_RTNL();
 438
 439         if (!ifa->ifa_local) {
 440                 inet_free_ifa(ifa);
 441                 return 0;
 442         }
 443
 444         ifa->ifa_flags &= ~IFA_F_SECONDARY;
 445         last_primary = &in_dev->ifa_list;
 446
 447         for (ifap = &in_dev->ifa_list; (ifa1 = *ifap) != NULL;
 448              ifap = &ifa1->ifa_next) {
 449                 if (!(ifa1->ifa_flags & IFA_F_SECONDARY) &&
 450                     ifa->ifa_scope <= ifa1->ifa_scope)
 451                         last_primary = &ifa1->ifa_next;
 452                 if (ifa1->ifa_mask == ifa->ifa_mask &&
 453                     inet_ifa_match(ifa1->ifa_address, ifa)) {
 454                         if (ifa1->ifa_local == ifa->ifa_local) {
 455                                 inet_free_ifa(ifa);
 456                                 return -EEXIST;
 457                         }
 458                         if (ifa1->ifa_scope != ifa->ifa_scope) {
 459                                 inet_free_ifa(ifa);
 460                                 return -EINVAL;
 461                         }
 462                         ifa->ifa_flags |= IFA_F_SECONDARY;
 463                 }
 464         }
 465
 466         if (!(ifa->ifa_flags & IFA_F_SECONDARY)) {
 467                 net_srandom(ifa->ifa_local);
 468                 ifap = last_primary;
 469         }
 470
 471         ifa->ifa_next = *ifap;
 472         *ifap = ifa;
 473
 474         inet_hash_insert(dev_net(in_dev->dev), ifa);
 475
 476         cancel_delayed_work(&check_lifetime_work);
 477         schedule_delayed_work(&check_lifetime_work, 0);
 478
 479         /* Send message first, then call notifier.
 480            Notifier will trigger FIB update, so that
 481            listeners of netlink will know about new ifaddr */
 482         rtmsg_ifa(RTM_NEWADDR, ifa, nlh, portid);
 483         blocking_notifier_call_chain(&inetaddr_chain, NETDEV_UP, ifa);
 484
 485         return 0;
 486 }
 487
 488 static int inet_insert_ifa(struct in_ifaddr *ifa)
 489 {
 490         return __inet_insert_ifa(ifa, NULL, 0);
 491 }
 492
 493 static int inet_set_ifa(struct net_device *dev, struct in_ifaddr *ifa)
 494 {
 495         struct in_device *in_dev = __in_dev_get_rtnl(dev);
 496
 497         ASSERT_RTNL();
 498
 499         if (!in_dev) {
 500                 inet_free_ifa(ifa);
 501                 return -ENOBUFS;
 502         }
 503         ipv4_devconf_setall(in_dev);
 504         neigh_parms_data_state_setall(in_dev->arp_parms);
 505         if (ifa->ifa_dev != in_dev) {
 506                 WARN_ON(ifa->ifa_dev);
 507                 in_dev_hold(in_dev);
 508                 ifa->ifa_dev = in_dev;
 509         }
 510         if (ipv4_is_loopback(ifa->ifa_local))
 511                 ifa->ifa_scope = RT_SCOPE_HOST;
 512         return inet_insert_ifa(ifa);
 513 }
 514
 515 /* Caller must hold RCU or RTNL :
 516  * We dont take a reference on found in_device
 517  */
 518 struct in_device *inetdev_by_index(struct net *net, int ifindex)
 519 {
 520         struct net_device *dev;
 521         struct in_device *in_dev = NULL;
 522
 523         rcu_read_lock();
 524         dev = dev_get_by_index_rcu(net, ifindex);
 525         if (dev)
 526                 in_dev = rcu_dereference_rtnl(dev->ip_ptr);
 527         rcu_read_unlock();
 528         return in_dev;
 529 }
 530 EXPORT_SYMBOL(inetdev_by_index);
 531
 532 /* Called only from RTNL semaphored context. No locks. */
 533
 534 struct in_ifaddr *inet_ifa_byprefix(struct in_device *in_dev, __be32 prefix,
 535                                     __be32 mask)
 536 {
 537         ASSERT_RTNL();
 538
 539         for_primary_ifa(in_dev) {
 540                 if (ifa->ifa_mask == mask && inet_ifa_match(prefix, ifa))
 541                         return ifa;
 542         } endfor_ifa(in_dev);
 543         return NULL;
 544 }
 545
 546 static int inet_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh)
 547 {
 548         struct net *net = sock_net(skb->sk);
 549         struct nlattr *tb[IFA_MAX+1];
 550         struct in_device *in_dev;
 551         struct ifaddrmsg *ifm;
 552         struct in_ifaddr *ifa, **ifap;
 553         int err = -EINVAL;
 554
 555         ASSERT_RTNL();
 556
 557         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv4_policy);
 558         if (err < 0)
 559                 goto errout;
 560
 561         ifm = nlmsg_data(nlh);
 562         in_dev = inetdev_by_index(net, ifm->ifa_index);
 563         if (in_dev == NULL) {
 564                 err = -ENODEV;
 565                 goto errout;
 566         }
 567
 568         for (ifap = &in_dev->ifa_list; (ifa = *ifap) != NULL;
 569              ifap = &ifa->ifa_next) {
 570                 if (tb[IFA_LOCAL] &&
 571                     ifa->ifa_local != nla_get_be32(tb[IFA_LOCAL]))
 572                         continue;
 573
 574                 if (tb[IFA_LABEL] && nla_strcmp(tb[IFA_LABEL], ifa->ifa_label))
 575                         continue;
 576
 577                 if (tb[IFA_ADDRESS] &&
 578                     (ifm->ifa_prefixlen != ifa->ifa_prefixlen ||
 579                     !inet_ifa_match(nla_get_be32(tb[IFA_ADDRESS]), ifa)))
 580                         continue;
 581
 582                 __inet_del_ifa(in_dev, ifap, 1, nlh, NETLINK_CB(skb).portid);
 583                 return 0;
 584         }
 585
 586         err = -EADDRNOTAVAIL;
 587 errout:
 588         return err;
 589 }
 590
 591 #define INFINITY_LIFE_TIME      0xFFFFFFFF
 592
 593 static void check_lifetime(struct work_struct *work)
 594 {
 595         unsigned long now, next, next_sec, next_sched;
 596         struct in_ifaddr *ifa;
 597         struct hlist_node *n;
 598         int i;
 599
 600         now = jiffies;
 601         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
 602
 603         for (i = 0; i < IN4_ADDR_HSIZE; i++) {
 604                 bool change_needed = false;
 605
 606                 rcu_read_lock();
 607                 hlist_for_each_entry_rcu(ifa, &inet_addr_lst[i], hash) {
 608                         unsigned long age;
 609
 610                         if (ifa->ifa_flags & IFA_F_PERMANENT)
 611                                 continue;
 612
 613                         /* We try to batch several events at once. */
 614                         age = (now - ifa->ifa_tstamp +
 615                                ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
 616
 617                         if (ifa->ifa_valid_lft != INFINITY_LIFE_TIME &&
 618                             age >= ifa->ifa_valid_lft) {
 619                                 change_needed = true;
 620                         } else if (ifa->ifa_preferred_lft ==
 621                                    INFINITY_LIFE_TIME) {
 622                                 continue;
 623                         } else if (age >= ifa->ifa_preferred_lft) {
 624                                 if (time_before(ifa->ifa_tstamp +
 625                                                 ifa->ifa_valid_lft * HZ, next))
 626                                         next = ifa->ifa_tstamp +
 627                                                ifa->ifa_valid_lft * HZ;
 628
 629                                 if (!(ifa->ifa_flags & IFA_F_DEPRECATED))
 630                                         change_needed = true;
 631                         } else if (time_before(ifa->ifa_tstamp +
 632                                                ifa->ifa_preferred_lft * HZ,
 633                                                next)) {
 634                                 next = ifa->ifa_tstamp +
 635                                        ifa->ifa_preferred_lft * HZ;
 636                         }
 637                 }
 638                 rcu_read_unlock();
 639                 if (!change_needed)
 640                         continue;
 641                 rtnl_lock();
 642                 hlist_for_each_entry_safe(ifa, n, &inet_addr_lst[i], hash) {
 643                         unsigned long age;
 644
 645                         if (ifa->ifa_flags & IFA_F_PERMANENT)
 646                                 continue;
 647
 648                         /* We try to batch several events at once. */
 649                         age = (now - ifa->ifa_tstamp +
 650                                ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
 651
 652                         if (ifa->ifa_valid_lft != INFINITY_LIFE_TIME &&
 653                             age >= ifa->ifa_valid_lft) {
 654                                 struct in_ifaddr **ifap;
 655
 656                                 for (ifap = &ifa->ifa_dev->ifa_list;
 657                                      *ifap != NULL; ifap = &(*ifap)->ifa_next) {
 658                                         if (*ifap == ifa) {
 659                                                 inet_del_ifa(ifa->ifa_dev,
 660                                                              ifap, 1);
 661                                                 break;
 662                                         }
 663                                 }
 664                         } else if (ifa->ifa_preferred_lft !=
 665                                    INFINITY_LIFE_TIME &&
 666                                    age >= ifa->ifa_preferred_lft &&
 667                                    !(ifa->ifa_flags & IFA_F_DEPRECATED)) {
 668                                 ifa->ifa_flags |= IFA_F_DEPRECATED;
 669                                 rtmsg_ifa(RTM_NEWADDR, ifa, NULL, 0);
 670                         }
 671                 }
 672                 rtnl_unlock();
 673         }
 674
 675         next_sec = round_jiffies_up(next);
 676         next_sched = next;
 677
 678         /* If rounded timeout is accurate enough, accept it. */
 679         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
 680                 next_sched = next_sec;
 681
 682         now = jiffies;
 683         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
 684         if (time_before(next_sched, now + ADDRCONF_TIMER_FUZZ_MAX))
 685                 next_sched = now + ADDRCONF_TIMER_FUZZ_MAX;
 686
 687         schedule_delayed_work(&check_lifetime_work, next_sched - now);
 688 }
 689
 690 static void set_ifa_lifetime(struct in_ifaddr *ifa, __u32 valid_lft,
 691                              __u32 prefered_lft)
 692 {
 693         unsigned long timeout;
 694
 695         ifa->ifa_flags &= ~(IFA_F_PERMANENT | IFA_F_DEPRECATED);
 696
 697         timeout = addrconf_timeout_fixup(valid_lft, HZ);
 698         if (addrconf_finite_timeout(timeout))
 699                 ifa->ifa_valid_lft = timeout;
 700         else
 701                 ifa->ifa_flags |= IFA_F_PERMANENT;
 702
 703         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
 704         if (addrconf_finite_timeout(timeout)) {
 705                 if (timeout == 0)
 706                         ifa->ifa_flags |= IFA_F_DEPRECATED;
 707                 ifa->ifa_preferred_lft = timeout;
 708         }
 709         ifa->ifa_tstamp = jiffies;
 710         if (!ifa->ifa_cstamp)
 711                 ifa->ifa_cstamp = ifa->ifa_tstamp;
 712 }
 713
 714 static struct in_ifaddr *rtm_to_ifaddr(struct net *net, struct nlmsghdr *nlh,
 715                                        __u32 *pvalid_lft, __u32 *pprefered_lft)
 716 {
 717         struct nlattr *tb[IFA_MAX+1];
 718         struct in_ifaddr *ifa;
 719         struct ifaddrmsg *ifm;
 720         struct net_device *dev;
 721         struct in_device *in_dev;
 722         int err;
 723
 724         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv4_policy);
 725         if (err < 0)
 726                 goto errout;
 727
 728         ifm = nlmsg_data(nlh);
 729         err = -EINVAL;
 730         if (ifm->ifa_prefixlen > 32 || tb[IFA_LOCAL] == NULL)
 731                 goto errout;
 732
 733         dev = __dev_get_by_index(net, ifm->ifa_index);
 734         err = -ENODEV;
 735         if (dev == NULL)
 736                 goto errout;
 737
 738         in_dev = __in_dev_get_rtnl(dev);
 739         err = -ENOBUFS;
 740         if (in_dev == NULL)
 741                 goto errout;
 742
 743         ifa = inet_alloc_ifa();
 744         if (ifa == NULL)
 745                 /*
 746                  * A potential indev allocation can be left alive, it stays
 747                  * assigned to its device and is destroy with it.
 748                  */
 749                 goto errout;
 750
 751         ipv4_devconf_setall(in_dev);
 752         neigh_parms_data_state_setall(in_dev->arp_parms);
 753         in_dev_hold(in_dev);
 754
 755         if (tb[IFA_ADDRESS] == NULL)
 756                 tb[IFA_ADDRESS] = tb[IFA_LOCAL];
 757
 758         INIT_HLIST_NODE(&ifa->hash);
 759         ifa->ifa_prefixlen = ifm->ifa_prefixlen;
 760         ifa->ifa_mask = inet_make_mask(ifm->ifa_prefixlen);
 761         ifa->ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) :
 762                                          ifm->ifa_flags;
 763         ifa->ifa_scope = ifm->ifa_scope;
 764         ifa->ifa_dev = in_dev;
 765
 766         ifa->ifa_local = nla_get_be32(tb[IFA_LOCAL]);
 767         ifa->ifa_address = nla_get_be32(tb[IFA_ADDRESS]);
 768
 769         if (tb[IFA_BROADCAST])
 770                 ifa->ifa_broadcast = nla_get_be32(tb[IFA_BROADCAST]);
 771
 772         if (tb[IFA_LABEL])
 773                 nla_strlcpy(ifa->ifa_label, tb[IFA_LABEL], IFNAMSIZ);
 774         else
 775                 memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
 776
 777         if (tb[IFA_CACHEINFO]) {
 778                 struct ifa_cacheinfo *ci;
 779
 780                 ci = nla_data(tb[IFA_CACHEINFO]);
 781                 if (!ci->ifa_valid || ci->ifa_prefered > ci->ifa_valid) {
 782                         err = -EINVAL;
 783                         goto errout_free;
 784                 }
 785                 *pvalid_lft = ci->ifa_valid;
 786                 *pprefered_lft = ci->ifa_prefered;
 787         }
 788
 789         return ifa;
 790
 791 errout_free:
 792         inet_free_ifa(ifa);
 793 errout:
 794         return ERR_PTR(err);
 795 }
 796
 797 static struct in_ifaddr *find_matching_ifa(struct in_ifaddr *ifa)
 798 {
 799         struct in_device *in_dev = ifa->ifa_dev;
 800         struct in_ifaddr *ifa1, **ifap;
 801
 802         if (!ifa->ifa_local)
 803                 return NULL;
 804
 805         for (ifap = &in_dev->ifa_list; (ifa1 = *ifap) != NULL;
 806              ifap = &ifa1->ifa_next) {
 807                 if (ifa1->ifa_mask == ifa->ifa_mask &&
 808                     inet_ifa_match(ifa1->ifa_address, ifa) &&
 809                     ifa1->ifa_local == ifa->ifa_local)
 810                         return ifa1;
 811         }
 812         return NULL;
 813 }
 814
 815 static int inet_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh)
 816 {
 817         struct net *net = sock_net(skb->sk);
 818         struct in_ifaddr *ifa;
 819         struct in_ifaddr *ifa_existing;
 820         __u32 valid_lft = INFINITY_LIFE_TIME;
 821         __u32 prefered_lft = INFINITY_LIFE_TIME;
 822
 823         ASSERT_RTNL();
 824
 825         ifa = rtm_to_ifaddr(net, nlh, &valid_lft, &prefered_lft);
 826         if (IS_ERR(ifa))
 827                 return PTR_ERR(ifa);
 828
 829         ifa_existing = find_matching_ifa(ifa);
 830         if (!ifa_existing) {
 831                 /* It would be best to check for !NLM_F_CREATE here but
 832                  * userspace alreay relies on not having to provide this.
 833                  */
 834                 set_ifa_lifetime(ifa, valid_lft, prefered_lft);
 835                 return __inet_insert_ifa(ifa, nlh, NETLINK_CB(skb).portid);
 836         } else {
 837                 inet_free_ifa(ifa);
 838
 839                 if (nlh->nlmsg_flags & NLM_F_EXCL ||
 840                     !(nlh->nlmsg_flags & NLM_F_REPLACE))
 841                         return -EEXIST;
 842                 ifa = ifa_existing;
 843                 set_ifa_lifetime(ifa, valid_lft, prefered_lft);
 844                 cancel_delayed_work(&check_lifetime_work);
 845                 schedule_delayed_work(&check_lifetime_work, 0);
 846                 rtmsg_ifa(RTM_NEWADDR, ifa, nlh, NETLINK_CB(skb).portid);
 847                 blocking_notifier_call_chain(&inetaddr_chain, NETDEV_UP, ifa);
 848         }
 849         return 0;
 850 }
 851
 852 /*
 853  *      Determine a default network mask, based on the IP address.
 854  */
 855
 856 static int inet_abc_len(__be32 addr)
 857 {
 858         int rc = -1;    /* Something else, probably a multicast. */
 859
 860         if (ipv4_is_zeronet(addr))
 861                 rc = 0;
 862         else {
 863                 __u32 haddr = ntohl(addr);
 864
 865                 if (IN_CLASSA(haddr))
 866                         rc = 8;
 867                 else if (IN_CLASSB(haddr))
 868                         rc = 16;
 869                 else if (IN_CLASSC(haddr))
 870                         rc = 24;
 871         }
 872
 873         return rc;
 874 }
 875
 876
 877 int devinet_ioctl(struct net *net, unsigned int cmd, void __user *arg)
 878 {
 879         struct ifreq ifr;
 880         struct sockaddr_in sin_orig;
 881         struct sockaddr_in *sin = (struct sockaddr_in *)&ifr.ifr_addr;
 882         struct in_device *in_dev;
 883         struct in_ifaddr **ifap = NULL;
 884         struct in_ifaddr *ifa = NULL;
 885         struct net_device *dev;
 886         char *colon;
 887         int ret = -EFAULT;
 888         int tryaddrmatch = 0;
 889
 890         /*
 891          *      Fetch the caller's info block into kernel space
 892          */
 893
 894         if (copy_from_user(&ifr, arg, sizeof(struct ifreq)))
 895                 goto out;
 896         ifr.ifr_name[IFNAMSIZ - 1] = 0;
 897
 898         /* save original address for comparison */
 899         memcpy(&sin_orig, sin, sizeof(*sin));
 900
 901         colon = strchr(ifr.ifr_name, ':');
 902         if (colon)
 903                 *colon = 0;
 904
 905         dev_load(net, ifr.ifr_name);
 906
 907         switch (cmd) {
 908         case SIOCGIFADDR:       /* Get interface address */
 909         case SIOCGIFBRDADDR:    /* Get the broadcast address */
 910         case SIOCGIFDSTADDR:    /* Get the destination address */
 911         case SIOCGIFNETMASK:    /* Get the netmask for the interface */
 912                 /* Note that these ioctls will not sleep,
 913                    so that we do not impose a lock.
 914                    One day we will be forced to put shlock here (I mean SMP)
 915                  */
 916                 tryaddrmatch = (sin_orig.sin_family == AF_INET);
 917                 memset(sin, 0, sizeof(*sin));
 918                 sin->sin_family = AF_INET;
 919                 break;
 920
 921         case SIOCSIFFLAGS:
 922                 ret = -EPERM;
 923                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
 924                         goto out;
 925                 break;
 926         case SIOCSIFADDR:       /* Set interface address (and family) */
 927         case SIOCSIFBRDADDR:    /* Set the broadcast address */
 928         case SIOCSIFDSTADDR:    /* Set the destination address */
 929         case SIOCSIFNETMASK:    /* Set the netmask for the interface */
 930                 ret = -EPERM;
 931                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
 932                         goto out;
 933                 ret = -EINVAL;
 934                 if (sin->sin_family != AF_INET)
 935                         goto out;
 936                 break;
 937         default:
 938                 ret = -EINVAL;
 939                 goto out;
 940         }
 941
 942         rtnl_lock();
 943
 944         ret = -ENODEV;
 945         dev = __dev_get_by_name(net, ifr.ifr_name);
 946         if (!dev)
 947                 goto done;
 948
 949         if (colon)
 950                 *colon = ':';
 951
 952         in_dev = __in_dev_get_rtnl(dev);
 953         if (in_dev) {
 954                 if (tryaddrmatch) {
 955                         /* Matthias Andree */
 956                         /* compare label and address (4.4BSD style) */
 957                         /* note: we only do this for a limited set of ioctls
 958                            and only if the original address family was AF_INET.
 959                            This is checked above. */
 960                         for (ifap = &in_dev->ifa_list; (ifa = *ifap) != NULL;
 961                              ifap = &ifa->ifa_next) {
 962                                 if (!strcmp(ifr.ifr_name, ifa->ifa_label) &&
 963                                     sin_orig.sin_addr.s_addr ==
 964                                                         ifa->ifa_local) {
 965                                         break; /* found */
 966                                 }
 967                         }
 968                 }
 969                 /* we didn't get a match, maybe the application is
 970                    4.3BSD-style and passed in junk so we fall back to
 971                    comparing just the label */
 972                 if (!ifa) {
 973                         for (ifap = &in_dev->ifa_list; (ifa = *ifap) != NULL;
 974                              ifap = &ifa->ifa_next)
 975                                 if (!strcmp(ifr.ifr_name, ifa->ifa_label))
 976                                         break;
 977                 }
 978         }
 979
 980         ret = -EADDRNOTAVAIL;
 981         if (!ifa && cmd != SIOCSIFADDR && cmd != SIOCSIFFLAGS)
 982                 goto done;
 983
 984         switch (cmd) {
 985         case SIOCGIFADDR:       /* Get interface address */
 986                 sin->sin_addr.s_addr = ifa->ifa_local;
 987                 goto rarok;
 988
 989         case SIOCGIFBRDADDR:    /* Get the broadcast address */
 990                 sin->sin_addr.s_addr = ifa->ifa_broadcast;
 991                 goto rarok;
 992
 993         case SIOCGIFDSTADDR:    /* Get the destination address */
 994                 sin->sin_addr.s_addr = ifa->ifa_address;
 995                 goto rarok;
 996
 997         case SIOCGIFNETMASK:    /* Get the netmask for the interface */
 998                 sin->sin_addr.s_addr = ifa->ifa_mask;
 999                 goto rarok;
1000
1001         case SIOCSIFFLAGS:
1002                 if (colon) {
1003                         ret = -EADDRNOTAVAIL;
1004                         if (!ifa)
1005                                 break;
1006                         ret = 0;
1007                         if (!(ifr.ifr_flags & IFF_UP))
1008                                 inet_del_ifa(in_dev, ifap, 1);
1009                         break;
1010                 }
1011                 ret = dev_change_flags(dev, ifr.ifr_flags);
1012                 break;
1013
1014         case SIOCSIFADDR:       /* Set interface address (and family) */
1015                 ret = -EINVAL;
1016                 if (inet_abc_len(sin->sin_addr.s_addr) < 0)
1017                         break;
1018
1019                 if (!ifa) {
1020                         ret = -ENOBUFS;
1021                         ifa = inet_alloc_ifa();
1022                         if (!ifa)
1023                                 break;
1024                         INIT_HLIST_NODE(&ifa->hash);
1025                         if (colon)
1026                                 memcpy(ifa->ifa_label, ifr.ifr_name, IFNAMSIZ);
1027                         else
1028                                 memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
1029                 } else {
1030                         ret = 0;
1031                         if (ifa->ifa_local == sin->sin_addr.s_addr)
1032                                 break;
1033                         inet_del_ifa(in_dev, ifap, 0);
1034                         ifa->ifa_broadcast = 0;
1035                         ifa->ifa_scope = 0;
1036                 }
1037
1038                 ifa->ifa_address = ifa->ifa_local = sin->sin_addr.s_addr;
1039
1040                 if (!(dev->flags & IFF_POINTOPOINT)) {
1041                         ifa->ifa_prefixlen = inet_abc_len(ifa->ifa_address);
1042                         ifa->ifa_mask = inet_make_mask(ifa->ifa_prefixlen);
1043                         if ((dev->flags & IFF_BROADCAST) &&
1044                             ifa->ifa_prefixlen < 31)
1045                                 ifa->ifa_broadcast = ifa->ifa_address |
1046                                                      ~ifa->ifa_mask;
1047                 } else {
1048                         ifa->ifa_prefixlen = 32;
1049                         ifa->ifa_mask = inet_make_mask(32);
1050                 }
1051                 set_ifa_lifetime(ifa, INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
1052                 ret = inet_set_ifa(dev, ifa);
1053                 break;
1054
1055         case SIOCSIFBRDADDR:    /* Set the broadcast address */
1056                 ret = 0;
1057                 if (ifa->ifa_broadcast != sin->sin_addr.s_addr) {
1058                         inet_del_ifa(in_dev, ifap, 0);
1059                         ifa->ifa_broadcast = sin->sin_addr.s_addr;
1060                         inet_insert_ifa(ifa);
1061                 }
1062                 break;
1063
1064         case SIOCSIFDSTADDR:    /* Set the destination address */
1065                 ret = 0;
1066                 if (ifa->ifa_address == sin->sin_addr.s_addr)
1067                         break;
1068                 ret = -EINVAL;
1069                 if (inet_abc_len(sin->sin_addr.s_addr) < 0)
1070                         break;
1071                 ret = 0;
1072                 inet_del_ifa(in_dev, ifap, 0);
1073                 ifa->ifa_address = sin->sin_addr.s_addr;
1074                 inet_insert_ifa(ifa);
1075                 break;
1076
1077         case SIOCSIFNETMASK:    /* Set the netmask for the interface */
1078
1079                 /*
1080                  *      The mask we set must be legal.
1081                  */
1082                 ret = -EINVAL;
1083                 if (bad_mask(sin->sin_addr.s_addr, 0))
1084                         break;
1085                 ret = 0;
1086                 if (ifa->ifa_mask != sin->sin_addr.s_addr) {
1087                         __be32 old_mask = ifa->ifa_mask;
1088                         inet_del_ifa(in_dev, ifap, 0);
1089                         ifa->ifa_mask = sin->sin_addr.s_addr;
1090                         ifa->ifa_prefixlen = inet_mask_len(ifa->ifa_mask);
1091
1092                         /* See if current broadcast address matches
1093                          * with current netmask, then recalculate
1094                          * the broadcast address. Otherwise it's a
1095                          * funny address, so don't touch it since
1096                          * the user seems to know what (s)he's doing...
1097                          */
1098                         if ((dev->flags & IFF_BROADCAST) &&
1099                             (ifa->ifa_prefixlen < 31) &&
1100                             (ifa->ifa_broadcast ==
1101                              (ifa->ifa_local|~old_mask))) {
1102                                 ifa->ifa_broadcast = (ifa->ifa_local |
1103                                                       ~sin->sin_addr.s_addr);
1104                         }
1105                         inet_insert_ifa(ifa);
1106                 }
1107                 break;
1108         }
1109 done:
1110         rtnl_unlock();
1111 out:
1112         return ret;
1113 rarok:
1114         rtnl_unlock();
1115         ret = copy_to_user(arg, &ifr, sizeof(struct ifreq)) ? -EFAULT : 0;
1116         goto out;
1117 }
1118
1119 static int inet_gifconf(struct net_device *dev, char __user *buf, int len)
1120 {
1121         struct in_device *in_dev = __in_dev_get_rtnl(dev);
1122         struct in_ifaddr *ifa;
1123         struct ifreq ifr;
1124         int done = 0;
1125
1126         if (!in_dev)
1127                 goto out;
1128
1129         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
1130                 if (!buf) {
1131                         done += sizeof(ifr);
1132                         continue;
1133                 }
1134                 if (len < (int) sizeof(ifr))
1135                         break;
1136                 memset(&ifr, 0, sizeof(struct ifreq));
1137                 strcpy(ifr.ifr_name, ifa->ifa_label);
1138
1139                 (*(struct sockaddr_in *)&ifr.ifr_addr).sin_family = AF_INET;
1140                 (*(struct sockaddr_in *)&ifr.ifr_addr).sin_addr.s_addr =
1141                                                                 ifa->ifa_local;
1142
1143                 if (copy_to_user(buf, &ifr, sizeof(struct ifreq))) {
1144                         done = -EFAULT;
1145                         break;
1146                 }
1147                 buf  += sizeof(struct ifreq);
1148                 len  -= sizeof(struct ifreq);
1149                 done += sizeof(struct ifreq);
1150         }
1151 out:
1152         return done;
1153 }
1154
1155 __be32 inet_select_addr(const struct net_device *dev, __be32 dst, int scope)
1156 {
1157         __be32 addr = 0;
1158         struct in_device *in_dev;
1159         struct net *net = dev_net(dev);
1160
1161         rcu_read_lock();
1162         in_dev = __in_dev_get_rcu(dev);
1163         if (!in_dev)
1164                 goto no_in_dev;
1165
1166         for_primary_ifa(in_dev) {
1167                 if (ifa->ifa_scope > scope)
1168                         continue;
1169                 if (!dst || inet_ifa_match(dst, ifa)) {
1170                         addr = ifa->ifa_local;
1171                         break;
1172                 }
1173                 if (!addr)
1174                         addr = ifa->ifa_local;
1175         } endfor_ifa(in_dev);
1176
1177         if (addr)
1178                 goto out_unlock;
1179 no_in_dev:
1180
1181         /* Not loopback addresses on loopback should be preferred
1182            in this case. It is importnat that lo is the first interface
1183            in dev_base list.
1184          */
1185         for_each_netdev_rcu(net, dev) {
1186                 in_dev = __in_dev_get_rcu(dev);
1187                 if (!in_dev)
1188                         continue;
1189
1190                 for_primary_ifa(in_dev) {
1191                         if (ifa->ifa_scope != RT_SCOPE_LINK &&
1192                             ifa->ifa_scope <= scope) {
1193                                 addr = ifa->ifa_local;
1194                                 goto out_unlock;
1195                         }
1196                 } endfor_ifa(in_dev);
1197         }
1198 out_unlock:
1199         rcu_read_unlock();
1200         return addr;
1201 }
1202 EXPORT_SYMBOL(inet_select_addr);
1203
1204 static __be32 confirm_addr_indev(struct in_device *in_dev, __be32 dst,
1205                               __be32 local, int scope)
1206 {
1207         int same = 0;
1208         __be32 addr = 0;
1209
1210         for_ifa(in_dev) {
1211                 if (!addr &&
1212                     (local == ifa->ifa_local || !local) &&
1213                     ifa->ifa_scope <= scope) {
1214                         addr = ifa->ifa_local;
1215                         if (same)
1216                                 break;
1217                 }
1218                 if (!same) {
1219                         same = (!local || inet_ifa_match(local, ifa)) &&
1220                                 (!dst || inet_ifa_match(dst, ifa));
1221                         if (same && addr) {
1222                                 if (local || !dst)
1223                                         break;
1224                                 /* Is the selected addr into dst subnet? */
1225                                 if (inet_ifa_match(addr, ifa))
1226                                         break;
1227                                 /* No, then can we use new local src? */
1228                                 if (ifa->ifa_scope <= scope) {
1229                                         addr = ifa->ifa_local;
1230                                         break;
1231                                 }
1232                                 /* search for large dst subnet for addr */
1233                                 same = 0;
1234                         }
1235                 }
1236         } endfor_ifa(in_dev);
1237
1238         return same ? addr : 0;
1239 }
1240
1241 /*
1242  * Confirm that local IP address exists using wildcards:
1243  * - net: netns to check, cannot be NULL
1244  * - in_dev: only on this interface, NULL=any interface
1245  * - dst: only in the same subnet as dst, 0=any dst
1246  * - local: address, 0=autoselect the local address
1247  * - scope: maximum allowed scope value for the local address
1248  */
1249 __be32 inet_confirm_addr(struct net *net, struct in_device *in_dev,
1250                          __be32 dst, __be32 local, int scope)
1251 {
1252         __be32 addr = 0;
1253         struct net_device *dev;
1254
1255         if (in_dev != NULL)
1256                 return confirm_addr_indev(in_dev, dst, local, scope);
1257
1258         rcu_read_lock();
1259         for_each_netdev_rcu(net, dev) {
1260                 in_dev = __in_dev_get_rcu(dev);
1261                 if (in_dev) {
1262                         addr = confirm_addr_indev(in_dev, dst, local, scope);
1263                         if (addr)
1264                                 break;
1265                 }
1266         }
1267         rcu_read_unlock();
1268
1269         return addr;
1270 }
1271 EXPORT_SYMBOL(inet_confirm_addr);
1272
1273 /*
1274  *      Device notifier
1275  */
1276
1277 int register_inetaddr_notifier(struct notifier_block *nb)
1278 {
1279         return blocking_notifier_chain_register(&inetaddr_chain, nb);
1280 }
1281 EXPORT_SYMBOL(register_inetaddr_notifier);
1282
1283 int unregister_inetaddr_notifier(struct notifier_block *nb)
1284 {
1285         return blocking_notifier_chain_unregister(&inetaddr_chain, nb);
1286 }
1287 EXPORT_SYMBOL(unregister_inetaddr_notifier);
1288
1289 /* Rename ifa_labels for a device name change. Make some effort to preserve
1290  * existing alias numbering and to create unique labels if possible.
1291 */
1292 static void inetdev_changename(struct net_device *dev, struct in_device *in_dev)
1293 {
1294         struct in_ifaddr *ifa;
1295         int named = 0;
1296
1297         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
1298                 char old[IFNAMSIZ], *dot;
1299
1300                 memcpy(old, ifa->ifa_label, IFNAMSIZ);
1301                 memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
1302                 if (named++ == 0)
1303                         goto skip;
1304                 dot = strchr(old, ':');
1305                 if (dot == NULL) {
1306                         sprintf(old, ":%d", named);
1307                         dot = old;
1308                 }
1309                 if (strlen(dot) + strlen(dev->name) < IFNAMSIZ)
1310                         strcat(ifa->ifa_label, dot);
1311                 else
1312                         strcpy(ifa->ifa_label + (IFNAMSIZ - strlen(dot) - 1), dot);
1313 skip:
1314                 rtmsg_ifa(RTM_NEWADDR, ifa, NULL, 0);
1315         }
1316 }
1317
1318 static bool inetdev_valid_mtu(unsigned int mtu)
1319 {
1320         return mtu >= 68;
1321 }
1322
1323 static void inetdev_send_gratuitous_arp(struct net_device *dev,
1324                                         struct in_device *in_dev)
1325
1326 {
1327         struct in_ifaddr *ifa;
1328
1329         for (ifa = in_dev->ifa_list; ifa;
1330              ifa = ifa->ifa_next) {
1331                 arp_send(ARPOP_REQUEST, ETH_P_ARP,
1332                          ifa->ifa_local, dev,
1333                          ifa->ifa_local, NULL,
1334                          dev->dev_addr, NULL);
1335         }
1336 }
1337
1338 /* Called only under RTNL semaphore */
1339
1340 static int inetdev_event(struct notifier_block *this, unsigned long event,
1341                          void *ptr)
1342 {
1343         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1344         struct in_device *in_dev = __in_dev_get_rtnl(dev);
1345
1346         ASSERT_RTNL();
1347
1348         if (!in_dev) {
1349                 if (event == NETDEV_REGISTER) {
1350                         in_dev = inetdev_init(dev);
1351                         if (!in_dev)
1352                                 return notifier_from_errno(-ENOMEM);
1353                         if (dev->flags & IFF_LOOPBACK) {
1354                                 IN_DEV_CONF_SET(in_dev, NOXFRM, 1);
1355                                 IN_DEV_CONF_SET(in_dev, NOPOLICY, 1);
1356                         }
1357                 } else if (event == NETDEV_CHANGEMTU) {
1358                         /* Re-enabling IP */
1359                         if (inetdev_valid_mtu(dev->mtu))
1360                                 in_dev = inetdev_init(dev);
1361                 }
1362                 goto out;
1363         }
1364
1365         switch (event) {
1366         case NETDEV_REGISTER:
1367                 pr_debug("%s: bug\n", __func__);
1368                 RCU_INIT_POINTER(dev->ip_ptr, NULL);
1369                 break;
1370         case NETDEV_UP:
1371                 if (!inetdev_valid_mtu(dev->mtu))
1372                         break;
1373                 if (dev->flags & IFF_LOOPBACK) {
1374                         struct in_ifaddr *ifa = inet_alloc_ifa();
1375
1376                         if (ifa) {
1377                                 INIT_HLIST_NODE(&ifa->hash);
1378                                 ifa->ifa_local =
1379                                   ifa->ifa_address = htonl(INADDR_LOOPBACK);
1380                                 ifa->ifa_prefixlen = 8;
1381                                 ifa->ifa_mask = inet_make_mask(8);
1382                                 in_dev_hold(in_dev);
1383                                 ifa->ifa_dev = in_dev;
1384                                 ifa->ifa_scope = RT_SCOPE_HOST;
1385                                 memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
1386                                 set_ifa_lifetime(ifa, INFINITY_LIFE_TIME,
1387                                                  INFINITY_LIFE_TIME);
1388                                 inet_insert_ifa(ifa);
1389                         }
1390                 }
1391                 ip_mc_up(in_dev);
1392                 /* fall through */
1393         case NETDEV_CHANGEADDR:
1394                 if (!IN_DEV_ARP_NOTIFY(in_dev))
1395                         break;
1396                 /* fall through */
1397         case NETDEV_NOTIFY_PEERS:
1398                 /* Send gratuitous ARP to notify of link change */
1399                 inetdev_send_gratuitous_arp(dev, in_dev);
1400                 break;
1401         case NETDEV_DOWN:
1402                 ip_mc_down(in_dev);
1403                 break;
1404         case NETDEV_PRE_TYPE_CHANGE:
1405                 ip_mc_unmap(in_dev);
1406                 break;
1407         case NETDEV_POST_TYPE_CHANGE:
1408                 ip_mc_remap(in_dev);
1409                 break;
1410         case NETDEV_CHANGEMTU:
1411                 if (inetdev_valid_mtu(dev->mtu))
1412                         break;
1413                 /* disable IP when MTU is not enough */
1414         case NETDEV_UNREGISTER:
1415                 inetdev_destroy(in_dev);
1416                 break;
1417         case NETDEV_CHANGENAME:
1418                 /* Do not notify about label change, this event is
1419                  * not interesting to applications using netlink.
1420                  */
1421                 inetdev_changename(dev, in_dev);
1422
1423                 devinet_sysctl_unregister(in_dev);
1424                 devinet_sysctl_register(in_dev);
1425                 break;
1426         }
1427 out:
1428         return NOTIFY_DONE;
1429 }
1430
1431 static struct notifier_block ip_netdev_notifier = {
1432         .notifier_call = inetdev_event,
1433 };
1434
1435 static size_t inet_nlmsg_size(void)
1436 {
1437         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
1438                + nla_total_size(4) /* IFA_ADDRESS */
1439                + nla_total_size(4) /* IFA_LOCAL */
1440                + nla_total_size(4) /* IFA_BROADCAST */
1441                + nla_total_size(IFNAMSIZ) /* IFA_LABEL */
1442                + nla_total_size(4);  /* IFA_FLAGS */
1443 }
1444
1445 static inline u32 cstamp_delta(unsigned long cstamp)
1446 {
1447         return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
1448 }
1449
1450 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
1451                          unsigned long tstamp, u32 preferred, u32 valid)
1452 {
1453         struct ifa_cacheinfo ci;
1454
1455         ci.cstamp = cstamp_delta(cstamp);
1456         ci.tstamp = cstamp_delta(tstamp);
1457         ci.ifa_prefered = preferred;
1458         ci.ifa_valid = valid;
1459
1460         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
1461 }
1462
1463 static int inet_fill_ifaddr(struct sk_buff *skb, struct in_ifaddr *ifa,
1464                             u32 portid, u32 seq, int event, unsigned int flags)
1465 {
1466         struct ifaddrmsg *ifm;
1467         struct nlmsghdr  *nlh;
1468         u32 preferred, valid;
1469
1470         nlh = nlmsg_put(skb, portid, seq, event, sizeof(*ifm), flags);
1471         if (nlh == NULL)
1472                 return -EMSGSIZE;
1473
1474         ifm = nlmsg_data(nlh);
1475         ifm->ifa_family = AF_INET;
1476         ifm->ifa_prefixlen = ifa->ifa_prefixlen;
1477         ifm->ifa_flags = ifa->ifa_flags;
1478         ifm->ifa_scope = ifa->ifa_scope;
1479         ifm->ifa_index = ifa->ifa_dev->dev->ifindex;
1480
1481         if (!(ifm->ifa_flags & IFA_F_PERMANENT)) {
1482                 preferred = ifa->ifa_preferred_lft;
1483                 valid = ifa->ifa_valid_lft;
1484                 if (preferred != INFINITY_LIFE_TIME) {
1485                         long tval = (jiffies - ifa->ifa_tstamp) / HZ;
1486
1487                         if (preferred > tval)
1488                                 preferred -= tval;
1489                         else
1490                                 preferred = 0;
1491                         if (valid != INFINITY_LIFE_TIME) {
1492                                 if (valid > tval)
1493                                         valid -= tval;
1494                                 else
1495                                         valid = 0;
1496                         }
1497                 }
1498         } else {
1499                 preferred = INFINITY_LIFE_TIME;
1500                 valid = INFINITY_LIFE_TIME;
1501         }
1502         if ((ifa->ifa_address &&
1503              nla_put_be32(skb, IFA_ADDRESS, ifa->ifa_address)) ||
1504             (ifa->ifa_local &&
1505              nla_put_be32(skb, IFA_LOCAL, ifa->ifa_local)) ||
1506             (ifa->ifa_broadcast &&
1507              nla_put_be32(skb, IFA_BROADCAST, ifa->ifa_broadcast)) ||
1508             (ifa->ifa_label[0] &&
1509              nla_put_string(skb, IFA_LABEL, ifa->ifa_label)) ||
1510             nla_put_u32(skb, IFA_FLAGS, ifa->ifa_flags) ||
1511             put_cacheinfo(skb, ifa->ifa_cstamp, ifa->ifa_tstamp,
1512                           preferred, valid))
1513                 goto nla_put_failure;
1514
1515         return nlmsg_end(skb, nlh);
1516
1517 nla_put_failure:
1518         nlmsg_cancel(skb, nlh);
1519         return -EMSGSIZE;
1520 }
1521
1522 static int inet_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
1523 {
1524         struct net *net = sock_net(skb->sk);
1525         int h, s_h;
1526         int idx, s_idx;
1527         int ip_idx, s_ip_idx;
1528         struct net_device *dev;
1529         struct in_device *in_dev;
1530         struct in_ifaddr *ifa;
1531         struct hlist_head *head;
1532
1533         s_h = cb->args[0];
1534         s_idx = idx = cb->args[1];
1535         s_ip_idx = ip_idx = cb->args[2];
1536
1537         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1538                 idx = 0;
1539                 head = &net->dev_index_head[h];
1540                 rcu_read_lock();
1541                 cb->seq = atomic_read(&net->ipv4.dev_addr_genid) ^
1542                           net->dev_base_seq;
1543                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
1544                         if (idx < s_idx)
1545                                 goto cont;
1546                         if (h > s_h || idx > s_idx)
1547                                 s_ip_idx = 0;
1548                         in_dev = __in_dev_get_rcu(dev);
1549                         if (!in_dev)
1550                                 goto cont;
1551
1552                         for (ifa = in_dev->ifa_list, ip_idx = 0; ifa;
1553                              ifa = ifa->ifa_next, ip_idx++) {
1554                                 if (ip_idx < s_ip_idx)
1555                                         continue;
1556                                 if (inet_fill_ifaddr(skb, ifa,
1557                                              NETLINK_CB(cb->skb).portid,
1558                                              cb->nlh->nlmsg_seq,
1559                                              RTM_NEWADDR, NLM_F_MULTI) <= 0) {
1560                                         rcu_read_unlock();
1561                                         goto done;
1562                                 }
1563                                 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1564                         }
1565 cont:
1566                         idx++;
1567                 }
1568                 rcu_read_unlock();
1569         }
1570
1571 done:
1572         cb->args[0] = h;
1573         cb->args[1] = idx;
1574         cb->args[2] = ip_idx;
1575
1576         return skb->len;
1577 }
1578
1579 static void rtmsg_ifa(int event, struct in_ifaddr *ifa, struct nlmsghdr *nlh,
1580                       u32 portid)
1581 {
1582         struct sk_buff *skb;
1583         u32 seq = nlh ? nlh->nlmsg_seq : 0;
1584         int err = -ENOBUFS;
1585         struct net *net;
1586
1587         net = dev_net(ifa->ifa_dev->dev);
1588         skb = nlmsg_new(inet_nlmsg_size(), GFP_KERNEL);
1589         if (skb == NULL)
1590                 goto errout;
1591
1592         err = inet_fill_ifaddr(skb, ifa, portid, seq, event, 0);
1593         if (err < 0) {
1594                 /* -EMSGSIZE implies BUG in inet_nlmsg_size() */
1595                 WARN_ON(err == -EMSGSIZE);
1596                 kfree_skb(skb);
1597                 goto errout;
1598         }
1599         rtnl_notify(skb, net, portid, RTNLGRP_IPV4_IFADDR, nlh, GFP_KERNEL);
1600         return;
1601 errout:
1602         if (err < 0)
1603                 rtnl_set_sk_err(net, RTNLGRP_IPV4_IFADDR, err);
1604 }
1605
1606 static size_t inet_get_link_af_size(const struct net_device *dev)
1607 {
1608         struct in_device *in_dev = rcu_dereference_rtnl(dev->ip_ptr);
1609
1610         if (!in_dev)
1611                 return 0;
1612
1613         return nla_total_size(IPV4_DEVCONF_MAX * 4); /* IFLA_INET_CONF */
1614 }
1615
1616 static int inet_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
1617 {
1618         struct in_device *in_dev = rcu_dereference_rtnl(dev->ip_ptr);
1619         struct nlattr *nla;
1620         int i;
1621
1622         if (!in_dev)
1623                 return -ENODATA;
1624
1625         nla = nla_reserve(skb, IFLA_INET_CONF, IPV4_DEVCONF_MAX * 4);
1626         if (nla == NULL)
1627                 return -EMSGSIZE;
1628
1629         for (i = 0; i < IPV4_DEVCONF_MAX; i++)
1630                 ((u32 *) nla_data(nla))[i] = in_dev->cnf.data[i];
1631
1632         return 0;
1633 }
1634
1635 static const struct nla_policy inet_af_policy[IFLA_INET_MAX+1] = {
1636         [IFLA_INET_CONF]        = { .type = NLA_NESTED },
1637 };
1638
1639 static int inet_validate_link_af(const struct net_device *dev,
1640                                  const struct nlattr *nla)
1641 {
1642         struct nlattr *a, *tb[IFLA_INET_MAX+1];
1643         int err, rem;
1644
1645         if (dev && !__in_dev_get_rtnl(dev))
1646                 return -EAFNOSUPPORT;
1647
1648         err = nla_parse_nested(tb, IFLA_INET_MAX, nla, inet_af_policy);
1649         if (err < 0)
1650                 return err;
1651
1652         if (tb[IFLA_INET_CONF]) {
1653                 nla_for_each_nested(a, tb[IFLA_INET_CONF], rem) {
1654                         int cfgid = nla_type(a);
1655
1656                         if (nla_len(a) < 4)
1657                                 return -EINVAL;
1658
1659                         if (cfgid <= 0 || cfgid > IPV4_DEVCONF_MAX)
1660                                 return -EINVAL;
1661                 }
1662         }
1663
1664         return 0;
1665 }
1666
1667 static int inet_set_link_af(struct net_device *dev, const struct nlattr *nla)
1668 {
1669         struct in_device *in_dev = __in_dev_get_rtnl(dev);
1670         struct nlattr *a, *tb[IFLA_INET_MAX+1];
1671         int rem;
1672
1673         if (!in_dev)
1674                 return -EAFNOSUPPORT;
1675
1676         if (nla_parse_nested(tb, IFLA_INET_MAX, nla, NULL) < 0)
1677                 BUG();
1678
1679         if (tb[IFLA_INET_CONF]) {
1680                 nla_for_each_nested(a, tb[IFLA_INET_CONF], rem)
1681                         ipv4_devconf_set(in_dev, nla_type(a), nla_get_u32(a));
1682         }
1683
1684         return 0;
1685 }
1686
1687 static int inet_netconf_msgsize_devconf(int type)
1688 {
1689         int size = NLMSG_ALIGN(sizeof(struct netconfmsg))
1690                    + nla_total_size(4); /* NETCONFA_IFINDEX */
1691
1692         /* type -1 is used for ALL */
1693         if (type == -1 || type == NETCONFA_FORWARDING)
1694                 size += nla_total_size(4);
1695         if (type == -1 || type == NETCONFA_RP_FILTER)
1696                 size += nla_total_size(4);
1697         if (type == -1 || type == NETCONFA_MC_FORWARDING)
1698                 size += nla_total_size(4);
1699
1700         return size;
1701 }
1702
1703 static int inet_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
1704                                      struct ipv4_devconf *devconf, u32 portid,
1705                                      u32 seq, int event, unsigned int flags,
1706                                      int type)
1707 {
1708         struct nlmsghdr  *nlh;
1709         struct netconfmsg *ncm;
1710
1711         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
1712                         flags);
1713         if (nlh == NULL)
1714                 return -EMSGSIZE;
1715
1716         ncm = nlmsg_data(nlh);
1717         ncm->ncm_family = AF_INET;
1718
1719         if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
1720                 goto nla_put_failure;
1721
1722         /* type -1 is used for ALL */
1723         if ((type == -1 || type == NETCONFA_FORWARDING) &&
1724             nla_put_s32(skb, NETCONFA_FORWARDING,
1725                         IPV4_DEVCONF(*devconf, FORWARDING)) < 0)
1726                 goto nla_put_failure;
1727         if ((type == -1 || type == NETCONFA_RP_FILTER) &&
1728             nla_put_s32(skb, NETCONFA_RP_FILTER,
1729                         IPV4_DEVCONF(*devconf, RP_FILTER)) < 0)
1730                 goto nla_put_failure;
1731         if ((type == -1 || type == NETCONFA_MC_FORWARDING) &&
1732             nla_put_s32(skb, NETCONFA_MC_FORWARDING,
1733                         IPV4_DEVCONF(*devconf, MC_FORWARDING)) < 0)
1734                 goto nla_put_failure;
1735
1736         return nlmsg_end(skb, nlh);
1737
1738 nla_put_failure:
1739         nlmsg_cancel(skb, nlh);
1740         return -EMSGSIZE;
1741 }
1742
1743 void inet_netconf_notify_devconf(struct net *net, int type, int ifindex,
1744                                  struct ipv4_devconf *devconf)
1745 {
1746         struct sk_buff *skb;
1747         int err = -ENOBUFS;
1748
1749         skb = nlmsg_new(inet_netconf_msgsize_devconf(type), GFP_ATOMIC);
1750         if (skb == NULL)
1751                 goto errout;
1752
1753         err = inet_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
1754                                         RTM_NEWNETCONF, 0, type);
1755         if (err < 0) {
1756                 /* -EMSGSIZE implies BUG in inet_netconf_msgsize_devconf() */
1757                 WARN_ON(err == -EMSGSIZE);
1758                 kfree_skb(skb);
1759                 goto errout;
1760         }
1761         rtnl_notify(skb, net, 0, RTNLGRP_IPV4_NETCONF, NULL, GFP_ATOMIC);
1762         return;
1763 errout:
1764         if (err < 0)
1765                 rtnl_set_sk_err(net, RTNLGRP_IPV4_NETCONF, err);
1766 }
1767
1768 static const struct nla_policy devconf_ipv4_policy[NETCONFA_MAX+1] = {
1769         [NETCONFA_IFINDEX]      = { .len = sizeof(int) },
1770         [NETCONFA_FORWARDING]   = { .len = sizeof(int) },
1771         [NETCONFA_RP_FILTER]    = { .len = sizeof(int) },
1772 };
1773
1774 static int inet_netconf_get_devconf(struct sk_buff *in_skb,
1775                                     struct nlmsghdr *nlh)
1776 {
1777         struct net *net = sock_net(in_skb->sk);
1778         struct nlattr *tb[NETCONFA_MAX+1];
1779         struct netconfmsg *ncm;
1780         struct sk_buff *skb;
1781         struct ipv4_devconf *devconf;
1782         struct in_device *in_dev;
1783         struct net_device *dev;
1784         int ifindex;
1785         int err;
1786
1787         err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX,
1788                           devconf_ipv4_policy);
1789         if (err < 0)
1790                 goto errout;
1791
1792         err = EINVAL;
1793         if (!tb[NETCONFA_IFINDEX])
1794                 goto errout;
1795
1796         ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
1797         switch (ifindex) {
1798         case NETCONFA_IFINDEX_ALL:
1799                 devconf = net->ipv4.devconf_all;
1800                 break;
1801         case NETCONFA_IFINDEX_DEFAULT:
1802                 devconf = net->ipv4.devconf_dflt;
1803                 break;
1804         default:
1805                 dev = __dev_get_by_index(net, ifindex);
1806                 if (dev == NULL)
1807                         goto errout;
1808                 in_dev = __in_dev_get_rtnl(dev);
1809                 if (in_dev == NULL)
1810                         goto errout;
1811                 devconf = &in_dev->cnf;
1812                 break;
1813         }
1814
1815         err = -ENOBUFS;
1816         skb = nlmsg_new(inet_netconf_msgsize_devconf(-1), GFP_ATOMIC);
1817         if (skb == NULL)
1818                 goto errout;
1819
1820         err = inet_netconf_fill_devconf(skb, ifindex, devconf,
1821                                         NETLINK_CB(in_skb).portid,
1822                                         nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
1823                                         -1);
1824         if (err < 0) {
1825                 /* -EMSGSIZE implies BUG in inet_netconf_msgsize_devconf() */
1826                 WARN_ON(err == -EMSGSIZE);
1827                 kfree_skb(skb);
1828                 goto errout;
1829         }
1830         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
1831 errout:
1832         return err;
1833 }
1834
1835 static int inet_netconf_dump_devconf(struct sk_buff *skb,
1836                                      struct netlink_callback *cb)
1837 {
1838         struct net *net = sock_net(skb->sk);
1839         int h, s_h;
1840         int idx, s_idx;
1841         struct net_device *dev;
1842         struct in_device *in_dev;
1843         struct hlist_head *head;
1844
1845         s_h = cb->args[0];
1846         s_idx = idx = cb->args[1];
1847
1848         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1849                 idx = 0;
1850                 head = &net->dev_index_head[h];
1851                 rcu_read_lock();
1852                 cb->seq = atomic_read(&net->ipv4.dev_addr_genid) ^
1853                           net->dev_base_seq;
1854                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
1855                         if (idx < s_idx)
1856                                 goto cont;
1857                         in_dev = __in_dev_get_rcu(dev);
1858                         if (!in_dev)
1859                                 goto cont;
1860
1861                         if (inet_netconf_fill_devconf(skb, dev->ifindex,
1862                                                       &in_dev->cnf,
1863                                                       NETLINK_CB(cb->skb).portid,
1864                                                       cb->nlh->nlmsg_seq,
1865                                                       RTM_NEWNETCONF,
1866                                                       NLM_F_MULTI,
1867                                                       -1) <= 0) {
1868                                 rcu_read_unlock();
1869                                 goto done;
1870                         }
1871                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1872 cont:
1873                         idx++;
1874                 }
1875                 rcu_read_unlock();
1876         }
1877         if (h == NETDEV_HASHENTRIES) {
1878                 if (inet_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
1879                                               net->ipv4.devconf_all,
1880                                               NETLINK_CB(cb->skb).portid,
1881                                               cb->nlh->nlmsg_seq,
1882                                               RTM_NEWNETCONF, NLM_F_MULTI,
1883                                               -1) <= 0)
1884                         goto done;
1885                 else
1886                         h++;
1887         }
1888         if (h == NETDEV_HASHENTRIES + 1) {
1889                 if (inet_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
1890                                               net->ipv4.devconf_dflt,
1891                                               NETLINK_CB(cb->skb).portid,
1892                                               cb->nlh->nlmsg_seq,
1893                                               RTM_NEWNETCONF, NLM_F_MULTI,
1894                                               -1) <= 0)
1895                         goto done;
1896                 else
1897                         h++;
1898         }
1899 done:
1900         cb->args[0] = h;
1901         cb->args[1] = idx;
1902
1903         return skb->len;
1904 }
1905
1906 #ifdef CONFIG_SYSCTL
1907
1908 static void devinet_copy_dflt_conf(struct net *net, int i)
1909 {
1910         struct net_device *dev;
1911
1912         rcu_read_lock();
1913         for_each_netdev_rcu(net, dev) {
1914                 struct in_device *in_dev;
1915
1916                 in_dev = __in_dev_get_rcu(dev);
1917                 if (in_dev && !test_bit(i, in_dev->cnf.state))
1918                         in_dev->cnf.data[i] = net->ipv4.devconf_dflt->data[i];
1919         }
1920         rcu_read_unlock();
1921 }
1922
1923 /* called with RTNL locked */
1924 static void inet_forward_change(struct net *net)
1925 {
1926         struct net_device *dev;
1927         int on = IPV4_DEVCONF_ALL(net, FORWARDING);
1928
1929         IPV4_DEVCONF_ALL(net, ACCEPT_REDIRECTS) = !on;
1930         IPV4_DEVCONF_DFLT(net, FORWARDING) = on;
1931         inet_netconf_notify_devconf(net, NETCONFA_FORWARDING,
1932                                     NETCONFA_IFINDEX_ALL,
1933                                     net->ipv4.devconf_all);
1934         inet_netconf_notify_devconf(net, NETCONFA_FORWARDING,
1935                                     NETCONFA_IFINDEX_DEFAULT,
1936                                     net->ipv4.devconf_dflt);
1937
1938         for_each_netdev(net, dev) {
1939                 struct in_device *in_dev;
1940                 if (on)
1941                         dev_disable_lro(dev);
1942                 rcu_read_lock();
1943                 in_dev = __in_dev_get_rcu(dev);
1944                 if (in_dev) {
1945                         IN_DEV_CONF_SET(in_dev, FORWARDING, on);
1946                         inet_netconf_notify_devconf(net, NETCONFA_FORWARDING,
1947                                                     dev->ifindex, &in_dev->cnf);
1948                 }
1949                 rcu_read_unlock();
1950         }
1951 }
1952
1953 static int devinet_conf_proc(struct ctl_table *ctl, int write,
1954                              void __user *buffer,
1955                              size_t *lenp, loff_t *ppos)
1956 {
1957         int old_value = *(int *)ctl->data;
1958         int ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
1959         int new_value = *(int *)ctl->data;
1960
1961         if (write) {
1962                 struct ipv4_devconf *cnf = ctl->extra1;
1963                 struct net *net = ctl->extra2;
1964                 int i = (int *)ctl->data - cnf->data;
1965
1966                 set_bit(i, cnf->state);
1967
1968                 if (cnf == net->ipv4.devconf_dflt)
1969                         devinet_copy_dflt_conf(net, i);
1970                 if (i == IPV4_DEVCONF_ACCEPT_LOCAL - 1 ||
1971                     i == IPV4_DEVCONF_ROUTE_LOCALNET - 1)
1972                         if ((new_value == 0) && (old_value != 0))
1973                                 rt_cache_flush(net);
1974                 if (i == IPV4_DEVCONF_RP_FILTER - 1 &&
1975                     new_value != old_value) {
1976                         int ifindex;
1977
1978                         if (cnf == net->ipv4.devconf_dflt)
1979                                 ifindex = NETCONFA_IFINDEX_DEFAULT;
1980                         else if (cnf == net->ipv4.devconf_all)
1981                                 ifindex = NETCONFA_IFINDEX_ALL;
1982                         else {
1983                                 struct in_device *idev =
1984                                         container_of(cnf, struct in_device,
1985                                                      cnf);
1986                                 ifindex = idev->dev->ifindex;
1987                         }
1988                         inet_netconf_notify_devconf(net, NETCONFA_RP_FILTER,
1989                                                     ifindex, cnf);
1990                 }
1991         }
1992
1993         return ret;
1994 }
1995
1996 static int devinet_sysctl_forward(struct ctl_table *ctl, int write,
1997                                   void __user *buffer,
1998                                   size_t *lenp, loff_t *ppos)
1999 {
2000         int *valp = ctl->data;
2001         int val = *valp;
2002         loff_t pos = *ppos;
2003         int ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
2004
2005         if (write && *valp != val) {
2006                 struct net *net = ctl->extra2;
2007
2008                 if (valp != &IPV4_DEVCONF_DFLT(net, FORWARDING)) {
2009                         if (!rtnl_trylock()) {
2010                                 /* Restore the original values before restarting */
2011                                 *valp = val;
2012                                 *ppos = pos;
2013                                 return restart_syscall();
2014                         }
2015                         if (valp == &IPV4_DEVCONF_ALL(net, FORWARDING)) {
2016                                 inet_forward_change(net);
2017                         } else {
2018                                 struct ipv4_devconf *cnf = ctl->extra1;
2019                                 struct in_device *idev =
2020                                         container_of(cnf, struct in_device, cnf);
2021                                 if (*valp)
2022                                         dev_disable_lro(idev->dev);
2023                                 inet_netconf_notify_devconf(net,
2024                                                             NETCONFA_FORWARDING,
2025                                                             idev->dev->ifindex,
2026                                                             cnf);
2027                         }
2028                         rtnl_unlock();
2029                         rt_cache_flush(net);
2030                 } else
2031                         inet_netconf_notify_devconf(net, NETCONFA_FORWARDING,
2032                                                     NETCONFA_IFINDEX_DEFAULT,
2033                                                     net->ipv4.devconf_dflt);
2034         }
2035
2036         return ret;
2037 }
2038
2039 static int ipv4_doint_and_flush(struct ctl_table *ctl, int write,
2040                                 void __user *buffer,
2041                                 size_t *lenp, loff_t *ppos)
2042 {
2043         int *valp = ctl->data;
2044         int val = *valp;
2045         int ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
2046         struct net *net = ctl->extra2;
2047
2048         if (write && *valp != val)
2049                 rt_cache_flush(net);
2050
2051         return ret;
2052 }
2053
2054 #define DEVINET_SYSCTL_ENTRY(attr, name, mval, proc) \
2055         { \
2056                 .procname       = name, \
2057                 .data           = ipv4_devconf.data + \
2058                                   IPV4_DEVCONF_ ## attr - 1, \
2059                 .maxlen         = sizeof(int), \
2060                 .mode           = mval, \
2061                 .proc_handler   = proc, \
2062                 .extra1         = &ipv4_devconf, \
2063         }
2064
2065 #define DEVINET_SYSCTL_RW_ENTRY(attr, name) \
2066         DEVINET_SYSCTL_ENTRY(attr, name, 0644, devinet_conf_proc)
2067
2068 #define DEVINET_SYSCTL_RO_ENTRY(attr, name) \
2069         DEVINET_SYSCTL_ENTRY(attr, name, 0444, devinet_conf_proc)
2070
2071 #define DEVINET_SYSCTL_COMPLEX_ENTRY(attr, name, proc) \
2072         DEVINET_SYSCTL_ENTRY(attr, name, 0644, proc)
2073
2074 #define DEVINET_SYSCTL_FLUSHING_ENTRY(attr, name) \
2075         DEVINET_SYSCTL_COMPLEX_ENTRY(attr, name, ipv4_doint_and_flush)
2076
2077 static struct devinet_sysctl_table {
2078         struct ctl_table_header *sysctl_header;
2079         struct ctl_table devinet_vars[__IPV4_DEVCONF_MAX];
2080 } devinet_sysctl = {
2081         .devinet_vars = {
2082                 DEVINET_SYSCTL_COMPLEX_ENTRY(FORWARDING, "forwarding",
2083                                              devinet_sysctl_forward),
2084                 DEVINET_SYSCTL_RO_ENTRY(MC_FORWARDING, "mc_forwarding"),
2085
2086                 DEVINET_SYSCTL_RW_ENTRY(ACCEPT_REDIRECTS, "accept_redirects"),
2087                 DEVINET_SYSCTL_RW_ENTRY(SECURE_REDIRECTS, "secure_redirects"),
2088                 DEVINET_SYSCTL_RW_ENTRY(SHARED_MEDIA, "shared_media"),
2089                 DEVINET_SYSCTL_RW_ENTRY(RP_FILTER, "rp_filter"),
2090                 DEVINET_SYSCTL_RW_ENTRY(SEND_REDIRECTS, "send_redirects"),
2091                 DEVINET_SYSCTL_RW_ENTRY(ACCEPT_SOURCE_ROUTE,
2092                                         "accept_source_route"),
2093                 DEVINET_SYSCTL_RW_ENTRY(ACCEPT_LOCAL, "accept_local"),
2094                 DEVINET_SYSCTL_RW_ENTRY(SRC_VMARK, "src_valid_mark"),
2095                 DEVINET_SYSCTL_RW_ENTRY(PROXY_ARP, "proxy_arp"),
2096                 DEVINET_SYSCTL_RW_ENTRY(MEDIUM_ID, "medium_id"),
2097                 DEVINET_SYSCTL_RW_ENTRY(BOOTP_RELAY, "bootp_relay"),
2098                 DEVINET_SYSCTL_RW_ENTRY(LOG_MARTIANS, "log_martians"),
2099                 DEVINET_SYSCTL_RW_ENTRY(TAG, "tag"),
2100                 DEVINET_SYSCTL_RW_ENTRY(ARPFILTER, "arp_filter"),
2101                 DEVINET_SYSCTL_RW_ENTRY(ARP_ANNOUNCE, "arp_announce"),
2102                 DEVINET_SYSCTL_RW_ENTRY(ARP_IGNORE, "arp_ignore"),
2103                 DEVINET_SYSCTL_RW_ENTRY(ARP_ACCEPT, "arp_accept"),
2104                 DEVINET_SYSCTL_RW_ENTRY(ARP_NOTIFY, "arp_notify"),
2105                 DEVINET_SYSCTL_RW_ENTRY(PROXY_ARP_PVLAN, "proxy_arp_pvlan"),
2106                 DEVINET_SYSCTL_RW_ENTRY(FORCE_IGMP_VERSION,
2107                                         "force_igmp_version"),
2108                 DEVINET_SYSCTL_RW_ENTRY(IGMPV2_UNSOLICITED_REPORT_INTERVAL,
2109                                         "igmpv2_unsolicited_report_interval"),
2110                 DEVINET_SYSCTL_RW_ENTRY(IGMPV3_UNSOLICITED_REPORT_INTERVAL,
2111                                         "igmpv3_unsolicited_report_interval"),
2112
2113                 DEVINET_SYSCTL_FLUSHING_ENTRY(NOXFRM, "disable_xfrm"),
2114                 DEVINET_SYSCTL_FLUSHING_ENTRY(NOPOLICY, "disable_policy"),
2115                 DEVINET_SYSCTL_FLUSHING_ENTRY(PROMOTE_SECONDARIES,
2116                                               "promote_secondaries"),
2117                 DEVINET_SYSCTL_FLUSHING_ENTRY(ROUTE_LOCALNET,
2118                                               "route_localnet"),
2119         },
2120 };
2121
2122 static int __devinet_sysctl_register(struct net *net, char *dev_name,
2123                                         struct ipv4_devconf *p)
2124 {
2125         int i;
2126         struct devinet_sysctl_table *t;
2127         char path[sizeof("net/ipv4/conf/") + IFNAMSIZ];
2128
2129         t = kmemdup(&devinet_sysctl, sizeof(*t), GFP_KERNEL);
2130         if (!t)
2131                 goto out;
2132
2133         for (i = 0; i < ARRAY_SIZE(t->devinet_vars) - 1; i++) {
2134                 t->devinet_vars[i].data += (char *)p - (char *)&ipv4_devconf;
2135                 t->devinet_vars[i].extra1 = p;
2136                 t->devinet_vars[i].extra2 = net;
2137         }
2138
2139         snprintf(path, sizeof(path), "net/ipv4/conf/%s", dev_name);
2140
2141         t->sysctl_header = register_net_sysctl(net, path, t->devinet_vars);
2142         if (!t->sysctl_header)
2143                 goto free;
2144
2145         p->sysctl = t;
2146         return 0;
2147
2148 free:
2149         kfree(t);
2150 out:
2151         return -ENOBUFS;
2152 }
2153
2154 static void __devinet_sysctl_unregister(struct ipv4_devconf *cnf)
2155 {
2156         struct devinet_sysctl_table *t = cnf->sysctl;
2157
2158         if (t == NULL)
2159                 return;
2160
2161         cnf->sysctl = NULL;
2162         unregister_net_sysctl_table(t->sysctl_header);
2163         kfree(t);
2164 }
2165
2166 static void devinet_sysctl_register(struct in_device *idev)
2167 {
2168         neigh_sysctl_register(idev->dev, idev->arp_parms, NULL);
2169         __devinet_sysctl_register(dev_net(idev->dev), idev->dev->name,
2170                                         &idev->cnf);
2171 }
2172
2173 static void devinet_sysctl_unregister(struct in_device *idev)
2174 {
2175         __devinet_sysctl_unregister(&idev->cnf);
2176         neigh_sysctl_unregister(idev->arp_parms);
2177 }
2178
2179 static struct ctl_table ctl_forward_entry[] = {
2180         {
2181                 .procname       = "ip_forward",
2182                 .data           = &ipv4_devconf.data[
2183                                         IPV4_DEVCONF_FORWARDING - 1],
2184                 .maxlen         = sizeof(int),
2185                 .mode           = 0644,
2186                 .proc_handler   = devinet_sysctl_forward,
2187                 .extra1         = &ipv4_devconf,
2188                 .extra2         = &init_net,
2189         },
2190         { },
2191 };
2192 #endif
2193
2194 static __net_init int devinet_init_net(struct net *net)
2195 {
2196         int err;
2197         struct ipv4_devconf *all, *dflt;
2198 #ifdef CONFIG_SYSCTL
2199         struct ctl_table *tbl = ctl_forward_entry;
2200         struct ctl_table_header *forw_hdr;
2201 #endif
2202
2203         err = -ENOMEM;
2204         all = &ipv4_devconf;
2205         dflt = &ipv4_devconf_dflt;
2206
2207         if (!net_eq(net, &init_net)) {
2208                 all = kmemdup(all, sizeof(ipv4_devconf), GFP_KERNEL);
2209                 if (all == NULL)
2210                         goto err_alloc_all;
2211
2212                 dflt = kmemdup(dflt, sizeof(ipv4_devconf_dflt), GFP_KERNEL);
2213                 if (dflt == NULL)
2214                         goto err_alloc_dflt;
2215
2216 #ifdef CONFIG_SYSCTL
2217                 tbl = kmemdup(tbl, sizeof(ctl_forward_entry), GFP_KERNEL);
2218                 if (tbl == NULL)
2219                         goto err_alloc_ctl;
2220
2221                 tbl[0].data = &all->data[IPV4_DEVCONF_FORWARDING - 1];
2222                 tbl[0].extra1 = all;
2223                 tbl[0].extra2 = net;
2224 #endif
2225         }
2226
2227 #ifdef CONFIG_SYSCTL
2228         err = __devinet_sysctl_register(net, "all", all);
2229         if (err < 0)
2230                 goto err_reg_all;
2231
2232         err = __devinet_sysctl_register(net, "default", dflt);
2233         if (err < 0)
2234                 goto err_reg_dflt;
2235
2236         err = -ENOMEM;
2237         forw_hdr = register_net_sysctl(net, "net/ipv4", tbl);
2238         if (forw_hdr == NULL)
2239                 goto err_reg_ctl;
2240         net->ipv4.forw_hdr = forw_hdr;
2241 #endif
2242
2243         net->ipv4.devconf_all = all;
2244         net->ipv4.devconf_dflt = dflt;
2245         return 0;
2246
2247 #ifdef CONFIG_SYSCTL
2248 err_reg_ctl:
2249         __devinet_sysctl_unregister(dflt);
2250 err_reg_dflt:
2251         __devinet_sysctl_unregister(all);
2252 err_reg_all:
2253         if (tbl != ctl_forward_entry)
2254                 kfree(tbl);
2255 err_alloc_ctl:
2256 #endif
2257         if (dflt != &ipv4_devconf_dflt)
2258                 kfree(dflt);
2259 err_alloc_dflt:
2260         if (all != &ipv4_devconf)
2261                 kfree(all);
2262 err_alloc_all:
2263         return err;
2264 }
2265
2266 static __net_exit void devinet_exit_net(struct net *net)
2267 {
2268 #ifdef CONFIG_SYSCTL
2269         struct ctl_table *tbl;
2270
2271         tbl = net->ipv4.forw_hdr->ctl_table_arg;
2272         unregister_net_sysctl_table(net->ipv4.forw_hdr);
2273         __devinet_sysctl_unregister(net->ipv4.devconf_dflt);
2274         __devinet_sysctl_unregister(net->ipv4.devconf_all);
2275         kfree(tbl);
2276 #endif
2277         kfree(net->ipv4.devconf_dflt);
2278         kfree(net->ipv4.devconf_all);
2279 }
2280
2281 static __net_initdata struct pernet_operations devinet_ops = {
2282         .init = devinet_init_net,
2283         .exit = devinet_exit_net,
2284 };
2285
2286 static struct rtnl_af_ops inet_af_ops = {
2287         .family           = AF_INET,
2288         .fill_link_af     = inet_fill_link_af,
2289         .get_link_af_size = inet_get_link_af_size,
2290         .validate_link_af = inet_validate_link_af,
2291         .set_link_af      = inet_set_link_af,
2292 };
2293
2294 void __init devinet_init(void)
2295 {
2296         int i;
2297
2298         for (i = 0; i < IN4_ADDR_HSIZE; i++)
2299                 INIT_HLIST_HEAD(&inet_addr_lst[i]);
2300
2301         register_pernet_subsys(&devinet_ops);
2302
2303         register_gifconf(PF_INET, inet_gifconf);
2304         register_netdevice_notifier(&ip_netdev_notifier);
2305
2306         schedule_delayed_work(&check_lifetime_work, 0);
2307
2308         rtnl_af_register(&inet_af_ops);
2309
2310         rtnl_register(PF_INET, RTM_NEWADDR, inet_rtm_newaddr, NULL, NULL);
2311         rtnl_register(PF_INET, RTM_DELADDR, inet_rtm_deladdr, NULL, NULL);
2312         rtnl_register(PF_INET, RTM_GETADDR, NULL, inet_dump_ifaddr, NULL);
2313         rtnl_register(PF_INET, RTM_GETNETCONF, inet_netconf_get_devconf,
2314                       inet_netconf_dump_devconf, NULL);
2315 }
2316