net/ipv4/devinet.c

   1 // SPDX-License-Identifier: GPL-2.0-or-later
   2 /*
   3  *      NET3    IP device support routines.
   4  *
   5  *      Derived from the IP parts of dev.c 1.0.19
   6  *              Authors:        Ross Biro
   7  *                              Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
   8  *                              Mark Evans, <evansmp@uhura.aston.ac.uk>
   9  *
  10  *      Additional Authors:
  11  *              Alan Cox, <gw4pts@gw4pts.ampr.org>
  12  *              Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
  13  *
  14  *      Changes:
  15  *              Alexey Kuznetsov:       pa_* fields are replaced with ifaddr
  16  *                                      lists.
  17  *              Cyrus Durgin:           updated for kmod
  18  *              Matthias Andree:        in devinet_ioctl, compare label and
  19  *                                      address (4.4BSD alias style support),
  20  *                                      fall back to comparing just the label
  21  *                                      if no match found.
  22  */
  23
  24
  25 #include <linux/uaccess.h>
  26 #include <linux/bitops.h>
  27 #include <linux/capability.h>
  28 #include <linux/module.h>
  29 #include <linux/types.h>
  30 #include <linux/kernel.h>
  31 #include <linux/sched/signal.h>
  32 #include <linux/string.h>
  33 #include <linux/mm.h>
  34 #include <linux/socket.h>
  35 #include <linux/sockios.h>
  36 #include <linux/in.h>
  37 #include <linux/errno.h>
  38 #include <linux/interrupt.h>
  39 #include <linux/if_addr.h>
  40 #include <linux/if_ether.h>
  41 #include <linux/inet.h>
  42 #include <linux/netdevice.h>
  43 #include <linux/etherdevice.h>
  44 #include <linux/skbuff.h>
  45 #include <linux/init.h>
  46 #include <linux/notifier.h>
  47 #include <linux/inetdevice.h>
  48 #include <linux/igmp.h>
  49 #include <linux/slab.h>
  50 #include <linux/hash.h>
  51 #ifdef CONFIG_SYSCTL
  52 #include <linux/sysctl.h>
  53 #endif
  54 #include <linux/kmod.h>
  55 #include <linux/netconf.h>
  56
  57 #include <net/arp.h>
  58 #include <net/ip.h>
  59 #include <net/route.h>
  60 #include <net/ip_fib.h>
  61 #include <net/rtnetlink.h>
  62 #include <net/net_namespace.h>
  63 #include <net/addrconf.h>
  64
  65 #define IPV6ONLY_FLAGS  \
  66                 (IFA_F_NODAD | IFA_F_OPTIMISTIC | IFA_F_DADFAILED | \
  67                  IFA_F_HOMEADDRESS | IFA_F_TENTATIVE | \
  68                  IFA_F_MANAGETEMPADDR | IFA_F_STABLE_PRIVACY)
  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         [IFA_RT_PRIORITY]       = { .type = NLA_U32 },
 104         [IFA_TARGET_NETNSID]    = { .type = NLA_S32 },
 105 };
 106
 107 struct inet_fill_args {
 108         u32 portid;
 109         u32 seq;
 110         int event;
 111         unsigned int flags;
 112         int netnsid;
 113         int ifindex;
 114 };
 115
 116 #define IN4_ADDR_HSIZE_SHIFT    8
 117 #define IN4_ADDR_HSIZE          (1U << IN4_ADDR_HSIZE_SHIFT)
 118
 119 static struct hlist_head inet_addr_lst[IN4_ADDR_HSIZE];
 120
 121 static u32 inet_addr_hash(const struct net *net, __be32 addr)
 122 {
 123         u32 val = (__force u32) addr ^ net_hash_mix(net);
 124
 125         return hash_32(val, IN4_ADDR_HSIZE_SHIFT);
 126 }
 127
 128 static void inet_hash_insert(struct net *net, struct in_ifaddr *ifa)
 129 {
 130         u32 hash = inet_addr_hash(net, ifa->ifa_local);
 131
 132         ASSERT_RTNL();
 133         hlist_add_head_rcu(&ifa->hash, &inet_addr_lst[hash]);
 134 }
 135
 136 static void inet_hash_remove(struct in_ifaddr *ifa)
 137 {
 138         ASSERT_RTNL();
 139         hlist_del_init_rcu(&ifa->hash);
 140 }
 141
 142 /**
 143  * __ip_dev_find - find the first device with a given source address.
 144  * @net: the net namespace
 145  * @addr: the source address
 146  * @devref: if true, take a reference on the found device
 147  *
 148  * If a caller uses devref=false, it should be protected by RCU, or RTNL
 149  */
 150 struct net_device *__ip_dev_find(struct net *net, __be32 addr, bool devref)
 151 {
 152         struct net_device *result = NULL;
 153         struct in_ifaddr *ifa;
 154
 155         rcu_read_lock();
 156         ifa = inet_lookup_ifaddr_rcu(net, addr);
 157         if (!ifa) {
 158                 struct flowi4 fl4 = { .daddr = addr };
 159                 struct fib_result res = { 0 };
 160                 struct fib_table *local;
 161
 162                 /* Fallback to FIB local table so that communication
 163                  * over loopback subnets work.
 164                  */
 165                 local = fib_get_table(net, RT_TABLE_LOCAL);
 166                 if (local &&
 167                     !fib_table_lookup(local, &fl4, &res, FIB_LOOKUP_NOREF) &&
 168                     res.type == RTN_LOCAL)
 169                         result = FIB_RES_DEV(res);
 170         } else {
 171                 result = ifa->ifa_dev->dev;
 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 /* called under RCU lock */
 181 struct in_ifaddr *inet_lookup_ifaddr_rcu(struct net *net, __be32 addr)
 182 {
 183         u32 hash = inet_addr_hash(net, addr);
 184         struct in_ifaddr *ifa;
 185
 186         hlist_for_each_entry_rcu(ifa, &inet_addr_lst[hash], hash)
 187                 if (ifa->ifa_local == addr &&
 188                     net_eq(dev_net(ifa->ifa_dev->dev), net))
 189                         return ifa;
 190
 191         return NULL;
 192 }
 193
 194 static void rtmsg_ifa(int event, struct in_ifaddr *, struct nlmsghdr *, u32);
 195
 196 static BLOCKING_NOTIFIER_HEAD(inetaddr_chain);
 197 static BLOCKING_NOTIFIER_HEAD(inetaddr_validator_chain);
 198 static void inet_del_ifa(struct in_device *in_dev,
 199                          struct in_ifaddr __rcu **ifap,
 200                          int destroy);
 201 #ifdef CONFIG_SYSCTL
 202 static int devinet_sysctl_register(struct in_device *idev);
 203 static void devinet_sysctl_unregister(struct in_device *idev);
 204 #else
 205 static int devinet_sysctl_register(struct in_device *idev)
 206 {
 207         return 0;
 208 }
 209 static void devinet_sysctl_unregister(struct in_device *idev)
 210 {
 211 }
 212 #endif
 213
 214 /* Locks all the inet devices. */
 215
 216 static struct in_ifaddr *inet_alloc_ifa(void)
 217 {
 218         return kzalloc(sizeof(struct in_ifaddr), GFP_KERNEL);
 219 }
 220
 221 static void inet_rcu_free_ifa(struct rcu_head *head)
 222 {
 223         struct in_ifaddr *ifa = container_of(head, struct in_ifaddr, rcu_head);
 224         if (ifa->ifa_dev)
 225                 in_dev_put(ifa->ifa_dev);
 226         kfree(ifa);
 227 }
 228
 229 static void inet_free_ifa(struct in_ifaddr *ifa)
 230 {
 231         call_rcu(&ifa->rcu_head, inet_rcu_free_ifa);
 232 }
 233
 234 void in_dev_finish_destroy(struct in_device *idev)
 235 {
 236         struct net_device *dev = idev->dev;
 237
 238         WARN_ON(idev->ifa_list);
 239         WARN_ON(idev->mc_list);
 240         kfree(rcu_dereference_protected(idev->mc_hash, 1));
 241 #ifdef NET_REFCNT_DEBUG
 242         pr_debug("%s: %p=%s\n", __func__, idev, dev ? dev->name : "NIL");
 243 #endif
 244         dev_put(dev);
 245         if (!idev->dead)
 246                 pr_err("Freeing alive in_device %p\n", idev);
 247         else
 248                 kfree(idev);
 249 }
 250 EXPORT_SYMBOL(in_dev_finish_destroy);
 251
 252 static struct in_device *inetdev_init(struct net_device *dev)
 253 {
 254         struct in_device *in_dev;
 255         int err = -ENOMEM;
 256
 257         ASSERT_RTNL();
 258
 259         in_dev = kzalloc(sizeof(*in_dev), GFP_KERNEL);
 260         if (!in_dev)
 261                 goto out;
 262         memcpy(&in_dev->cnf, dev_net(dev)->ipv4.devconf_dflt,
 263                         sizeof(in_dev->cnf));
 264         in_dev->cnf.sysctl = NULL;
 265         in_dev->dev = dev;
 266         in_dev->arp_parms = neigh_parms_alloc(dev, &arp_tbl);
 267         if (!in_dev->arp_parms)
 268                 goto out_kfree;
 269         if (IPV4_DEVCONF(in_dev->cnf, FORWARDING))
 270                 dev_disable_lro(dev);
 271         /* Reference in_dev->dev */
 272         dev_hold(dev);
 273         /* Account for reference dev->ip_ptr (below) */
 274         refcount_set(&in_dev->refcnt, 1);
 275
 276         err = devinet_sysctl_register(in_dev);
 277         if (err) {
 278                 in_dev->dead = 1;
 279                 in_dev_put(in_dev);
 280                 in_dev = NULL;
 281                 goto out;
 282         }
 283         ip_mc_init_dev(in_dev);
 284         if (dev->flags & IFF_UP)
 285                 ip_mc_up(in_dev);
 286
 287         /* we can receive as soon as ip_ptr is set -- do this last */
 288         rcu_assign_pointer(dev->ip_ptr, in_dev);
 289 out:
 290         return in_dev ?: ERR_PTR(err);
 291 out_kfree:
 292         kfree(in_dev);
 293         in_dev = NULL;
 294         goto out;
 295 }
 296
 297 static void in_dev_rcu_put(struct rcu_head *head)
 298 {
 299         struct in_device *idev = container_of(head, struct in_device, rcu_head);
 300         in_dev_put(idev);
 301 }
 302
 303 static void inetdev_destroy(struct in_device *in_dev)
 304 {
 305         struct net_device *dev;
 306         struct in_ifaddr *ifa;
 307
 308         ASSERT_RTNL();
 309
 310         dev = in_dev->dev;
 311
 312         in_dev->dead = 1;
 313
 314         ip_mc_destroy_dev(in_dev);
 315
 316         while ((ifa = rtnl_dereference(in_dev->ifa_list)) != NULL) {
 317                 inet_del_ifa(in_dev, &in_dev->ifa_list, 0);
 318                 inet_free_ifa(ifa);
 319         }
 320
 321         RCU_INIT_POINTER(dev->ip_ptr, NULL);
 322
 323         devinet_sysctl_unregister(in_dev);
 324         neigh_parms_release(&arp_tbl, in_dev->arp_parms);
 325         arp_ifdown(dev);
 326
 327         call_rcu(&in_dev->rcu_head, in_dev_rcu_put);
 328 }
 329
 330 int inet_addr_onlink(struct in_device *in_dev, __be32 a, __be32 b)
 331 {
 332         const struct in_ifaddr *ifa;
 333
 334         rcu_read_lock();
 335         in_dev_for_each_ifa_rcu(ifa, in_dev) {
 336                 if (inet_ifa_match(a, ifa)) {
 337                         if (!b || inet_ifa_match(b, ifa)) {
 338                                 rcu_read_unlock();
 339                                 return 1;
 340                         }
 341                 }
 342         }
 343         rcu_read_unlock();
 344         return 0;
 345 }
 346
 347 static void __inet_del_ifa(struct in_device *in_dev,
 348                            struct in_ifaddr __rcu **ifap,
 349                            int destroy, struct nlmsghdr *nlh, u32 portid)
 350 {
 351         struct in_ifaddr *promote = NULL;
 352         struct in_ifaddr *ifa, *ifa1;
 353         struct in_ifaddr *last_prim;
 354         struct in_ifaddr *prev_prom = NULL;
 355         int do_promote = IN_DEV_PROMOTE_SECONDARIES(in_dev);
 356
 357         ASSERT_RTNL();
 358
 359         ifa1 = rtnl_dereference(*ifap);
 360         last_prim = rtnl_dereference(in_dev->ifa_list);
 361         if (in_dev->dead)
 362                 goto no_promotions;
 363
 364         /* 1. Deleting primary ifaddr forces deletion all secondaries
 365          * unless alias promotion is set
 366          **/
 367
 368         if (!(ifa1->ifa_flags & IFA_F_SECONDARY)) {
 369                 struct in_ifaddr __rcu **ifap1 = &ifa1->ifa_next;
 370
 371                 while ((ifa = rtnl_dereference(*ifap1)) != NULL) {
 372                         if (!(ifa->ifa_flags & IFA_F_SECONDARY) &&
 373                             ifa1->ifa_scope <= ifa->ifa_scope)
 374                                 last_prim = ifa;
 375
 376                         if (!(ifa->ifa_flags & IFA_F_SECONDARY) ||
 377                             ifa1->ifa_mask != ifa->ifa_mask ||
 378                             !inet_ifa_match(ifa1->ifa_address, ifa)) {
 379                                 ifap1 = &ifa->ifa_next;
 380                                 prev_prom = ifa;
 381                                 continue;
 382                         }
 383
 384                         if (!do_promote) {
 385                                 inet_hash_remove(ifa);
 386                                 *ifap1 = ifa->ifa_next;
 387
 388                                 rtmsg_ifa(RTM_DELADDR, ifa, nlh, portid);
 389                                 blocking_notifier_call_chain(&inetaddr_chain,
 390                                                 NETDEV_DOWN, ifa);
 391                                 inet_free_ifa(ifa);
 392                         } else {
 393                                 promote = ifa;
 394                                 break;
 395                         }
 396                 }
 397         }
 398
 399         /* On promotion all secondaries from subnet are changing
 400          * the primary IP, we must remove all their routes silently
 401          * and later to add them back with new prefsrc. Do this
 402          * while all addresses are on the device list.
 403          */
 404         for (ifa = promote; ifa; ifa = rtnl_dereference(ifa->ifa_next)) {
 405                 if (ifa1->ifa_mask == ifa->ifa_mask &&
 406                     inet_ifa_match(ifa1->ifa_address, ifa))
 407                         fib_del_ifaddr(ifa, ifa1);
 408         }
 409
 410 no_promotions:
 411         /* 2. Unlink it */
 412
 413         *ifap = ifa1->ifa_next;
 414         inet_hash_remove(ifa1);
 415
 416         /* 3. Announce address deletion */
 417
 418         /* Send message first, then call notifier.
 419            At first sight, FIB update triggered by notifier
 420            will refer to already deleted ifaddr, that could confuse
 421            netlink listeners. It is not true: look, gated sees
 422            that route deleted and if it still thinks that ifaddr
 423            is valid, it will try to restore deleted routes... Grr.
 424            So that, this order is correct.
 425          */
 426         rtmsg_ifa(RTM_DELADDR, ifa1, nlh, portid);
 427         blocking_notifier_call_chain(&inetaddr_chain, NETDEV_DOWN, ifa1);
 428
 429         if (promote) {
 430                 struct in_ifaddr *next_sec;
 431
 432                 next_sec = rtnl_dereference(promote->ifa_next);
 433                 if (prev_prom) {
 434                         struct in_ifaddr *last_sec;
 435
 436                         rcu_assign_pointer(prev_prom->ifa_next, next_sec);
 437
 438                         last_sec = rtnl_dereference(last_prim->ifa_next);
 439                         rcu_assign_pointer(promote->ifa_next, last_sec);
 440                         rcu_assign_pointer(last_prim->ifa_next, promote);
 441                 }
 442
 443                 promote->ifa_flags &= ~IFA_F_SECONDARY;
 444                 rtmsg_ifa(RTM_NEWADDR, promote, nlh, portid);
 445                 blocking_notifier_call_chain(&inetaddr_chain,
 446                                 NETDEV_UP, promote);
 447                 for (ifa = next_sec; ifa;
 448                      ifa = rtnl_dereference(ifa->ifa_next)) {
 449                         if (ifa1->ifa_mask != ifa->ifa_mask ||
 450                             !inet_ifa_match(ifa1->ifa_address, ifa))
 451                                         continue;
 452                         fib_add_ifaddr(ifa);
 453                 }
 454
 455         }
 456         if (destroy)
 457                 inet_free_ifa(ifa1);
 458 }
 459
 460 static void inet_del_ifa(struct in_device *in_dev,
 461                          struct in_ifaddr __rcu **ifap,
 462                          int destroy)
 463 {
 464         __inet_del_ifa(in_dev, ifap, destroy, NULL, 0);
 465 }
 466
 467 static void check_lifetime(struct work_struct *work);
 468
 469 static DECLARE_DELAYED_WORK(check_lifetime_work, check_lifetime);
 470
 471 static int __inet_insert_ifa(struct in_ifaddr *ifa, struct nlmsghdr *nlh,
 472                              u32 portid, struct netlink_ext_ack *extack)
 473 {
 474         struct in_ifaddr __rcu **last_primary, **ifap;
 475         struct in_device *in_dev = ifa->ifa_dev;
 476         struct in_validator_info ivi;
 477         struct in_ifaddr *ifa1;
 478         int ret;
 479
 480         ASSERT_RTNL();
 481
 482         if (!ifa->ifa_local) {
 483                 inet_free_ifa(ifa);
 484                 return 0;
 485         }
 486
 487         ifa->ifa_flags &= ~IFA_F_SECONDARY;
 488         last_primary = &in_dev->ifa_list;
 489
 490         /* Don't set IPv6 only flags to IPv4 addresses */
 491         ifa->ifa_flags &= ~IPV6ONLY_FLAGS;
 492
 493         ifap = &in_dev->ifa_list;
 494         ifa1 = rtnl_dereference(*ifap);
 495
 496         while (ifa1) {
 497                 if (!(ifa1->ifa_flags & IFA_F_SECONDARY) &&
 498                     ifa->ifa_scope <= ifa1->ifa_scope)
 499                         last_primary = &ifa1->ifa_next;
 500                 if (ifa1->ifa_mask == ifa->ifa_mask &&
 501                     inet_ifa_match(ifa1->ifa_address, ifa)) {
 502                         if (ifa1->ifa_local == ifa->ifa_local) {
 503                                 inet_free_ifa(ifa);
 504                                 return -EEXIST;
 505                         }
 506                         if (ifa1->ifa_scope != ifa->ifa_scope) {
 507                                 inet_free_ifa(ifa);
 508                                 return -EINVAL;
 509                         }
 510                         ifa->ifa_flags |= IFA_F_SECONDARY;
 511                 }
 512
 513                 ifap = &ifa1->ifa_next;
 514                 ifa1 = rtnl_dereference(*ifap);
 515         }
 516
 517         /* Allow any devices that wish to register ifaddr validtors to weigh
 518          * in now, before changes are committed.  The rntl lock is serializing
 519          * access here, so the state should not change between a validator call
 520          * and a final notify on commit.  This isn't invoked on promotion under
 521          * the assumption that validators are checking the address itself, and
 522          * not the flags.
 523          */
 524         ivi.ivi_addr = ifa->ifa_address;
 525         ivi.ivi_dev = ifa->ifa_dev;
 526         ivi.extack = extack;
 527         ret = blocking_notifier_call_chain(&inetaddr_validator_chain,
 528                                            NETDEV_UP, &ivi);
 529         ret = notifier_to_errno(ret);
 530         if (ret) {
 531                 inet_free_ifa(ifa);
 532                 return ret;
 533         }
 534
 535         if (!(ifa->ifa_flags & IFA_F_SECONDARY)) {
 536                 prandom_seed((__force u32) ifa->ifa_local);
 537                 ifap = last_primary;
 538         }
 539
 540         rcu_assign_pointer(ifa->ifa_next, *ifap);
 541         rcu_assign_pointer(*ifap, ifa);
 542
 543         inet_hash_insert(dev_net(in_dev->dev), ifa);
 544
 545         cancel_delayed_work(&check_lifetime_work);
 546         queue_delayed_work(system_power_efficient_wq, &check_lifetime_work, 0);
 547
 548         /* Send message first, then call notifier.
 549            Notifier will trigger FIB update, so that
 550            listeners of netlink will know about new ifaddr */
 551         rtmsg_ifa(RTM_NEWADDR, ifa, nlh, portid);
 552         blocking_notifier_call_chain(&inetaddr_chain, NETDEV_UP, ifa);
 553
 554         return 0;
 555 }
 556
 557 static int inet_insert_ifa(struct in_ifaddr *ifa)
 558 {
 559         return __inet_insert_ifa(ifa, NULL, 0, NULL);
 560 }
 561
 562 static int inet_set_ifa(struct net_device *dev, struct in_ifaddr *ifa)
 563 {
 564         struct in_device *in_dev = __in_dev_get_rtnl(dev);
 565
 566         ASSERT_RTNL();
 567
 568         if (!in_dev) {
 569                 inet_free_ifa(ifa);
 570                 return -ENOBUFS;
 571         }
 572         ipv4_devconf_setall(in_dev);
 573         neigh_parms_data_state_setall(in_dev->arp_parms);
 574         if (ifa->ifa_dev != in_dev) {
 575                 WARN_ON(ifa->ifa_dev);
 576                 in_dev_hold(in_dev);
 577                 ifa->ifa_dev = in_dev;
 578         }
 579         if (ipv4_is_loopback(ifa->ifa_local))
 580                 ifa->ifa_scope = RT_SCOPE_HOST;
 581         return inet_insert_ifa(ifa);
 582 }
 583
 584 /* Caller must hold RCU or RTNL :
 585  * We dont take a reference on found in_device
 586  */
 587 struct in_device *inetdev_by_index(struct net *net, int ifindex)
 588 {
 589         struct net_device *dev;
 590         struct in_device *in_dev = NULL;
 591
 592         rcu_read_lock();
 593         dev = dev_get_by_index_rcu(net, ifindex);
 594         if (dev)
 595                 in_dev = rcu_dereference_rtnl(dev->ip_ptr);
 596         rcu_read_unlock();
 597         return in_dev;
 598 }
 599 EXPORT_SYMBOL(inetdev_by_index);
 600
 601 /* Called only from RTNL semaphored context. No locks. */
 602
 603 struct in_ifaddr *inet_ifa_byprefix(struct in_device *in_dev, __be32 prefix,
 604                                     __be32 mask)
 605 {
 606         struct in_ifaddr *ifa;
 607
 608         ASSERT_RTNL();
 609
 610         in_dev_for_each_ifa_rtnl(ifa, in_dev) {
 611                 if (ifa->ifa_mask == mask && inet_ifa_match(prefix, ifa))
 612                         return ifa;
 613         }
 614         return NULL;
 615 }
 616
 617 static int ip_mc_config(struct sock *sk, bool join, const struct in_ifaddr *ifa)
 618 {
 619         struct ip_mreqn mreq = {
 620                 .imr_multiaddr.s_addr = ifa->ifa_address,
 621                 .imr_ifindex = ifa->ifa_dev->dev->ifindex,
 622         };
 623         int ret;
 624
 625         ASSERT_RTNL();
 626
 627         lock_sock(sk);
 628         if (join)
 629                 ret = ip_mc_join_group(sk, &mreq);
 630         else
 631                 ret = ip_mc_leave_group(sk, &mreq);
 632         release_sock(sk);
 633
 634         return ret;
 635 }
 636
 637 static int inet_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
 638                             struct netlink_ext_ack *extack)
 639 {
 640         struct net *net = sock_net(skb->sk);
 641         struct in_ifaddr __rcu **ifap;
 642         struct nlattr *tb[IFA_MAX+1];
 643         struct in_device *in_dev;
 644         struct ifaddrmsg *ifm;
 645         struct in_ifaddr *ifa;
 646
 647         int err = -EINVAL;
 648
 649         ASSERT_RTNL();
 650
 651         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
 652                                      ifa_ipv4_policy, extack);
 653         if (err < 0)
 654                 goto errout;
 655
 656         ifm = nlmsg_data(nlh);
 657         in_dev = inetdev_by_index(net, ifm->ifa_index);
 658         if (!in_dev) {
 659                 err = -ENODEV;
 660                 goto errout;
 661         }
 662
 663         for (ifap = &in_dev->ifa_list; (ifa = rtnl_dereference(*ifap)) != NULL;
 664              ifap = &ifa->ifa_next) {
 665                 if (tb[IFA_LOCAL] &&
 666                     ifa->ifa_local != nla_get_in_addr(tb[IFA_LOCAL]))
 667                         continue;
 668
 669                 if (tb[IFA_LABEL] && nla_strcmp(tb[IFA_LABEL], ifa->ifa_label))
 670                         continue;
 671
 672                 if (tb[IFA_ADDRESS] &&
 673                     (ifm->ifa_prefixlen != ifa->ifa_prefixlen ||
 674                     !inet_ifa_match(nla_get_in_addr(tb[IFA_ADDRESS]), ifa)))
 675                         continue;
 676
 677                 if (ipv4_is_multicast(ifa->ifa_address))
 678                         ip_mc_config(net->ipv4.mc_autojoin_sk, false, ifa);
 679                 __inet_del_ifa(in_dev, ifap, 1, nlh, NETLINK_CB(skb).portid);
 680                 return 0;
 681         }
 682
 683         err = -EADDRNOTAVAIL;
 684 errout:
 685         return err;
 686 }
 687
 688 #define INFINITY_LIFE_TIME      0xFFFFFFFF
 689
 690 static void check_lifetime(struct work_struct *work)
 691 {
 692         unsigned long now, next, next_sec, next_sched;
 693         struct in_ifaddr *ifa;
 694         struct hlist_node *n;
 695         int i;
 696
 697         now = jiffies;
 698         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
 699
 700         for (i = 0; i < IN4_ADDR_HSIZE; i++) {
 701                 bool change_needed = false;
 702
 703                 rcu_read_lock();
 704                 hlist_for_each_entry_rcu(ifa, &inet_addr_lst[i], hash) {
 705                         unsigned long age;
 706
 707                         if (ifa->ifa_flags & IFA_F_PERMANENT)
 708                                 continue;
 709
 710                         /* We try to batch several events at once. */
 711                         age = (now - ifa->ifa_tstamp +
 712                                ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
 713
 714                         if (ifa->ifa_valid_lft != INFINITY_LIFE_TIME &&
 715                             age >= ifa->ifa_valid_lft) {
 716                                 change_needed = true;
 717                         } else if (ifa->ifa_preferred_lft ==
 718                                    INFINITY_LIFE_TIME) {
 719                                 continue;
 720                         } else if (age >= ifa->ifa_preferred_lft) {
 721                                 if (time_before(ifa->ifa_tstamp +
 722                                                 ifa->ifa_valid_lft * HZ, next))
 723                                         next = ifa->ifa_tstamp +
 724                                                ifa->ifa_valid_lft * HZ;
 725
 726                                 if (!(ifa->ifa_flags & IFA_F_DEPRECATED))
 727                                         change_needed = true;
 728                         } else if (time_before(ifa->ifa_tstamp +
 729                                                ifa->ifa_preferred_lft * HZ,
 730                                                next)) {
 731                                 next = ifa->ifa_tstamp +
 732                                        ifa->ifa_preferred_lft * HZ;
 733                         }
 734                 }
 735                 rcu_read_unlock();
 736                 if (!change_needed)
 737                         continue;
 738                 rtnl_lock();
 739                 hlist_for_each_entry_safe(ifa, n, &inet_addr_lst[i], hash) {
 740                         unsigned long age;
 741
 742                         if (ifa->ifa_flags & IFA_F_PERMANENT)
 743                                 continue;
 744
 745                         /* We try to batch several events at once. */
 746                         age = (now - ifa->ifa_tstamp +
 747                                ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
 748
 749                         if (ifa->ifa_valid_lft != INFINITY_LIFE_TIME &&
 750                             age >= ifa->ifa_valid_lft) {
 751                                 struct in_ifaddr __rcu **ifap;
 752                                 struct in_ifaddr *tmp;
 753
 754                                 ifap = &ifa->ifa_dev->ifa_list;
 755                                 tmp = rtnl_dereference(*ifap);
 756                                 while (tmp) {
 757                                         if (tmp == ifa) {
 758                                                 inet_del_ifa(ifa->ifa_dev,
 759                                                              ifap, 1);
 760                                                 break;
 761                                         }
 762                                         ifap = &tmp->ifa_next;
 763                                         tmp = rtnl_dereference(*ifap);
 764                                 }
 765                         } else if (ifa->ifa_preferred_lft !=
 766                                    INFINITY_LIFE_TIME &&
 767                                    age >= ifa->ifa_preferred_lft &&
 768                                    !(ifa->ifa_flags & IFA_F_DEPRECATED)) {
 769                                 ifa->ifa_flags |= IFA_F_DEPRECATED;
 770                                 rtmsg_ifa(RTM_NEWADDR, ifa, NULL, 0);
 771                         }
 772                 }
 773                 rtnl_unlock();
 774         }
 775
 776         next_sec = round_jiffies_up(next);
 777         next_sched = next;
 778
 779         /* If rounded timeout is accurate enough, accept it. */
 780         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
 781                 next_sched = next_sec;
 782
 783         now = jiffies;
 784         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
 785         if (time_before(next_sched, now + ADDRCONF_TIMER_FUZZ_MAX))
 786                 next_sched = now + ADDRCONF_TIMER_FUZZ_MAX;
 787
 788         queue_delayed_work(system_power_efficient_wq, &check_lifetime_work,
 789                         next_sched - now);
 790 }
 791
 792 static void set_ifa_lifetime(struct in_ifaddr *ifa, __u32 valid_lft,
 793                              __u32 prefered_lft)
 794 {
 795         unsigned long timeout;
 796
 797         ifa->ifa_flags &= ~(IFA_F_PERMANENT | IFA_F_DEPRECATED);
 798
 799         timeout = addrconf_timeout_fixup(valid_lft, HZ);
 800         if (addrconf_finite_timeout(timeout))
 801                 ifa->ifa_valid_lft = timeout;
 802         else
 803                 ifa->ifa_flags |= IFA_F_PERMANENT;
 804
 805         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
 806         if (addrconf_finite_timeout(timeout)) {
 807                 if (timeout == 0)
 808                         ifa->ifa_flags |= IFA_F_DEPRECATED;
 809                 ifa->ifa_preferred_lft = timeout;
 810         }
 811         ifa->ifa_tstamp = jiffies;
 812         if (!ifa->ifa_cstamp)
 813                 ifa->ifa_cstamp = ifa->ifa_tstamp;
 814 }
 815
 816 static struct in_ifaddr *rtm_to_ifaddr(struct net *net, struct nlmsghdr *nlh,
 817                                        __u32 *pvalid_lft, __u32 *pprefered_lft,
 818                                        struct netlink_ext_ack *extack)
 819 {
 820         struct nlattr *tb[IFA_MAX+1];
 821         struct in_ifaddr *ifa;
 822         struct ifaddrmsg *ifm;
 823         struct net_device *dev;
 824         struct in_device *in_dev;
 825         int err;
 826
 827         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
 828                                      ifa_ipv4_policy, extack);
 829         if (err < 0)
 830                 goto errout;
 831
 832         ifm = nlmsg_data(nlh);
 833         err = -EINVAL;
 834         if (ifm->ifa_prefixlen > 32 || !tb[IFA_LOCAL])
 835                 goto errout;
 836
 837         dev = __dev_get_by_index(net, ifm->ifa_index);
 838         err = -ENODEV;
 839         if (!dev)
 840                 goto errout;
 841
 842         in_dev = __in_dev_get_rtnl(dev);
 843         err = -ENOBUFS;
 844         if (!in_dev)
 845                 goto errout;
 846
 847         ifa = inet_alloc_ifa();
 848         if (!ifa)
 849                 /*
 850                  * A potential indev allocation can be left alive, it stays
 851                  * assigned to its device and is destroy with it.
 852                  */
 853                 goto errout;
 854
 855         ipv4_devconf_setall(in_dev);
 856         neigh_parms_data_state_setall(in_dev->arp_parms);
 857         in_dev_hold(in_dev);
 858
 859         if (!tb[IFA_ADDRESS])
 860                 tb[IFA_ADDRESS] = tb[IFA_LOCAL];
 861
 862         INIT_HLIST_NODE(&ifa->hash);
 863         ifa->ifa_prefixlen = ifm->ifa_prefixlen;
 864         ifa->ifa_mask = inet_make_mask(ifm->ifa_prefixlen);
 865         ifa->ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) :
 866                                          ifm->ifa_flags;
 867         ifa->ifa_scope = ifm->ifa_scope;
 868         ifa->ifa_dev = in_dev;
 869
 870         ifa->ifa_local = nla_get_in_addr(tb[IFA_LOCAL]);
 871         ifa->ifa_address = nla_get_in_addr(tb[IFA_ADDRESS]);
 872
 873         if (tb[IFA_BROADCAST])
 874                 ifa->ifa_broadcast = nla_get_in_addr(tb[IFA_BROADCAST]);
 875
 876         if (tb[IFA_LABEL])
 877                 nla_strlcpy(ifa->ifa_label, tb[IFA_LABEL], IFNAMSIZ);
 878         else
 879                 memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
 880
 881         if (tb[IFA_RT_PRIORITY])
 882                 ifa->ifa_rt_priority = nla_get_u32(tb[IFA_RT_PRIORITY]);
 883
 884         if (tb[IFA_CACHEINFO]) {
 885                 struct ifa_cacheinfo *ci;
 886
 887                 ci = nla_data(tb[IFA_CACHEINFO]);
 888                 if (!ci->ifa_valid || ci->ifa_prefered > ci->ifa_valid) {
 889                         err = -EINVAL;
 890                         goto errout_free;
 891                 }
 892                 *pvalid_lft = ci->ifa_valid;
 893                 *pprefered_lft = ci->ifa_prefered;
 894         }
 895
 896         return ifa;
 897
 898 errout_free:
 899         inet_free_ifa(ifa);
 900 errout:
 901         return ERR_PTR(err);
 902 }
 903
 904 static struct in_ifaddr *find_matching_ifa(struct in_ifaddr *ifa)
 905 {
 906         struct in_device *in_dev = ifa->ifa_dev;
 907         struct in_ifaddr *ifa1;
 908
 909         if (!ifa->ifa_local)
 910                 return NULL;
 911
 912         in_dev_for_each_ifa_rtnl(ifa1, in_dev) {
 913                 if (ifa1->ifa_mask == ifa->ifa_mask &&
 914                     inet_ifa_match(ifa1->ifa_address, ifa) &&
 915                     ifa1->ifa_local == ifa->ifa_local)
 916                         return ifa1;
 917         }
 918         return NULL;
 919 }
 920
 921 static int inet_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
 922                             struct netlink_ext_ack *extack)
 923 {
 924         struct net *net = sock_net(skb->sk);
 925         struct in_ifaddr *ifa;
 926         struct in_ifaddr *ifa_existing;
 927         __u32 valid_lft = INFINITY_LIFE_TIME;
 928         __u32 prefered_lft = INFINITY_LIFE_TIME;
 929
 930         ASSERT_RTNL();
 931
 932         ifa = rtm_to_ifaddr(net, nlh, &valid_lft, &prefered_lft, extack);
 933         if (IS_ERR(ifa))
 934                 return PTR_ERR(ifa);
 935
 936         ifa_existing = find_matching_ifa(ifa);
 937         if (!ifa_existing) {
 938                 /* It would be best to check for !NLM_F_CREATE here but
 939                  * userspace already relies on not having to provide this.
 940                  */
 941                 set_ifa_lifetime(ifa, valid_lft, prefered_lft);
 942                 if (ifa->ifa_flags & IFA_F_MCAUTOJOIN) {
 943                         int ret = ip_mc_config(net->ipv4.mc_autojoin_sk,
 944                                                true, ifa);
 945
 946                         if (ret < 0) {
 947                                 inet_free_ifa(ifa);
 948                                 return ret;
 949                         }
 950                 }
 951                 return __inet_insert_ifa(ifa, nlh, NETLINK_CB(skb).portid,
 952                                          extack);
 953         } else {
 954                 u32 new_metric = ifa->ifa_rt_priority;
 955
 956                 inet_free_ifa(ifa);
 957
 958                 if (nlh->nlmsg_flags & NLM_F_EXCL ||
 959                     !(nlh->nlmsg_flags & NLM_F_REPLACE))
 960                         return -EEXIST;
 961                 ifa = ifa_existing;
 962
 963                 if (ifa->ifa_rt_priority != new_metric) {
 964                         fib_modify_prefix_metric(ifa, new_metric);
 965                         ifa->ifa_rt_priority = new_metric;
 966                 }
 967
 968                 set_ifa_lifetime(ifa, valid_lft, prefered_lft);
 969                 cancel_delayed_work(&check_lifetime_work);
 970                 queue_delayed_work(system_power_efficient_wq,
 971                                 &check_lifetime_work, 0);
 972                 rtmsg_ifa(RTM_NEWADDR, ifa, nlh, NETLINK_CB(skb).portid);
 973         }
 974         return 0;
 975 }
 976
 977 /*
 978  *      Determine a default network mask, based on the IP address.
 979  */
 980
 981 static int inet_abc_len(__be32 addr)
 982 {
 983         int rc = -1;    /* Something else, probably a multicast. */
 984
 985         if (ipv4_is_zeronet(addr) || ipv4_is_lbcast(addr))
 986                 rc = 0;
 987         else {
 988                 __u32 haddr = ntohl(addr);
 989                 if (IN_CLASSA(haddr))
 990                         rc = 8;
 991                 else if (IN_CLASSB(haddr))
 992                         rc = 16;
 993                 else if (IN_CLASSC(haddr))
 994                         rc = 24;
 995                 else if (IN_CLASSE(haddr))
 996                         rc = 32;
 997         }
 998
 999         return rc;
1000 }
1001
1002
1003 int devinet_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr)
1004 {
1005         struct sockaddr_in sin_orig;
1006         struct sockaddr_in *sin = (struct sockaddr_in *)&ifr->ifr_addr;
1007         struct in_ifaddr __rcu **ifap = NULL;
1008         struct in_device *in_dev;
1009         struct in_ifaddr *ifa = NULL;
1010         struct net_device *dev;
1011         char *colon;
1012         int ret = -EFAULT;
1013         int tryaddrmatch = 0;
1014
1015         ifr->ifr_name[IFNAMSIZ - 1] = 0;
1016
1017         /* save original address for comparison */
1018         memcpy(&sin_orig, sin, sizeof(*sin));
1019
1020         colon = strchr(ifr->ifr_name, ':');
1021         if (colon)
1022                 *colon = 0;
1023
1024         dev_load(net, ifr->ifr_name);
1025
1026         switch (cmd) {
1027         case SIOCGIFADDR:       /* Get interface address */
1028         case SIOCGIFBRDADDR:    /* Get the broadcast address */
1029         case SIOCGIFDSTADDR:    /* Get the destination address */
1030         case SIOCGIFNETMASK:    /* Get the netmask for the interface */
1031                 /* Note that these ioctls will not sleep,
1032                    so that we do not impose a lock.
1033                    One day we will be forced to put shlock here (I mean SMP)
1034                  */
1035                 tryaddrmatch = (sin_orig.sin_family == AF_INET);
1036                 memset(sin, 0, sizeof(*sin));
1037                 sin->sin_family = AF_INET;
1038                 break;
1039
1040         case SIOCSIFFLAGS:
1041                 ret = -EPERM;
1042                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1043                         goto out;
1044                 break;
1045         case SIOCSIFADDR:       /* Set interface address (and family) */
1046         case SIOCSIFBRDADDR:    /* Set the broadcast address */
1047         case SIOCSIFDSTADDR:    /* Set the destination address */
1048         case SIOCSIFNETMASK:    /* Set the netmask for the interface */
1049                 ret = -EPERM;
1050                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1051                         goto out;
1052                 ret = -EINVAL;
1053                 if (sin->sin_family != AF_INET)
1054                         goto out;
1055                 break;
1056         default:
1057                 ret = -EINVAL;
1058                 goto out;
1059         }
1060
1061         rtnl_lock();
1062
1063         ret = -ENODEV;
1064         dev = __dev_get_by_name(net, ifr->ifr_name);
1065         if (!dev)
1066                 goto done;
1067
1068         if (colon)
1069                 *colon = ':';
1070
1071         in_dev = __in_dev_get_rtnl(dev);
1072         if (in_dev) {
1073                 if (tryaddrmatch) {
1074                         /* Matthias Andree */
1075                         /* compare label and address (4.4BSD style) */
1076                         /* note: we only do this for a limited set of ioctls
1077                            and only if the original address family was AF_INET.
1078                            This is checked above. */
1079
1080                         for (ifap = &in_dev->ifa_list;
1081                              (ifa = rtnl_dereference(*ifap)) != NULL;
1082                              ifap = &ifa->ifa_next) {
1083                                 if (!strcmp(ifr->ifr_name, ifa->ifa_label) &&
1084                                     sin_orig.sin_addr.s_addr ==
1085                                                         ifa->ifa_local) {
1086                                         break; /* found */
1087                                 }
1088                         }
1089                 }
1090                 /* we didn't get a match, maybe the application is
1091                    4.3BSD-style and passed in junk so we fall back to
1092                    comparing just the label */
1093                 if (!ifa) {
1094                         for (ifap = &in_dev->ifa_list;
1095                              (ifa = rtnl_dereference(*ifap)) != NULL;
1096                              ifap = &ifa->ifa_next)
1097                                 if (!strcmp(ifr->ifr_name, ifa->ifa_label))
1098                                         break;
1099                 }
1100         }
1101
1102         ret = -EADDRNOTAVAIL;
1103         if (!ifa && cmd != SIOCSIFADDR && cmd != SIOCSIFFLAGS)
1104                 goto done;
1105
1106         switch (cmd) {
1107         case SIOCGIFADDR:       /* Get interface address */
1108                 ret = 0;
1109                 sin->sin_addr.s_addr = ifa->ifa_local;
1110                 break;
1111
1112         case SIOCGIFBRDADDR:    /* Get the broadcast address */
1113                 ret = 0;
1114                 sin->sin_addr.s_addr = ifa->ifa_broadcast;
1115                 break;
1116
1117         case SIOCGIFDSTADDR:    /* Get the destination address */
1118                 ret = 0;
1119                 sin->sin_addr.s_addr = ifa->ifa_address;
1120                 break;
1121
1122         case SIOCGIFNETMASK:    /* Get the netmask for the interface */
1123                 ret = 0;
1124                 sin->sin_addr.s_addr = ifa->ifa_mask;
1125                 break;
1126
1127         case SIOCSIFFLAGS:
1128                 if (colon) {
1129                         ret = -EADDRNOTAVAIL;
1130                         if (!ifa)
1131                                 break;
1132                         ret = 0;
1133                         if (!(ifr->ifr_flags & IFF_UP))
1134                                 inet_del_ifa(in_dev, ifap, 1);
1135                         break;
1136                 }
1137                 ret = dev_change_flags(dev, ifr->ifr_flags, NULL);
1138                 break;
1139
1140         case SIOCSIFADDR:       /* Set interface address (and family) */
1141                 ret = -EINVAL;
1142                 if (inet_abc_len(sin->sin_addr.s_addr) < 0)
1143                         break;
1144
1145                 if (!ifa) {
1146                         ret = -ENOBUFS;
1147                         ifa = inet_alloc_ifa();
1148                         if (!ifa)
1149                                 break;
1150                         INIT_HLIST_NODE(&ifa->hash);
1151                         if (colon)
1152                                 memcpy(ifa->ifa_label, ifr->ifr_name, IFNAMSIZ);
1153                         else
1154                                 memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
1155                 } else {
1156                         ret = 0;
1157                         if (ifa->ifa_local == sin->sin_addr.s_addr)
1158                                 break;
1159                         inet_del_ifa(in_dev, ifap, 0);
1160                         ifa->ifa_broadcast = 0;
1161                         ifa->ifa_scope = 0;
1162                 }
1163
1164                 ifa->ifa_address = ifa->ifa_local = sin->sin_addr.s_addr;
1165
1166                 if (!(dev->flags & IFF_POINTOPOINT)) {
1167                         ifa->ifa_prefixlen = inet_abc_len(ifa->ifa_address);
1168                         ifa->ifa_mask = inet_make_mask(ifa->ifa_prefixlen);
1169                         if ((dev->flags & IFF_BROADCAST) &&
1170                             ifa->ifa_prefixlen < 31)
1171                                 ifa->ifa_broadcast = ifa->ifa_address |
1172                                                      ~ifa->ifa_mask;
1173                 } else {
1174                         ifa->ifa_prefixlen = 32;
1175                         ifa->ifa_mask = inet_make_mask(32);
1176                 }
1177                 set_ifa_lifetime(ifa, INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
1178                 ret = inet_set_ifa(dev, ifa);
1179                 break;
1180
1181         case SIOCSIFBRDADDR:    /* Set the broadcast address */
1182                 ret = 0;
1183                 if (ifa->ifa_broadcast != sin->sin_addr.s_addr) {
1184                         inet_del_ifa(in_dev, ifap, 0);
1185                         ifa->ifa_broadcast = sin->sin_addr.s_addr;
1186                         inet_insert_ifa(ifa);
1187                 }
1188                 break;
1189
1190         case SIOCSIFDSTADDR:    /* Set the destination address */
1191                 ret = 0;
1192                 if (ifa->ifa_address == sin->sin_addr.s_addr)
1193                         break;
1194                 ret = -EINVAL;
1195                 if (inet_abc_len(sin->sin_addr.s_addr) < 0)
1196                         break;
1197                 ret = 0;
1198                 inet_del_ifa(in_dev, ifap, 0);
1199                 ifa->ifa_address = sin->sin_addr.s_addr;
1200                 inet_insert_ifa(ifa);
1201                 break;
1202
1203         case SIOCSIFNETMASK:    /* Set the netmask for the interface */
1204
1205                 /*
1206                  *      The mask we set must be legal.
1207                  */
1208                 ret = -EINVAL;
1209                 if (bad_mask(sin->sin_addr.s_addr, 0))
1210                         break;
1211                 ret = 0;
1212                 if (ifa->ifa_mask != sin->sin_addr.s_addr) {
1213                         __be32 old_mask = ifa->ifa_mask;
1214                         inet_del_ifa(in_dev, ifap, 0);
1215                         ifa->ifa_mask = sin->sin_addr.s_addr;
1216                         ifa->ifa_prefixlen = inet_mask_len(ifa->ifa_mask);
1217
1218                         /* See if current broadcast address matches
1219                          * with current netmask, then recalculate
1220                          * the broadcast address. Otherwise it's a
1221                          * funny address, so don't touch it since
1222                          * the user seems to know what (s)he's doing...
1223                          */
1224                         if ((dev->flags & IFF_BROADCAST) &&
1225                             (ifa->ifa_prefixlen < 31) &&
1226                             (ifa->ifa_broadcast ==
1227                              (ifa->ifa_local|~old_mask))) {
1228                                 ifa->ifa_broadcast = (ifa->ifa_local |
1229                                                       ~sin->sin_addr.s_addr);
1230                         }
1231                         inet_insert_ifa(ifa);
1232                 }
1233                 break;
1234         }
1235 done:
1236         rtnl_unlock();
1237 out:
1238         return ret;
1239 }
1240
1241 static int inet_gifconf(struct net_device *dev, char __user *buf, int len, int size)
1242 {
1243         struct in_device *in_dev = __in_dev_get_rtnl(dev);
1244         const struct in_ifaddr *ifa;
1245         struct ifreq ifr;
1246         int done = 0;
1247
1248         if (WARN_ON(size > sizeof(struct ifreq)))
1249                 goto out;
1250
1251         if (!in_dev)
1252                 goto out;
1253
1254         in_dev_for_each_ifa_rtnl(ifa, in_dev) {
1255                 if (!buf) {
1256                         done += size;
1257                         continue;
1258                 }
1259                 if (len < size)
1260                         break;
1261                 memset(&ifr, 0, sizeof(struct ifreq));
1262                 strcpy(ifr.ifr_name, ifa->ifa_label);
1263
1264                 (*(struct sockaddr_in *)&ifr.ifr_addr).sin_family = AF_INET;
1265                 (*(struct sockaddr_in *)&ifr.ifr_addr).sin_addr.s_addr =
1266                                                                 ifa->ifa_local;
1267
1268                 if (copy_to_user(buf + done, &ifr, size)) {
1269                         done = -EFAULT;
1270                         break;
1271                 }
1272                 len  -= size;
1273                 done += size;
1274         }
1275 out:
1276         return done;
1277 }
1278
1279 static __be32 in_dev_select_addr(const struct in_device *in_dev,
1280                                  int scope)
1281 {
1282         const struct in_ifaddr *ifa;
1283
1284         in_dev_for_each_ifa_rcu(ifa, in_dev) {
1285                 if (ifa->ifa_flags & IFA_F_SECONDARY)
1286                         continue;
1287                 if (ifa->ifa_scope != RT_SCOPE_LINK &&
1288                     ifa->ifa_scope <= scope)
1289                         return ifa->ifa_local;
1290         }
1291
1292         return 0;
1293 }
1294
1295 __be32 inet_select_addr(const struct net_device *dev, __be32 dst, int scope)
1296 {
1297         const struct in_ifaddr *ifa;
1298         __be32 addr = 0;
1299         unsigned char localnet_scope = RT_SCOPE_HOST;
1300         struct in_device *in_dev;
1301         struct net *net = dev_net(dev);
1302         int master_idx;
1303
1304         rcu_read_lock();
1305         in_dev = __in_dev_get_rcu(dev);
1306         if (!in_dev)
1307                 goto no_in_dev;
1308
1309         if (unlikely(IN_DEV_ROUTE_LOCALNET(in_dev)))
1310                 localnet_scope = RT_SCOPE_LINK;
1311
1312         in_dev_for_each_ifa_rcu(ifa, in_dev) {
1313                 if (ifa->ifa_flags & IFA_F_SECONDARY)
1314                         continue;
1315                 if (min(ifa->ifa_scope, localnet_scope) > scope)
1316                         continue;
1317                 if (!dst || inet_ifa_match(dst, ifa)) {
1318                         addr = ifa->ifa_local;
1319                         break;
1320                 }
1321                 if (!addr)
1322                         addr = ifa->ifa_local;
1323         }
1324
1325         if (addr)
1326                 goto out_unlock;
1327 no_in_dev:
1328         master_idx = l3mdev_master_ifindex_rcu(dev);
1329
1330         /* For VRFs, the VRF device takes the place of the loopback device,
1331          * with addresses on it being preferred.  Note in such cases the
1332          * loopback device will be among the devices that fail the master_idx
1333          * equality check in the loop below.
1334          */
1335         if (master_idx &&
1336             (dev = dev_get_by_index_rcu(net, master_idx)) &&
1337             (in_dev = __in_dev_get_rcu(dev))) {
1338                 addr = in_dev_select_addr(in_dev, scope);
1339                 if (addr)
1340                         goto out_unlock;
1341         }
1342
1343         /* Not loopback addresses on loopback should be preferred
1344            in this case. It is important that lo is the first interface
1345            in dev_base list.
1346          */
1347         for_each_netdev_rcu(net, dev) {
1348                 if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1349                         continue;
1350
1351                 in_dev = __in_dev_get_rcu(dev);
1352                 if (!in_dev)
1353                         continue;
1354
1355                 addr = in_dev_select_addr(in_dev, scope);
1356                 if (addr)
1357                         goto out_unlock;
1358         }
1359 out_unlock:
1360         rcu_read_unlock();
1361         return addr;
1362 }
1363 EXPORT_SYMBOL(inet_select_addr);
1364
1365 static __be32 confirm_addr_indev(struct in_device *in_dev, __be32 dst,
1366                               __be32 local, int scope)
1367 {
1368         unsigned char localnet_scope = RT_SCOPE_HOST;
1369         const struct in_ifaddr *ifa;
1370         __be32 addr = 0;
1371         int same = 0;
1372
1373         if (unlikely(IN_DEV_ROUTE_LOCALNET(in_dev)))
1374                 localnet_scope = RT_SCOPE_LINK;
1375
1376         in_dev_for_each_ifa_rcu(ifa, in_dev) {
1377                 unsigned char min_scope = min(ifa->ifa_scope, localnet_scope);
1378
1379                 if (!addr &&
1380                     (local == ifa->ifa_local || !local) &&
1381                     min_scope <= scope) {
1382                         addr = ifa->ifa_local;
1383                         if (same)
1384                                 break;
1385                 }
1386                 if (!same) {
1387                         same = (!local || inet_ifa_match(local, ifa)) &&
1388                                 (!dst || inet_ifa_match(dst, ifa));
1389                         if (same && addr) {
1390                                 if (local || !dst)
1391                                         break;
1392                                 /* Is the selected addr into dst subnet? */
1393                                 if (inet_ifa_match(addr, ifa))
1394                                         break;
1395                                 /* No, then can we use new local src? */
1396                                 if (min_scope <= scope) {
1397                                         addr = ifa->ifa_local;
1398                                         break;
1399                                 }
1400                                 /* search for large dst subnet for addr */
1401                                 same = 0;
1402                         }
1403                 }
1404         }
1405
1406         return same ? addr : 0;
1407 }
1408
1409 /*
1410  * Confirm that local IP address exists using wildcards:
1411  * - net: netns to check, cannot be NULL
1412  * - in_dev: only on this interface, NULL=any interface
1413  * - dst: only in the same subnet as dst, 0=any dst
1414  * - local: address, 0=autoselect the local address
1415  * - scope: maximum allowed scope value for the local address
1416  */
1417 __be32 inet_confirm_addr(struct net *net, struct in_device *in_dev,
1418                          __be32 dst, __be32 local, int scope)
1419 {
1420         __be32 addr = 0;
1421         struct net_device *dev;
1422
1423         if (in_dev)
1424                 return confirm_addr_indev(in_dev, dst, local, scope);
1425
1426         rcu_read_lock();
1427         for_each_netdev_rcu(net, dev) {
1428                 in_dev = __in_dev_get_rcu(dev);
1429                 if (in_dev) {
1430                         addr = confirm_addr_indev(in_dev, dst, local, scope);
1431                         if (addr)
1432                                 break;
1433                 }
1434         }
1435         rcu_read_unlock();
1436
1437         return addr;
1438 }
1439 EXPORT_SYMBOL(inet_confirm_addr);
1440
1441 /*
1442  *      Device notifier
1443  */
1444
1445 int register_inetaddr_notifier(struct notifier_block *nb)
1446 {
1447         return blocking_notifier_chain_register(&inetaddr_chain, nb);
1448 }
1449 EXPORT_SYMBOL(register_inetaddr_notifier);
1450
1451 int unregister_inetaddr_notifier(struct notifier_block *nb)
1452 {
1453         return blocking_notifier_chain_unregister(&inetaddr_chain, nb);
1454 }
1455 EXPORT_SYMBOL(unregister_inetaddr_notifier);
1456
1457 int register_inetaddr_validator_notifier(struct notifier_block *nb)
1458 {
1459         return blocking_notifier_chain_register(&inetaddr_validator_chain, nb);
1460 }
1461 EXPORT_SYMBOL(register_inetaddr_validator_notifier);
1462
1463 int unregister_inetaddr_validator_notifier(struct notifier_block *nb)
1464 {
1465         return blocking_notifier_chain_unregister(&inetaddr_validator_chain,
1466             nb);
1467 }
1468 EXPORT_SYMBOL(unregister_inetaddr_validator_notifier);
1469
1470 /* Rename ifa_labels for a device name change. Make some effort to preserve
1471  * existing alias numbering and to create unique labels if possible.
1472 */
1473 static void inetdev_changename(struct net_device *dev, struct in_device *in_dev)
1474 {
1475         struct in_ifaddr *ifa;
1476         int named = 0;
1477
1478         in_dev_for_each_ifa_rtnl(ifa, in_dev) {
1479                 char old[IFNAMSIZ], *dot;
1480
1481                 memcpy(old, ifa->ifa_label, IFNAMSIZ);
1482                 memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
1483                 if (named++ == 0)
1484                         goto skip;
1485                 dot = strchr(old, ':');
1486                 if (!dot) {
1487                         sprintf(old, ":%d", named);
1488                         dot = old;
1489                 }
1490                 if (strlen(dot) + strlen(dev->name) < IFNAMSIZ)
1491                         strcat(ifa->ifa_label, dot);
1492                 else
1493                         strcpy(ifa->ifa_label + (IFNAMSIZ - strlen(dot) - 1), dot);
1494 skip:
1495                 rtmsg_ifa(RTM_NEWADDR, ifa, NULL, 0);
1496         }
1497 }
1498
1499 static void inetdev_send_gratuitous_arp(struct net_device *dev,
1500                                         struct in_device *in_dev)
1501
1502 {
1503         const struct in_ifaddr *ifa;
1504
1505         in_dev_for_each_ifa_rtnl(ifa, in_dev) {
1506                 arp_send(ARPOP_REQUEST, ETH_P_ARP,
1507                          ifa->ifa_local, dev,
1508                          ifa->ifa_local, NULL,
1509                          dev->dev_addr, NULL);
1510         }
1511 }
1512
1513 /* Called only under RTNL semaphore */
1514
1515 static int inetdev_event(struct notifier_block *this, unsigned long event,
1516                          void *ptr)
1517 {
1518         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1519         struct in_device *in_dev = __in_dev_get_rtnl(dev);
1520
1521         ASSERT_RTNL();
1522
1523         if (!in_dev) {
1524                 if (event == NETDEV_REGISTER) {
1525                         in_dev = inetdev_init(dev);
1526                         if (IS_ERR(in_dev))
1527                                 return notifier_from_errno(PTR_ERR(in_dev));
1528                         if (dev->flags & IFF_LOOPBACK) {
1529                                 IN_DEV_CONF_SET(in_dev, NOXFRM, 1);
1530                                 IN_DEV_CONF_SET(in_dev, NOPOLICY, 1);
1531                         }
1532                 } else if (event == NETDEV_CHANGEMTU) {
1533                         /* Re-enabling IP */
1534                         if (inetdev_valid_mtu(dev->mtu))
1535                                 in_dev = inetdev_init(dev);
1536                 }
1537                 goto out;
1538         }
1539
1540         switch (event) {
1541         case NETDEV_REGISTER:
1542                 pr_debug("%s: bug\n", __func__);
1543                 RCU_INIT_POINTER(dev->ip_ptr, NULL);
1544                 break;
1545         case NETDEV_UP:
1546                 if (!inetdev_valid_mtu(dev->mtu))
1547                         break;
1548                 if (dev->flags & IFF_LOOPBACK) {
1549                         struct in_ifaddr *ifa = inet_alloc_ifa();
1550
1551                         if (ifa) {
1552                                 INIT_HLIST_NODE(&ifa->hash);
1553                                 ifa->ifa_local =
1554                                   ifa->ifa_address = htonl(INADDR_LOOPBACK);
1555                                 ifa->ifa_prefixlen = 8;
1556                                 ifa->ifa_mask = inet_make_mask(8);
1557                                 in_dev_hold(in_dev);
1558                                 ifa->ifa_dev = in_dev;
1559                                 ifa->ifa_scope = RT_SCOPE_HOST;
1560                                 memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
1561                                 set_ifa_lifetime(ifa, INFINITY_LIFE_TIME,
1562                                                  INFINITY_LIFE_TIME);
1563                                 ipv4_devconf_setall(in_dev);
1564                                 neigh_parms_data_state_setall(in_dev->arp_parms);
1565                                 inet_insert_ifa(ifa);
1566                         }
1567                 }
1568                 ip_mc_up(in_dev);
1569                 fallthrough;
1570         case NETDEV_CHANGEADDR:
1571                 if (!IN_DEV_ARP_NOTIFY(in_dev))
1572                         break;
1573                 fallthrough;
1574         case NETDEV_NOTIFY_PEERS:
1575                 /* Send gratuitous ARP to notify of link change */
1576                 inetdev_send_gratuitous_arp(dev, in_dev);
1577                 break;
1578         case NETDEV_DOWN:
1579                 ip_mc_down(in_dev);
1580                 break;
1581         case NETDEV_PRE_TYPE_CHANGE:
1582                 ip_mc_unmap(in_dev);
1583                 break;
1584         case NETDEV_POST_TYPE_CHANGE:
1585                 ip_mc_remap(in_dev);
1586                 break;
1587         case NETDEV_CHANGEMTU:
1588                 if (inetdev_valid_mtu(dev->mtu))
1589                         break;
1590                 /* disable IP when MTU is not enough */
1591                 fallthrough;
1592         case NETDEV_UNREGISTER:
1593                 inetdev_destroy(in_dev);
1594                 break;
1595         case NETDEV_CHANGENAME:
1596                 /* Do not notify about label change, this event is
1597                  * not interesting to applications using netlink.
1598                  */
1599                 inetdev_changename(dev, in_dev);
1600
1601                 devinet_sysctl_unregister(in_dev);
1602                 devinet_sysctl_register(in_dev);
1603                 break;
1604         }
1605 out:
1606         return NOTIFY_DONE;
1607 }
1608
1609 static struct notifier_block ip_netdev_notifier = {
1610         .notifier_call = inetdev_event,
1611 };
1612
1613 static size_t inet_nlmsg_size(void)
1614 {
1615         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
1616                + nla_total_size(4) /* IFA_ADDRESS */
1617                + nla_total_size(4) /* IFA_LOCAL */
1618                + nla_total_size(4) /* IFA_BROADCAST */
1619                + nla_total_size(IFNAMSIZ) /* IFA_LABEL */
1620                + nla_total_size(4)  /* IFA_FLAGS */
1621                + nla_total_size(4)  /* IFA_RT_PRIORITY */
1622                + nla_total_size(sizeof(struct ifa_cacheinfo)); /* IFA_CACHEINFO */
1623 }
1624
1625 static inline u32 cstamp_delta(unsigned long cstamp)
1626 {
1627         return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
1628 }
1629
1630 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
1631                          unsigned long tstamp, u32 preferred, u32 valid)
1632 {
1633         struct ifa_cacheinfo ci;
1634
1635         ci.cstamp = cstamp_delta(cstamp);
1636         ci.tstamp = cstamp_delta(tstamp);
1637         ci.ifa_prefered = preferred;
1638         ci.ifa_valid = valid;
1639
1640         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
1641 }
1642
1643 static int inet_fill_ifaddr(struct sk_buff *skb, struct in_ifaddr *ifa,
1644                             struct inet_fill_args *args)
1645 {
1646         struct ifaddrmsg *ifm;
1647         struct nlmsghdr  *nlh;
1648         u32 preferred, valid;
1649
1650         nlh = nlmsg_put(skb, args->portid, args->seq, args->event, sizeof(*ifm),
1651                         args->flags);
1652         if (!nlh)
1653                 return -EMSGSIZE;
1654
1655         ifm = nlmsg_data(nlh);
1656         ifm->ifa_family = AF_INET;
1657         ifm->ifa_prefixlen = ifa->ifa_prefixlen;
1658         ifm->ifa_flags = ifa->ifa_flags;
1659         ifm->ifa_scope = ifa->ifa_scope;
1660         ifm->ifa_index = ifa->ifa_dev->dev->ifindex;
1661
1662         if (args->netnsid >= 0 &&
1663             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
1664                 goto nla_put_failure;
1665
1666         if (!(ifm->ifa_flags & IFA_F_PERMANENT)) {
1667                 preferred = ifa->ifa_preferred_lft;
1668                 valid = ifa->ifa_valid_lft;
1669                 if (preferred != INFINITY_LIFE_TIME) {
1670                         long tval = (jiffies - ifa->ifa_tstamp) / HZ;
1671
1672                         if (preferred > tval)
1673                                 preferred -= tval;
1674                         else
1675                                 preferred = 0;
1676                         if (valid != INFINITY_LIFE_TIME) {
1677                                 if (valid > tval)
1678                                         valid -= tval;
1679                                 else
1680                                         valid = 0;
1681                         }
1682                 }
1683         } else {
1684                 preferred = INFINITY_LIFE_TIME;
1685                 valid = INFINITY_LIFE_TIME;
1686         }
1687         if ((ifa->ifa_address &&
1688              nla_put_in_addr(skb, IFA_ADDRESS, ifa->ifa_address)) ||
1689             (ifa->ifa_local &&
1690              nla_put_in_addr(skb, IFA_LOCAL, ifa->ifa_local)) ||
1691             (ifa->ifa_broadcast &&
1692              nla_put_in_addr(skb, IFA_BROADCAST, ifa->ifa_broadcast)) ||
1693             (ifa->ifa_label[0] &&
1694              nla_put_string(skb, IFA_LABEL, ifa->ifa_label)) ||
1695             nla_put_u32(skb, IFA_FLAGS, ifa->ifa_flags) ||
1696             (ifa->ifa_rt_priority &&
1697              nla_put_u32(skb, IFA_RT_PRIORITY, ifa->ifa_rt_priority)) ||
1698             put_cacheinfo(skb, ifa->ifa_cstamp, ifa->ifa_tstamp,
1699                           preferred, valid))
1700                 goto nla_put_failure;
1701
1702         nlmsg_end(skb, nlh);
1703         return 0;
1704
1705 nla_put_failure:
1706         nlmsg_cancel(skb, nlh);
1707         return -EMSGSIZE;
1708 }
1709
1710 static int inet_valid_dump_ifaddr_req(const struct nlmsghdr *nlh,
1711                                       struct inet_fill_args *fillargs,
1712                                       struct net **tgt_net, struct sock *sk,
1713                                       struct netlink_callback *cb)
1714 {
1715         struct netlink_ext_ack *extack = cb->extack;
1716         struct nlattr *tb[IFA_MAX+1];
1717         struct ifaddrmsg *ifm;
1718         int err, i;
1719
1720         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
1721                 NL_SET_ERR_MSG(extack, "ipv4: Invalid header for address dump request");
1722                 return -EINVAL;
1723         }
1724
1725         ifm = nlmsg_data(nlh);
1726         if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
1727                 NL_SET_ERR_MSG(extack, "ipv4: Invalid values in header for address dump request");
1728                 return -EINVAL;
1729         }
1730
1731         fillargs->ifindex = ifm->ifa_index;
1732         if (fillargs->ifindex) {
1733                 cb->answer_flags |= NLM_F_DUMP_FILTERED;
1734                 fillargs->flags |= NLM_F_DUMP_FILTERED;
1735         }
1736
1737         err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
1738                                             ifa_ipv4_policy, extack);
1739         if (err < 0)
1740                 return err;
1741
1742         for (i = 0; i <= IFA_MAX; ++i) {
1743                 if (!tb[i])
1744                         continue;
1745
1746                 if (i == IFA_TARGET_NETNSID) {
1747                         struct net *net;
1748
1749                         fillargs->netnsid = nla_get_s32(tb[i]);
1750
1751                         net = rtnl_get_net_ns_capable(sk, fillargs->netnsid);
1752                         if (IS_ERR(net)) {
1753                                 fillargs->netnsid = -1;
1754                                 NL_SET_ERR_MSG(extack, "ipv4: Invalid target network namespace id");
1755                                 return PTR_ERR(net);
1756                         }
1757                         *tgt_net = net;
1758                 } else {
1759                         NL_SET_ERR_MSG(extack, "ipv4: Unsupported attribute in dump request");
1760                         return -EINVAL;
1761                 }
1762         }
1763
1764         return 0;
1765 }
1766
1767 static int in_dev_dump_addr(struct in_device *in_dev, struct sk_buff *skb,
1768                             struct netlink_callback *cb, int s_ip_idx,
1769                             struct inet_fill_args *fillargs)
1770 {
1771         struct in_ifaddr *ifa;
1772         int ip_idx = 0;
1773         int err;
1774
1775         in_dev_for_each_ifa_rtnl(ifa, in_dev) {
1776                 if (ip_idx < s_ip_idx) {
1777                         ip_idx++;
1778                         continue;
1779                 }
1780                 err = inet_fill_ifaddr(skb, ifa, fillargs);
1781                 if (err < 0)
1782                         goto done;
1783
1784                 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1785                 ip_idx++;
1786         }
1787         err = 0;
1788
1789 done:
1790         cb->args[2] = ip_idx;
1791
1792         return err;
1793 }
1794
1795 static int inet_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
1796 {
1797         const struct nlmsghdr *nlh = cb->nlh;
1798         struct inet_fill_args fillargs = {
1799                 .portid = NETLINK_CB(cb->skb).portid,
1800                 .seq = nlh->nlmsg_seq,
1801                 .event = RTM_NEWADDR,
1802                 .flags = NLM_F_MULTI,
1803                 .netnsid = -1,
1804         };
1805         struct net *net = sock_net(skb->sk);
1806         struct net *tgt_net = net;
1807         int h, s_h;
1808         int idx, s_idx;
1809         int s_ip_idx;
1810         struct net_device *dev;
1811         struct in_device *in_dev;
1812         struct hlist_head *head;
1813         int err = 0;
1814
1815         s_h = cb->args[0];
1816         s_idx = idx = cb->args[1];
1817         s_ip_idx = cb->args[2];
1818
1819         if (cb->strict_check) {
1820                 err = inet_valid_dump_ifaddr_req(nlh, &fillargs, &tgt_net,
1821                                                  skb->sk, cb);
1822                 if (err < 0)
1823                         goto put_tgt_net;
1824
1825                 err = 0;
1826                 if (fillargs.ifindex) {
1827                         dev = __dev_get_by_index(tgt_net, fillargs.ifindex);
1828                         if (!dev) {
1829                                 err = -ENODEV;
1830                                 goto put_tgt_net;
1831                         }
1832
1833                         in_dev = __in_dev_get_rtnl(dev);
1834                         if (in_dev) {
1835                                 err = in_dev_dump_addr(in_dev, skb, cb, s_ip_idx,
1836                                                        &fillargs);
1837                         }
1838                         goto put_tgt_net;
1839                 }
1840         }
1841
1842         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1843                 idx = 0;
1844                 head = &tgt_net->dev_index_head[h];
1845                 rcu_read_lock();
1846                 cb->seq = atomic_read(&tgt_net->ipv4.dev_addr_genid) ^
1847                           tgt_net->dev_base_seq;
1848                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
1849                         if (idx < s_idx)
1850                                 goto cont;
1851                         if (h > s_h || idx > s_idx)
1852                                 s_ip_idx = 0;
1853                         in_dev = __in_dev_get_rcu(dev);
1854                         if (!in_dev)
1855                                 goto cont;
1856
1857                         err = in_dev_dump_addr(in_dev, skb, cb, s_ip_idx,
1858                                                &fillargs);
1859                         if (err < 0) {
1860                                 rcu_read_unlock();
1861                                 goto done;
1862                         }
1863 cont:
1864                         idx++;
1865                 }
1866                 rcu_read_unlock();
1867         }
1868
1869 done:
1870         cb->args[0] = h;
1871         cb->args[1] = idx;
1872 put_tgt_net:
1873         if (fillargs.netnsid >= 0)
1874                 put_net(tgt_net);
1875
1876         return skb->len ? : err;
1877 }
1878
1879 static void rtmsg_ifa(int event, struct in_ifaddr *ifa, struct nlmsghdr *nlh,
1880                       u32 portid)
1881 {
1882         struct inet_fill_args fillargs = {
1883                 .portid = portid,
1884                 .seq = nlh ? nlh->nlmsg_seq : 0,
1885                 .event = event,
1886                 .flags = 0,
1887                 .netnsid = -1,
1888         };
1889         struct sk_buff *skb;
1890         int err = -ENOBUFS;
1891         struct net *net;
1892
1893         net = dev_net(ifa->ifa_dev->dev);
1894         skb = nlmsg_new(inet_nlmsg_size(), GFP_KERNEL);
1895         if (!skb)
1896                 goto errout;
1897
1898         err = inet_fill_ifaddr(skb, ifa, &fillargs);
1899         if (err < 0) {
1900                 /* -EMSGSIZE implies BUG in inet_nlmsg_size() */
1901                 WARN_ON(err == -EMSGSIZE);
1902                 kfree_skb(skb);
1903                 goto errout;
1904         }
1905         rtnl_notify(skb, net, portid, RTNLGRP_IPV4_IFADDR, nlh, GFP_KERNEL);
1906         return;
1907 errout:
1908         if (err < 0)
1909                 rtnl_set_sk_err(net, RTNLGRP_IPV4_IFADDR, err);
1910 }
1911
1912 static size_t inet_get_link_af_size(const struct net_device *dev,
1913                                     u32 ext_filter_mask)
1914 {
1915         struct in_device *in_dev = rcu_dereference_rtnl(dev->ip_ptr);
1916
1917         if (!in_dev)
1918                 return 0;
1919
1920         return nla_total_size(IPV4_DEVCONF_MAX * 4); /* IFLA_INET_CONF */
1921 }
1922
1923 static int inet_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
1924                              u32 ext_filter_mask)
1925 {
1926         struct in_device *in_dev = rcu_dereference_rtnl(dev->ip_ptr);
1927         struct nlattr *nla;
1928         int i;
1929
1930         if (!in_dev)
1931                 return -ENODATA;
1932
1933         nla = nla_reserve(skb, IFLA_INET_CONF, IPV4_DEVCONF_MAX * 4);
1934         if (!nla)
1935                 return -EMSGSIZE;
1936
1937         for (i = 0; i < IPV4_DEVCONF_MAX; i++)
1938                 ((u32 *) nla_data(nla))[i] = in_dev->cnf.data[i];
1939
1940         return 0;
1941 }
1942
1943 static const struct nla_policy inet_af_policy[IFLA_INET_MAX+1] = {
1944         [IFLA_INET_CONF]        = { .type = NLA_NESTED },
1945 };
1946
1947 static int inet_validate_link_af(const struct net_device *dev,
1948                                  const struct nlattr *nla)
1949 {
1950         struct nlattr *a, *tb[IFLA_INET_MAX+1];
1951         int err, rem;
1952
1953         if (dev && !__in_dev_get_rcu(dev))
1954                 return -EAFNOSUPPORT;
1955
1956         err = nla_parse_nested_deprecated(tb, IFLA_INET_MAX, nla,
1957                                           inet_af_policy, NULL);
1958         if (err < 0)
1959                 return err;
1960
1961         if (tb[IFLA_INET_CONF]) {
1962                 nla_for_each_nested(a, tb[IFLA_INET_CONF], rem) {
1963                         int cfgid = nla_type(a);
1964
1965                         if (nla_len(a) < 4)
1966                                 return -EINVAL;
1967
1968                         if (cfgid <= 0 || cfgid > IPV4_DEVCONF_MAX)
1969                                 return -EINVAL;
1970                 }
1971         }
1972
1973         return 0;
1974 }
1975
1976 static int inet_set_link_af(struct net_device *dev, const struct nlattr *nla)
1977 {
1978         struct in_device *in_dev = __in_dev_get_rcu(dev);
1979         struct nlattr *a, *tb[IFLA_INET_MAX+1];
1980         int rem;
1981
1982         if (!in_dev)
1983                 return -EAFNOSUPPORT;
1984
1985         if (nla_parse_nested_deprecated(tb, IFLA_INET_MAX, nla, NULL, NULL) < 0)
1986                 BUG();
1987
1988         if (tb[IFLA_INET_CONF]) {
1989                 nla_for_each_nested(a, tb[IFLA_INET_CONF], rem)
1990                         ipv4_devconf_set(in_dev, nla_type(a), nla_get_u32(a));
1991         }
1992
1993         return 0;
1994 }
1995
1996 static int inet_netconf_msgsize_devconf(int type)
1997 {
1998         int size = NLMSG_ALIGN(sizeof(struct netconfmsg))
1999                    + nla_total_size(4); /* NETCONFA_IFINDEX */
2000         bool all = false;
2001
2002         if (type == NETCONFA_ALL)
2003                 all = true;
2004
2005         if (all || type == NETCONFA_FORWARDING)
2006                 size += nla_total_size(4);
2007         if (all || type == NETCONFA_RP_FILTER)
2008                 size += nla_total_size(4);
2009         if (all || type == NETCONFA_MC_FORWARDING)
2010                 size += nla_total_size(4);
2011         if (all || type == NETCONFA_BC_FORWARDING)
2012                 size += nla_total_size(4);
2013         if (all || type == NETCONFA_PROXY_NEIGH)
2014                 size += nla_total_size(4);
2015         if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
2016                 size += nla_total_size(4);
2017
2018         return size;
2019 }
2020
2021 static int inet_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
2022                                      struct ipv4_devconf *devconf, u32 portid,
2023                                      u32 seq, int event, unsigned int flags,
2024                                      int type)
2025 {
2026         struct nlmsghdr  *nlh;
2027         struct netconfmsg *ncm;
2028         bool all = false;
2029
2030         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
2031                         flags);
2032         if (!nlh)
2033                 return -EMSGSIZE;
2034
2035         if (type == NETCONFA_ALL)
2036                 all = true;
2037
2038         ncm = nlmsg_data(nlh);
2039         ncm->ncm_family = AF_INET;
2040
2041         if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
2042                 goto nla_put_failure;
2043
2044         if (!devconf)
2045                 goto out;
2046
2047         if ((all || type == NETCONFA_FORWARDING) &&
2048             nla_put_s32(skb, NETCONFA_FORWARDING,
2049                         IPV4_DEVCONF(*devconf, FORWARDING)) < 0)
2050                 goto nla_put_failure;
2051         if ((all || type == NETCONFA_RP_FILTER) &&
2052             nla_put_s32(skb, NETCONFA_RP_FILTER,
2053                         IPV4_DEVCONF(*devconf, RP_FILTER)) < 0)
2054                 goto nla_put_failure;
2055         if ((all || type == NETCONFA_MC_FORWARDING) &&
2056             nla_put_s32(skb, NETCONFA_MC_FORWARDING,
2057                         IPV4_DEVCONF(*devconf, MC_FORWARDING)) < 0)
2058                 goto nla_put_failure;
2059         if ((all || type == NETCONFA_BC_FORWARDING) &&
2060             nla_put_s32(skb, NETCONFA_BC_FORWARDING,
2061                         IPV4_DEVCONF(*devconf, BC_FORWARDING)) < 0)
2062                 goto nla_put_failure;
2063         if ((all || type == NETCONFA_PROXY_NEIGH) &&
2064             nla_put_s32(skb, NETCONFA_PROXY_NEIGH,
2065                         IPV4_DEVCONF(*devconf, PROXY_ARP)) < 0)
2066                 goto nla_put_failure;
2067         if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
2068             nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
2069                         IPV4_DEVCONF(*devconf, IGNORE_ROUTES_WITH_LINKDOWN)) < 0)
2070                 goto nla_put_failure;
2071
2072 out:
2073         nlmsg_end(skb, nlh);
2074         return 0;
2075
2076 nla_put_failure:
2077         nlmsg_cancel(skb, nlh);
2078         return -EMSGSIZE;
2079 }
2080
2081 void inet_netconf_notify_devconf(struct net *net, int event, int type,
2082                                  int ifindex, struct ipv4_devconf *devconf)
2083 {
2084         struct sk_buff *skb;
2085         int err = -ENOBUFS;
2086
2087         skb = nlmsg_new(inet_netconf_msgsize_devconf(type), GFP_KERNEL);
2088         if (!skb)
2089                 goto errout;
2090
2091         err = inet_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
2092                                         event, 0, type);
2093         if (err < 0) {
2094                 /* -EMSGSIZE implies BUG in inet_netconf_msgsize_devconf() */
2095                 WARN_ON(err == -EMSGSIZE);
2096                 kfree_skb(skb);
2097                 goto errout;
2098         }
2099         rtnl_notify(skb, net, 0, RTNLGRP_IPV4_NETCONF, NULL, GFP_KERNEL);
2100         return;
2101 errout:
2102         if (err < 0)
2103                 rtnl_set_sk_err(net, RTNLGRP_IPV4_NETCONF, err);
2104 }
2105
2106 static const struct nla_policy devconf_ipv4_policy[NETCONFA_MAX+1] = {
2107         [NETCONFA_IFINDEX]      = { .len = sizeof(int) },
2108         [NETCONFA_FORWARDING]   = { .len = sizeof(int) },
2109         [NETCONFA_RP_FILTER]    = { .len = sizeof(int) },
2110         [NETCONFA_PROXY_NEIGH]  = { .len = sizeof(int) },
2111         [NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN]  = { .len = sizeof(int) },
2112 };
2113
2114 static int inet_netconf_valid_get_req(struct sk_buff *skb,
2115                                       const struct nlmsghdr *nlh,
2116                                       struct nlattr **tb,
2117                                       struct netlink_ext_ack *extack)
2118 {
2119         int i, err;
2120
2121         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(struct netconfmsg))) {
2122                 NL_SET_ERR_MSG(extack, "ipv4: Invalid header for netconf get request");
2123                 return -EINVAL;
2124         }
2125
2126         if (!netlink_strict_get_check(skb))
2127                 return nlmsg_parse_deprecated(nlh, sizeof(struct netconfmsg),
2128                                               tb, NETCONFA_MAX,
2129                                               devconf_ipv4_policy, extack);
2130
2131         err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct netconfmsg),
2132                                             tb, NETCONFA_MAX,
2133                                             devconf_ipv4_policy, extack);
2134         if (err)
2135                 return err;
2136
2137         for (i = 0; i <= NETCONFA_MAX; i++) {
2138                 if (!tb[i])
2139                         continue;
2140
2141                 switch (i) {
2142                 case NETCONFA_IFINDEX:
2143                         break;
2144                 default:
2145                         NL_SET_ERR_MSG(extack, "ipv4: Unsupported attribute in netconf get request");
2146                         return -EINVAL;
2147                 }
2148         }
2149
2150         return 0;
2151 }
2152
2153 static int inet_netconf_get_devconf(struct sk_buff *in_skb,
2154                                     struct nlmsghdr *nlh,
2155                                     struct netlink_ext_ack *extack)
2156 {
2157         struct net *net = sock_net(in_skb->sk);
2158         struct nlattr *tb[NETCONFA_MAX+1];
2159         struct sk_buff *skb;
2160         struct ipv4_devconf *devconf;
2161         struct in_device *in_dev;
2162         struct net_device *dev;
2163         int ifindex;
2164         int err;
2165
2166         err = inet_netconf_valid_get_req(in_skb, nlh, tb, extack);
2167         if (err)
2168                 goto errout;
2169
2170         err = -EINVAL;
2171         if (!tb[NETCONFA_IFINDEX])
2172                 goto errout;
2173
2174         ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
2175         switch (ifindex) {
2176         case NETCONFA_IFINDEX_ALL:
2177                 devconf = net->ipv4.devconf_all;
2178                 break;
2179         case NETCONFA_IFINDEX_DEFAULT:
2180                 devconf = net->ipv4.devconf_dflt;
2181                 break;
2182         default:
2183                 dev = __dev_get_by_index(net, ifindex);
2184                 if (!dev)
2185                         goto errout;
2186                 in_dev = __in_dev_get_rtnl(dev);
2187                 if (!in_dev)
2188                         goto errout;
2189                 devconf = &in_dev->cnf;
2190                 break;
2191         }
2192
2193         err = -ENOBUFS;
2194         skb = nlmsg_new(inet_netconf_msgsize_devconf(NETCONFA_ALL), GFP_KERNEL);
2195         if (!skb)
2196                 goto errout;
2197
2198         err = inet_netconf_fill_devconf(skb, ifindex, devconf,
2199                                         NETLINK_CB(in_skb).portid,
2200                                         nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
2201                                         NETCONFA_ALL);
2202         if (err < 0) {
2203                 /* -EMSGSIZE implies BUG in inet_netconf_msgsize_devconf() */
2204                 WARN_ON(err == -EMSGSIZE);
2205                 kfree_skb(skb);
2206                 goto errout;
2207         }
2208         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
2209 errout:
2210         return err;
2211 }
2212
2213 static int inet_netconf_dump_devconf(struct sk_buff *skb,
2214                                      struct netlink_callback *cb)
2215 {
2216         const struct nlmsghdr *nlh = cb->nlh;
2217         struct net *net = sock_net(skb->sk);
2218         int h, s_h;
2219         int idx, s_idx;
2220         struct net_device *dev;
2221         struct in_device *in_dev;
2222         struct hlist_head *head;
2223
2224         if (cb->strict_check) {
2225                 struct netlink_ext_ack *extack = cb->extack;
2226                 struct netconfmsg *ncm;
2227
2228                 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ncm))) {
2229                         NL_SET_ERR_MSG(extack, "ipv4: Invalid header for netconf dump request");
2230                         return -EINVAL;
2231                 }
2232
2233                 if (nlmsg_attrlen(nlh, sizeof(*ncm))) {
2234                         NL_SET_ERR_MSG(extack, "ipv4: Invalid data after header in netconf dump request");
2235                         return -EINVAL;
2236                 }
2237         }
2238
2239         s_h = cb->args[0];
2240         s_idx = idx = cb->args[1];
2241
2242         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
2243                 idx = 0;
2244                 head = &net->dev_index_head[h];
2245                 rcu_read_lock();
2246                 cb->seq = atomic_read(&net->ipv4.dev_addr_genid) ^
2247                           net->dev_base_seq;
2248                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
2249                         if (idx < s_idx)
2250                                 goto cont;
2251                         in_dev = __in_dev_get_rcu(dev);
2252                         if (!in_dev)
2253                                 goto cont;
2254
2255                         if (inet_netconf_fill_devconf(skb, dev->ifindex,
2256                                                       &in_dev->cnf,
2257                                                       NETLINK_CB(cb->skb).portid,
2258                                                       nlh->nlmsg_seq,
2259                                                       RTM_NEWNETCONF,
2260                                                       NLM_F_MULTI,
2261                                                       NETCONFA_ALL) < 0) {
2262                                 rcu_read_unlock();
2263                                 goto done;
2264                         }
2265                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
2266 cont:
2267                         idx++;
2268                 }
2269                 rcu_read_unlock();
2270         }
2271         if (h == NETDEV_HASHENTRIES) {
2272                 if (inet_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
2273                                               net->ipv4.devconf_all,
2274                                               NETLINK_CB(cb->skb).portid,
2275                                               nlh->nlmsg_seq,
2276                                               RTM_NEWNETCONF, NLM_F_MULTI,
2277                                               NETCONFA_ALL) < 0)
2278                         goto done;
2279                 else
2280                         h++;
2281         }
2282         if (h == NETDEV_HASHENTRIES + 1) {
2283                 if (inet_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
2284                                               net->ipv4.devconf_dflt,
2285                                               NETLINK_CB(cb->skb).portid,
2286                                               nlh->nlmsg_seq,
2287                                               RTM_NEWNETCONF, NLM_F_MULTI,
2288                                               NETCONFA_ALL) < 0)
2289                         goto done;
2290                 else
2291                         h++;
2292         }
2293 done:
2294         cb->args[0] = h;
2295         cb->args[1] = idx;
2296
2297         return skb->len;
2298 }
2299
2300 #ifdef CONFIG_SYSCTL
2301
2302 static void devinet_copy_dflt_conf(struct net *net, int i)
2303 {
2304         struct net_device *dev;
2305
2306         rcu_read_lock();
2307         for_each_netdev_rcu(net, dev) {
2308                 struct in_device *in_dev;
2309
2310                 in_dev = __in_dev_get_rcu(dev);
2311                 if (in_dev && !test_bit(i, in_dev->cnf.state))
2312                         in_dev->cnf.data[i] = net->ipv4.devconf_dflt->data[i];
2313         }
2314         rcu_read_unlock();
2315 }
2316
2317 /* called with RTNL locked */
2318 static void inet_forward_change(struct net *net)
2319 {
2320         struct net_device *dev;
2321         int on = IPV4_DEVCONF_ALL(net, FORWARDING);
2322
2323         IPV4_DEVCONF_ALL(net, ACCEPT_REDIRECTS) = !on;
2324         IPV4_DEVCONF_DFLT(net, FORWARDING) = on;
2325         inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
2326                                     NETCONFA_FORWARDING,
2327                                     NETCONFA_IFINDEX_ALL,
2328                                     net->ipv4.devconf_all);
2329         inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
2330                                     NETCONFA_FORWARDING,
2331                                     NETCONFA_IFINDEX_DEFAULT,
2332                                     net->ipv4.devconf_dflt);
2333
2334         for_each_netdev(net, dev) {
2335                 struct in_device *in_dev;
2336
2337                 if (on)
2338                         dev_disable_lro(dev);
2339
2340                 in_dev = __in_dev_get_rtnl(dev);
2341                 if (in_dev) {
2342                         IN_DEV_CONF_SET(in_dev, FORWARDING, on);
2343                         inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
2344                                                     NETCONFA_FORWARDING,
2345                                                     dev->ifindex, &in_dev->cnf);
2346                 }
2347         }
2348 }
2349
2350 static int devinet_conf_ifindex(struct net *net, struct ipv4_devconf *cnf)
2351 {
2352         if (cnf == net->ipv4.devconf_dflt)
2353                 return NETCONFA_IFINDEX_DEFAULT;
2354         else if (cnf == net->ipv4.devconf_all)
2355                 return NETCONFA_IFINDEX_ALL;
2356         else {
2357                 struct in_device *idev
2358                         = container_of(cnf, struct in_device, cnf);
2359                 return idev->dev->ifindex;
2360         }
2361 }
2362
2363 static int devinet_conf_proc(struct ctl_table *ctl, int write,
2364                              void __user *buffer,
2365                              size_t *lenp, loff_t *ppos)
2366 {
2367         int old_value = *(int *)ctl->data;
2368         int ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
2369         int new_value = *(int *)ctl->data;
2370
2371         if (write) {
2372                 struct ipv4_devconf *cnf = ctl->extra1;
2373                 struct net *net = ctl->extra2;
2374                 int i = (int *)ctl->data - cnf->data;
2375                 int ifindex;
2376
2377                 set_bit(i, cnf->state);
2378
2379                 if (cnf == net->ipv4.devconf_dflt)
2380                         devinet_copy_dflt_conf(net, i);
2381                 if (i == IPV4_DEVCONF_ACCEPT_LOCAL - 1 ||
2382                     i == IPV4_DEVCONF_ROUTE_LOCALNET - 1)
2383                         if ((new_value == 0) && (old_value != 0))
2384                                 rt_cache_flush(net);
2385
2386                 if (i == IPV4_DEVCONF_BC_FORWARDING - 1 &&
2387                     new_value != old_value)
2388                         rt_cache_flush(net);
2389
2390                 if (i == IPV4_DEVCONF_RP_FILTER - 1 &&
2391                     new_value != old_value) {
2392                         ifindex = devinet_conf_ifindex(net, cnf);
2393                         inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
2394                                                     NETCONFA_RP_FILTER,
2395                                                     ifindex, cnf);
2396                 }
2397                 if (i == IPV4_DEVCONF_PROXY_ARP - 1 &&
2398                     new_value != old_value) {
2399                         ifindex = devinet_conf_ifindex(net, cnf);
2400                         inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
2401                                                     NETCONFA_PROXY_NEIGH,
2402                                                     ifindex, cnf);
2403                 }
2404                 if (i == IPV4_DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN - 1 &&
2405                     new_value != old_value) {
2406                         ifindex = devinet_conf_ifindex(net, cnf);
2407                         inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
2408                                                     NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
2409                                                     ifindex, cnf);
2410                 }
2411         }
2412
2413         return ret;
2414 }
2415
2416 static int devinet_sysctl_forward(struct ctl_table *ctl, int write,
2417                                   void __user *buffer,
2418                                   size_t *lenp, loff_t *ppos)
2419 {
2420         int *valp = ctl->data;
2421         int val = *valp;
2422         loff_t pos = *ppos;
2423         int ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
2424
2425         if (write && *valp != val) {
2426                 struct net *net = ctl->extra2;
2427
2428                 if (valp != &IPV4_DEVCONF_DFLT(net, FORWARDING)) {
2429                         if (!rtnl_trylock()) {
2430                                 /* Restore the original values before restarting */
2431                                 *valp = val;
2432                                 *ppos = pos;
2433                                 return restart_syscall();
2434                         }
2435                         if (valp == &IPV4_DEVCONF_ALL(net, FORWARDING)) {
2436                                 inet_forward_change(net);
2437                         } else {
2438                                 struct ipv4_devconf *cnf = ctl->extra1;
2439                                 struct in_device *idev =
2440                                         container_of(cnf, struct in_device, cnf);
2441                                 if (*valp)
2442                                         dev_disable_lro(idev->dev);
2443                                 inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
2444                                                             NETCONFA_FORWARDING,
2445                                                             idev->dev->ifindex,
2446                                                             cnf);
2447                         }
2448                         rtnl_unlock();
2449                         rt_cache_flush(net);
2450                 } else
2451                         inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
2452                                                     NETCONFA_FORWARDING,
2453                                                     NETCONFA_IFINDEX_DEFAULT,
2454                                                     net->ipv4.devconf_dflt);
2455         }
2456
2457         return ret;
2458 }
2459
2460 static int ipv4_doint_and_flush(struct ctl_table *ctl, int write,
2461                                 void __user *buffer,
2462                                 size_t *lenp, loff_t *ppos)
2463 {
2464         int *valp = ctl->data;
2465         int val = *valp;
2466         int ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
2467         struct net *net = ctl->extra2;
2468
2469         if (write && *valp != val)
2470                 rt_cache_flush(net);
2471
2472         return ret;
2473 }
2474
2475 #define DEVINET_SYSCTL_ENTRY(attr, name, mval, proc) \
2476         { \
2477                 .procname       = name, \
2478                 .data           = ipv4_devconf.data + \
2479                                   IPV4_DEVCONF_ ## attr - 1, \
2480                 .maxlen         = sizeof(int), \
2481                 .mode           = mval, \
2482                 .proc_handler   = proc, \
2483                 .extra1         = &ipv4_devconf, \
2484         }
2485
2486 #define DEVINET_SYSCTL_RW_ENTRY(attr, name) \
2487         DEVINET_SYSCTL_ENTRY(attr, name, 0644, devinet_conf_proc)
2488
2489 #define DEVINET_SYSCTL_RO_ENTRY(attr, name) \
2490         DEVINET_SYSCTL_ENTRY(attr, name, 0444, devinet_conf_proc)
2491
2492 #define DEVINET_SYSCTL_COMPLEX_ENTRY(attr, name, proc) \
2493         DEVINET_SYSCTL_ENTRY(attr, name, 0644, proc)
2494
2495 #define DEVINET_SYSCTL_FLUSHING_ENTRY(attr, name) \
2496         DEVINET_SYSCTL_COMPLEX_ENTRY(attr, name, ipv4_doint_and_flush)
2497
2498 static struct devinet_sysctl_table {
2499         struct ctl_table_header *sysctl_header;
2500         struct ctl_table devinet_vars[__IPV4_DEVCONF_MAX];
2501 } devinet_sysctl = {
2502         .devinet_vars = {
2503                 DEVINET_SYSCTL_COMPLEX_ENTRY(FORWARDING, "forwarding",
2504                                              devinet_sysctl_forward),
2505                 DEVINET_SYSCTL_RO_ENTRY(MC_FORWARDING, "mc_forwarding"),
2506                 DEVINET_SYSCTL_RW_ENTRY(BC_FORWARDING, "bc_forwarding"),
2507
2508                 DEVINET_SYSCTL_RW_ENTRY(ACCEPT_REDIRECTS, "accept_redirects"),
2509                 DEVINET_SYSCTL_RW_ENTRY(SECURE_REDIRECTS, "secure_redirects"),
2510                 DEVINET_SYSCTL_RW_ENTRY(SHARED_MEDIA, "shared_media"),
2511                 DEVINET_SYSCTL_RW_ENTRY(RP_FILTER, "rp_filter"),
2512                 DEVINET_SYSCTL_RW_ENTRY(SEND_REDIRECTS, "send_redirects"),
2513                 DEVINET_SYSCTL_RW_ENTRY(ACCEPT_SOURCE_ROUTE,
2514                                         "accept_source_route"),
2515                 DEVINET_SYSCTL_RW_ENTRY(ACCEPT_LOCAL, "accept_local"),
2516                 DEVINET_SYSCTL_RW_ENTRY(SRC_VMARK, "src_valid_mark"),
2517                 DEVINET_SYSCTL_RW_ENTRY(PROXY_ARP, "proxy_arp"),
2518                 DEVINET_SYSCTL_RW_ENTRY(MEDIUM_ID, "medium_id"),
2519                 DEVINET_SYSCTL_RW_ENTRY(BOOTP_RELAY, "bootp_relay"),
2520                 DEVINET_SYSCTL_RW_ENTRY(LOG_MARTIANS, "log_martians"),
2521                 DEVINET_SYSCTL_RW_ENTRY(TAG, "tag"),
2522                 DEVINET_SYSCTL_RW_ENTRY(ARPFILTER, "arp_filter"),
2523                 DEVINET_SYSCTL_RW_ENTRY(ARP_ANNOUNCE, "arp_announce"),
2524                 DEVINET_SYSCTL_RW_ENTRY(ARP_IGNORE, "arp_ignore"),
2525                 DEVINET_SYSCTL_RW_ENTRY(ARP_ACCEPT, "arp_accept"),
2526                 DEVINET_SYSCTL_RW_ENTRY(ARP_NOTIFY, "arp_notify"),
2527                 DEVINET_SYSCTL_RW_ENTRY(PROXY_ARP_PVLAN, "proxy_arp_pvlan"),
2528                 DEVINET_SYSCTL_RW_ENTRY(FORCE_IGMP_VERSION,
2529                                         "force_igmp_version"),
2530                 DEVINET_SYSCTL_RW_ENTRY(IGMPV2_UNSOLICITED_REPORT_INTERVAL,
2531                                         "igmpv2_unsolicited_report_interval"),
2532                 DEVINET_SYSCTL_RW_ENTRY(IGMPV3_UNSOLICITED_REPORT_INTERVAL,
2533                                         "igmpv3_unsolicited_report_interval"),
2534                 DEVINET_SYSCTL_RW_ENTRY(IGNORE_ROUTES_WITH_LINKDOWN,
2535                                         "ignore_routes_with_linkdown"),
2536                 DEVINET_SYSCTL_RW_ENTRY(DROP_GRATUITOUS_ARP,
2537                                         "drop_gratuitous_arp"),
2538
2539                 DEVINET_SYSCTL_FLUSHING_ENTRY(NOXFRM, "disable_xfrm"),
2540                 DEVINET_SYSCTL_FLUSHING_ENTRY(NOPOLICY, "disable_policy"),
2541                 DEVINET_SYSCTL_FLUSHING_ENTRY(PROMOTE_SECONDARIES,
2542                                               "promote_secondaries"),
2543                 DEVINET_SYSCTL_FLUSHING_ENTRY(ROUTE_LOCALNET,
2544                                               "route_localnet"),
2545                 DEVINET_SYSCTL_FLUSHING_ENTRY(DROP_UNICAST_IN_L2_MULTICAST,
2546                                               "drop_unicast_in_l2_multicast"),
2547         },
2548 };
2549
2550 static int __devinet_sysctl_register(struct net *net, char *dev_name,
2551                                      int ifindex, struct ipv4_devconf *p)
2552 {
2553         int i;
2554         struct devinet_sysctl_table *t;
2555         char path[sizeof("net/ipv4/conf/") + IFNAMSIZ];
2556
2557         t = kmemdup(&devinet_sysctl, sizeof(*t), GFP_KERNEL);
2558         if (!t)
2559                 goto out;
2560
2561         for (i = 0; i < ARRAY_SIZE(t->devinet_vars) - 1; i++) {
2562                 t->devinet_vars[i].data += (char *)p - (char *)&ipv4_devconf;
2563                 t->devinet_vars[i].extra1 = p;
2564                 t->devinet_vars[i].extra2 = net;
2565         }
2566
2567         snprintf(path, sizeof(path), "net/ipv4/conf/%s", dev_name);
2568
2569         t->sysctl_header = register_net_sysctl(net, path, t->devinet_vars);
2570         if (!t->sysctl_header)
2571                 goto free;
2572
2573         p->sysctl = t;
2574
2575         inet_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
2576                                     ifindex, p);
2577         return 0;
2578
2579 free:
2580         kfree(t);
2581 out:
2582         return -ENOBUFS;
2583 }
2584
2585 static void __devinet_sysctl_unregister(struct net *net,
2586                                         struct ipv4_devconf *cnf, int ifindex)
2587 {
2588         struct devinet_sysctl_table *t = cnf->sysctl;
2589
2590         if (t) {
2591                 cnf->sysctl = NULL;
2592                 unregister_net_sysctl_table(t->sysctl_header);
2593                 kfree(t);
2594         }
2595
2596         inet_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
2597 }
2598
2599 static int devinet_sysctl_register(struct in_device *idev)
2600 {
2601         int err;
2602
2603         if (!sysctl_dev_name_is_allowed(idev->dev->name))
2604                 return -EINVAL;
2605
2606         err = neigh_sysctl_register(idev->dev, idev->arp_parms, NULL);
2607         if (err)
2608                 return err;
2609         err = __devinet_sysctl_register(dev_net(idev->dev), idev->dev->name,
2610                                         idev->dev->ifindex, &idev->cnf);
2611         if (err)
2612                 neigh_sysctl_unregister(idev->arp_parms);
2613         return err;
2614 }
2615
2616 static void devinet_sysctl_unregister(struct in_device *idev)
2617 {
2618         struct net *net = dev_net(idev->dev);
2619
2620         __devinet_sysctl_unregister(net, &idev->cnf, idev->dev->ifindex);
2621         neigh_sysctl_unregister(idev->arp_parms);
2622 }
2623
2624 static struct ctl_table ctl_forward_entry[] = {
2625         {
2626                 .procname       = "ip_forward",
2627                 .data           = &ipv4_devconf.data[
2628                                         IPV4_DEVCONF_FORWARDING - 1],
2629                 .maxlen         = sizeof(int),
2630                 .mode           = 0644,
2631                 .proc_handler   = devinet_sysctl_forward,
2632                 .extra1         = &ipv4_devconf,
2633                 .extra2         = &init_net,
2634         },
2635         { },
2636 };
2637 #endif
2638
2639 static __net_init int devinet_init_net(struct net *net)
2640 {
2641         int err;
2642         struct ipv4_devconf *all, *dflt;
2643 #ifdef CONFIG_SYSCTL
2644         struct ctl_table *tbl;
2645         struct ctl_table_header *forw_hdr;
2646 #endif
2647
2648         err = -ENOMEM;
2649         all = kmemdup(&ipv4_devconf, sizeof(ipv4_devconf), GFP_KERNEL);
2650         if (!all)
2651                 goto err_alloc_all;
2652
2653         dflt = kmemdup(&ipv4_devconf_dflt, sizeof(ipv4_devconf_dflt), GFP_KERNEL);
2654         if (!dflt)
2655                 goto err_alloc_dflt;
2656
2657 #ifdef CONFIG_SYSCTL
2658         tbl = kmemdup(ctl_forward_entry, sizeof(ctl_forward_entry), GFP_KERNEL);
2659         if (!tbl)
2660                 goto err_alloc_ctl;
2661
2662         tbl[0].data = &all->data[IPV4_DEVCONF_FORWARDING - 1];
2663         tbl[0].extra1 = all;
2664         tbl[0].extra2 = net;
2665 #endif
2666
2667         if ((!IS_ENABLED(CONFIG_SYSCTL) ||
2668              sysctl_devconf_inherit_init_net != 2) &&
2669             !net_eq(net, &init_net)) {
2670                 memcpy(all, init_net.ipv4.devconf_all, sizeof(ipv4_devconf));
2671                 memcpy(dflt, init_net.ipv4.devconf_dflt, sizeof(ipv4_devconf_dflt));
2672         }
2673
2674 #ifdef CONFIG_SYSCTL
2675         err = __devinet_sysctl_register(net, "all", NETCONFA_IFINDEX_ALL, all);
2676         if (err < 0)
2677                 goto err_reg_all;
2678
2679         err = __devinet_sysctl_register(net, "default",
2680                                         NETCONFA_IFINDEX_DEFAULT, dflt);
2681         if (err < 0)
2682                 goto err_reg_dflt;
2683
2684         err = -ENOMEM;
2685         forw_hdr = register_net_sysctl(net, "net/ipv4", tbl);
2686         if (!forw_hdr)
2687                 goto err_reg_ctl;
2688         net->ipv4.forw_hdr = forw_hdr;
2689 #endif
2690
2691         net->ipv4.devconf_all = all;
2692         net->ipv4.devconf_dflt = dflt;
2693         return 0;
2694
2695 #ifdef CONFIG_SYSCTL
2696 err_reg_ctl:
2697         __devinet_sysctl_unregister(net, dflt, NETCONFA_IFINDEX_DEFAULT);
2698 err_reg_dflt:
2699         __devinet_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
2700 err_reg_all:
2701         kfree(tbl);
2702 err_alloc_ctl:
2703 #endif
2704         kfree(dflt);
2705 err_alloc_dflt:
2706         kfree(all);
2707 err_alloc_all:
2708         return err;
2709 }
2710
2711 static __net_exit void devinet_exit_net(struct net *net)
2712 {
2713 #ifdef CONFIG_SYSCTL
2714         struct ctl_table *tbl;
2715
2716         tbl = net->ipv4.forw_hdr->ctl_table_arg;
2717         unregister_net_sysctl_table(net->ipv4.forw_hdr);
2718         __devinet_sysctl_unregister(net, net->ipv4.devconf_dflt,
2719                                     NETCONFA_IFINDEX_DEFAULT);
2720         __devinet_sysctl_unregister(net, net->ipv4.devconf_all,
2721                                     NETCONFA_IFINDEX_ALL);
2722         kfree(tbl);
2723 #endif
2724         kfree(net->ipv4.devconf_dflt);
2725         kfree(net->ipv4.devconf_all);
2726 }
2727
2728 static __net_initdata struct pernet_operations devinet_ops = {
2729         .init = devinet_init_net,
2730         .exit = devinet_exit_net,
2731 };
2732
2733 static struct rtnl_af_ops inet_af_ops __read_mostly = {
2734         .family           = AF_INET,
2735         .fill_link_af     = inet_fill_link_af,
2736         .get_link_af_size = inet_get_link_af_size,
2737         .validate_link_af = inet_validate_link_af,
2738         .set_link_af      = inet_set_link_af,
2739 };
2740
2741 void __init devinet_init(void)
2742 {
2743         int i;
2744
2745         for (i = 0; i < IN4_ADDR_HSIZE; i++)
2746                 INIT_HLIST_HEAD(&inet_addr_lst[i]);
2747
2748         register_pernet_subsys(&devinet_ops);
2749
2750         register_gifconf(PF_INET, inet_gifconf);
2751         register_netdevice_notifier(&ip_netdev_notifier);
2752
2753         queue_delayed_work(system_power_efficient_wq, &check_lifetime_work, 0);
2754
2755         rtnl_af_register(&inet_af_ops);
2756
2757         rtnl_register(PF_INET, RTM_NEWADDR, inet_rtm_newaddr, NULL, 0);
2758         rtnl_register(PF_INET, RTM_DELADDR, inet_rtm_deladdr, NULL, 0);
2759         rtnl_register(PF_INET, RTM_GETADDR, NULL, inet_dump_ifaddr, 0);
2760         rtnl_register(PF_INET, RTM_GETNETCONF, inet_netconf_get_devconf,
2761                       inet_netconf_dump_devconf, 0);
2762 }