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_ACCOUNT);
 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                 neigh_parms_release(&arp_tbl, in_dev->arp_parms);
 280                 in_dev_put(in_dev);
 281                 in_dev = NULL;
 282                 goto out;
 283         }
 284         ip_mc_init_dev(in_dev);
 285         if (dev->flags & IFF_UP)
 286                 ip_mc_up(in_dev);
 287
 288         /* we can receive as soon as ip_ptr is set -- do this last */
 289         rcu_assign_pointer(dev->ip_ptr, in_dev);
 290 out:
 291         return in_dev ?: ERR_PTR(err);
 292 out_kfree:
 293         kfree(in_dev);
 294         in_dev = NULL;
 295         goto out;
 296 }
 297
 298 static void in_dev_rcu_put(struct rcu_head *head)
 299 {
 300         struct in_device *idev = container_of(head, struct in_device, rcu_head);
 301         in_dev_put(idev);
 302 }
 303
 304 static void inetdev_destroy(struct in_device *in_dev)
 305 {
 306         struct net_device *dev;
 307         struct in_ifaddr *ifa;
 308
 309         ASSERT_RTNL();
 310
 311         dev = in_dev->dev;
 312
 313         in_dev->dead = 1;
 314
 315         ip_mc_destroy_dev(in_dev);
 316
 317         while ((ifa = rtnl_dereference(in_dev->ifa_list)) != NULL) {
 318                 inet_del_ifa(in_dev, &in_dev->ifa_list, 0);
 319                 inet_free_ifa(ifa);
 320         }
 321
 322         RCU_INIT_POINTER(dev->ip_ptr, NULL);
 323
 324         devinet_sysctl_unregister(in_dev);
 325         neigh_parms_release(&arp_tbl, in_dev->arp_parms);
 326         arp_ifdown(dev);
 327
 328         call_rcu(&in_dev->rcu_head, in_dev_rcu_put);
 329 }
 330
 331 int inet_addr_onlink(struct in_device *in_dev, __be32 a, __be32 b)
 332 {
 333         const struct in_ifaddr *ifa;
 334
 335         rcu_read_lock();
 336         in_dev_for_each_ifa_rcu(ifa, in_dev) {
 337                 if (inet_ifa_match(a, ifa)) {
 338                         if (!b || inet_ifa_match(b, ifa)) {
 339                                 rcu_read_unlock();
 340                                 return 1;
 341                         }
 342                 }
 343         }
 344         rcu_read_unlock();
 345         return 0;
 346 }
 347
 348 static void __inet_del_ifa(struct in_device *in_dev,
 349                            struct in_ifaddr __rcu **ifap,
 350                            int destroy, struct nlmsghdr *nlh, u32 portid)
 351 {
 352         struct in_ifaddr *promote = NULL;
 353         struct in_ifaddr *ifa, *ifa1;
 354         struct in_ifaddr *last_prim;
 355         struct in_ifaddr *prev_prom = NULL;
 356         int do_promote = IN_DEV_PROMOTE_SECONDARIES(in_dev);
 357
 358         ASSERT_RTNL();
 359
 360         ifa1 = rtnl_dereference(*ifap);
 361         last_prim = rtnl_dereference(in_dev->ifa_list);
 362         if (in_dev->dead)
 363                 goto no_promotions;
 364
 365         /* 1. Deleting primary ifaddr forces deletion all secondaries
 366          * unless alias promotion is set
 367          **/
 368
 369         if (!(ifa1->ifa_flags & IFA_F_SECONDARY)) {
 370                 struct in_ifaddr __rcu **ifap1 = &ifa1->ifa_next;
 371
 372                 while ((ifa = rtnl_dereference(*ifap1)) != NULL) {
 373                         if (!(ifa->ifa_flags & IFA_F_SECONDARY) &&
 374                             ifa1->ifa_scope <= ifa->ifa_scope)
 375                                 last_prim = ifa;
 376
 377                         if (!(ifa->ifa_flags & IFA_F_SECONDARY) ||
 378                             ifa1->ifa_mask != ifa->ifa_mask ||
 379                             !inet_ifa_match(ifa1->ifa_address, ifa)) {
 380                                 ifap1 = &ifa->ifa_next;
 381                                 prev_prom = ifa;
 382                                 continue;
 383                         }
 384
 385                         if (!do_promote) {
 386                                 inet_hash_remove(ifa);
 387                                 *ifap1 = ifa->ifa_next;
 388
 389                                 rtmsg_ifa(RTM_DELADDR, ifa, nlh, portid);
 390                                 blocking_notifier_call_chain(&inetaddr_chain,
 391                                                 NETDEV_DOWN, ifa);
 392                                 inet_free_ifa(ifa);
 393                         } else {
 394                                 promote = ifa;
 395                                 break;
 396                         }
 397                 }
 398         }
 399
 400         /* On promotion all secondaries from subnet are changing
 401          * the primary IP, we must remove all their routes silently
 402          * and later to add them back with new prefsrc. Do this
 403          * while all addresses are on the device list.
 404          */
 405         for (ifa = promote; ifa; ifa = rtnl_dereference(ifa->ifa_next)) {
 406                 if (ifa1->ifa_mask == ifa->ifa_mask &&
 407                     inet_ifa_match(ifa1->ifa_address, ifa))
 408                         fib_del_ifaddr(ifa, ifa1);
 409         }
 410
 411 no_promotions:
 412         /* 2. Unlink it */
 413
 414         *ifap = ifa1->ifa_next;
 415         inet_hash_remove(ifa1);
 416
 417         /* 3. Announce address deletion */
 418
 419         /* Send message first, then call notifier.
 420            At first sight, FIB update triggered by notifier
 421            will refer to already deleted ifaddr, that could confuse
 422            netlink listeners. It is not true: look, gated sees
 423            that route deleted and if it still thinks that ifaddr
 424            is valid, it will try to restore deleted routes... Grr.
 425            So that, this order is correct.
 426          */
 427         rtmsg_ifa(RTM_DELADDR, ifa1, nlh, portid);
 428         blocking_notifier_call_chain(&inetaddr_chain, NETDEV_DOWN, ifa1);
 429
 430         if (promote) {
 431                 struct in_ifaddr *next_sec;
 432
 433                 next_sec = rtnl_dereference(promote->ifa_next);
 434                 if (prev_prom) {
 435                         struct in_ifaddr *last_sec;
 436
 437                         rcu_assign_pointer(prev_prom->ifa_next, next_sec);
 438
 439                         last_sec = rtnl_dereference(last_prim->ifa_next);
 440                         rcu_assign_pointer(promote->ifa_next, last_sec);
 441                         rcu_assign_pointer(last_prim->ifa_next, promote);
 442                 }
 443
 444                 promote->ifa_flags &= ~IFA_F_SECONDARY;
 445                 rtmsg_ifa(RTM_NEWADDR, promote, nlh, portid);
 446                 blocking_notifier_call_chain(&inetaddr_chain,
 447                                 NETDEV_UP, promote);
 448                 for (ifa = next_sec; ifa;
 449                      ifa = rtnl_dereference(ifa->ifa_next)) {
 450                         if (ifa1->ifa_mask != ifa->ifa_mask ||
 451                             !inet_ifa_match(ifa1->ifa_address, ifa))
 452                                         continue;
 453                         fib_add_ifaddr(ifa);
 454                 }
 455
 456         }
 457         if (destroy)
 458                 inet_free_ifa(ifa1);
 459 }
 460
 461 static void inet_del_ifa(struct in_device *in_dev,
 462                          struct in_ifaddr __rcu **ifap,
 463                          int destroy)
 464 {
 465         __inet_del_ifa(in_dev, ifap, destroy, NULL, 0);
 466 }
 467
 468 static void check_lifetime(struct work_struct *work);
 469
 470 static DECLARE_DELAYED_WORK(check_lifetime_work, check_lifetime);
 471
 472 static int __inet_insert_ifa(struct in_ifaddr *ifa, struct nlmsghdr *nlh,
 473                              u32 portid, struct netlink_ext_ack *extack)
 474 {
 475         struct in_ifaddr __rcu **last_primary, **ifap;
 476         struct in_device *in_dev = ifa->ifa_dev;
 477         struct in_validator_info ivi;
 478         struct in_ifaddr *ifa1;
 479         int ret;
 480
 481         ASSERT_RTNL();
 482
 483         if (!ifa->ifa_local) {
 484                 inet_free_ifa(ifa);
 485                 return 0;
 486         }
 487
 488         ifa->ifa_flags &= ~IFA_F_SECONDARY;
 489         last_primary = &in_dev->ifa_list;
 490
 491         /* Don't set IPv6 only flags to IPv4 addresses */
 492         ifa->ifa_flags &= ~IPV6ONLY_FLAGS;
 493
 494         ifap = &in_dev->ifa_list;
 495         ifa1 = rtnl_dereference(*ifap);
 496
 497         while (ifa1) {
 498                 if (!(ifa1->ifa_flags & IFA_F_SECONDARY) &&
 499                     ifa->ifa_scope <= ifa1->ifa_scope)
 500                         last_primary = &ifa1->ifa_next;
 501                 if (ifa1->ifa_mask == ifa->ifa_mask &&
 502                     inet_ifa_match(ifa1->ifa_address, ifa)) {
 503                         if (ifa1->ifa_local == ifa->ifa_local) {
 504                                 inet_free_ifa(ifa);
 505                                 return -EEXIST;
 506                         }
 507                         if (ifa1->ifa_scope != ifa->ifa_scope) {
 508                                 inet_free_ifa(ifa);
 509                                 return -EINVAL;
 510                         }
 511                         ifa->ifa_flags |= IFA_F_SECONDARY;
 512                 }
 513
 514                 ifap = &ifa1->ifa_next;
 515                 ifa1 = rtnl_dereference(*ifap);
 516         }
 517
 518         /* Allow any devices that wish to register ifaddr validtors to weigh
 519          * in now, before changes are committed.  The rntl lock is serializing
 520          * access here, so the state should not change between a validator call
 521          * and a final notify on commit.  This isn't invoked on promotion under
 522          * the assumption that validators are checking the address itself, and
 523          * not the flags.
 524          */
 525         ivi.ivi_addr = ifa->ifa_address;
 526         ivi.ivi_dev = ifa->ifa_dev;
 527         ivi.extack = extack;
 528         ret = blocking_notifier_call_chain(&inetaddr_validator_chain,
 529                                            NETDEV_UP, &ivi);
 530         ret = notifier_to_errno(ret);
 531         if (ret) {
 532                 inet_free_ifa(ifa);
 533                 return ret;
 534         }
 535
 536         if (!(ifa->ifa_flags & IFA_F_SECONDARY)) {
 537                 prandom_seed((__force u32) ifa->ifa_local);
 538                 ifap = last_primary;
 539         }
 540
 541         rcu_assign_pointer(ifa->ifa_next, *ifap);
 542         rcu_assign_pointer(*ifap, ifa);
 543
 544         inet_hash_insert(dev_net(in_dev->dev), ifa);
 545
 546         cancel_delayed_work(&check_lifetime_work);
 547         queue_delayed_work(system_power_efficient_wq, &check_lifetime_work, 0);
 548
 549         /* Send message first, then call notifier.
 550            Notifier will trigger FIB update, so that
 551            listeners of netlink will know about new ifaddr */
 552         rtmsg_ifa(RTM_NEWADDR, ifa, nlh, portid);
 553         blocking_notifier_call_chain(&inetaddr_chain, NETDEV_UP, ifa);
 554
 555         return 0;
 556 }
 557
 558 static int inet_insert_ifa(struct in_ifaddr *ifa)
 559 {
 560         return __inet_insert_ifa(ifa, NULL, 0, NULL);
 561 }
 562
 563 static int inet_set_ifa(struct net_device *dev, struct in_ifaddr *ifa)
 564 {
 565         struct in_device *in_dev = __in_dev_get_rtnl(dev);
 566
 567         ASSERT_RTNL();
 568
 569         if (!in_dev) {
 570                 inet_free_ifa(ifa);
 571                 return -ENOBUFS;
 572         }
 573         ipv4_devconf_setall(in_dev);
 574         neigh_parms_data_state_setall(in_dev->arp_parms);
 575         if (ifa->ifa_dev != in_dev) {
 576                 WARN_ON(ifa->ifa_dev);
 577                 in_dev_hold(in_dev);
 578                 ifa->ifa_dev = in_dev;
 579         }
 580         if (ipv4_is_loopback(ifa->ifa_local))
 581                 ifa->ifa_scope = RT_SCOPE_HOST;
 582         return inet_insert_ifa(ifa);
 583 }
 584
 585 /* Caller must hold RCU or RTNL :
 586  * We dont take a reference on found in_device
 587  */
 588 struct in_device *inetdev_by_index(struct net *net, int ifindex)
 589 {
 590         struct net_device *dev;
 591         struct in_device *in_dev = NULL;
 592
 593         rcu_read_lock();
 594         dev = dev_get_by_index_rcu(net, ifindex);
 595         if (dev)
 596                 in_dev = rcu_dereference_rtnl(dev->ip_ptr);
 597         rcu_read_unlock();
 598         return in_dev;
 599 }
 600 EXPORT_SYMBOL(inetdev_by_index);
 601
 602 /* Called only from RTNL semaphored context. No locks. */
 603
 604 struct in_ifaddr *inet_ifa_byprefix(struct in_device *in_dev, __be32 prefix,
 605                                     __be32 mask)
 606 {
 607         struct in_ifaddr *ifa;
 608
 609         ASSERT_RTNL();
 610
 611         in_dev_for_each_ifa_rtnl(ifa, in_dev) {
 612                 if (ifa->ifa_mask == mask && inet_ifa_match(prefix, ifa))
 613                         return ifa;
 614         }
 615         return NULL;
 616 }
 617
 618 static int ip_mc_autojoin_config(struct net *net, bool join,
 619                                  const struct in_ifaddr *ifa)
 620 {
 621 #if defined(CONFIG_IP_MULTICAST)
 622         struct ip_mreqn mreq = {
 623                 .imr_multiaddr.s_addr = ifa->ifa_address,
 624                 .imr_ifindex = ifa->ifa_dev->dev->ifindex,
 625         };
 626         struct sock *sk = net->ipv4.mc_autojoin_sk;
 627         int ret;
 628
 629         ASSERT_RTNL();
 630
 631         lock_sock(sk);
 632         if (join)
 633                 ret = ip_mc_join_group(sk, &mreq);
 634         else
 635                 ret = ip_mc_leave_group(sk, &mreq);
 636         release_sock(sk);
 637
 638         return ret;
 639 #else
 640         return -EOPNOTSUPP;
 641 #endif
 642 }
 643
 644 static int inet_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
 645                             struct netlink_ext_ack *extack)
 646 {
 647         struct net *net = sock_net(skb->sk);
 648         struct in_ifaddr __rcu **ifap;
 649         struct nlattr *tb[IFA_MAX+1];
 650         struct in_device *in_dev;
 651         struct ifaddrmsg *ifm;
 652         struct in_ifaddr *ifa;
 653         int err;
 654
 655         ASSERT_RTNL();
 656
 657         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
 658                                      ifa_ipv4_policy, extack);
 659         if (err < 0)
 660                 goto errout;
 661
 662         ifm = nlmsg_data(nlh);
 663         in_dev = inetdev_by_index(net, ifm->ifa_index);
 664         if (!in_dev) {
 665                 err = -ENODEV;
 666                 goto errout;
 667         }
 668
 669         for (ifap = &in_dev->ifa_list; (ifa = rtnl_dereference(*ifap)) != NULL;
 670              ifap = &ifa->ifa_next) {
 671                 if (tb[IFA_LOCAL] &&
 672                     ifa->ifa_local != nla_get_in_addr(tb[IFA_LOCAL]))
 673                         continue;
 674
 675                 if (tb[IFA_LABEL] && nla_strcmp(tb[IFA_LABEL], ifa->ifa_label))
 676                         continue;
 677
 678                 if (tb[IFA_ADDRESS] &&
 679                     (ifm->ifa_prefixlen != ifa->ifa_prefixlen ||
 680                     !inet_ifa_match(nla_get_in_addr(tb[IFA_ADDRESS]), ifa)))
 681                         continue;
 682
 683                 if (ipv4_is_multicast(ifa->ifa_address))
 684                         ip_mc_autojoin_config(net, false, ifa);
 685                 __inet_del_ifa(in_dev, ifap, 1, nlh, NETLINK_CB(skb).portid);
 686                 return 0;
 687         }
 688
 689         err = -EADDRNOTAVAIL;
 690 errout:
 691         return err;
 692 }
 693
 694 #define INFINITY_LIFE_TIME      0xFFFFFFFF
 695
 696 static void check_lifetime(struct work_struct *work)
 697 {
 698         unsigned long now, next, next_sec, next_sched;
 699         struct in_ifaddr *ifa;
 700         struct hlist_node *n;
 701         int i;
 702
 703         now = jiffies;
 704         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
 705
 706         for (i = 0; i < IN4_ADDR_HSIZE; i++) {
 707                 bool change_needed = false;
 708
 709                 rcu_read_lock();
 710                 hlist_for_each_entry_rcu(ifa, &inet_addr_lst[i], hash) {
 711                         unsigned long age;
 712
 713                         if (ifa->ifa_flags & IFA_F_PERMANENT)
 714                                 continue;
 715
 716                         /* We try to batch several events at once. */
 717                         age = (now - ifa->ifa_tstamp +
 718                                ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
 719
 720                         if (ifa->ifa_valid_lft != INFINITY_LIFE_TIME &&
 721                             age >= ifa->ifa_valid_lft) {
 722                                 change_needed = true;
 723                         } else if (ifa->ifa_preferred_lft ==
 724                                    INFINITY_LIFE_TIME) {
 725                                 continue;
 726                         } else if (age >= ifa->ifa_preferred_lft) {
 727                                 if (time_before(ifa->ifa_tstamp +
 728                                                 ifa->ifa_valid_lft * HZ, next))
 729                                         next = ifa->ifa_tstamp +
 730                                                ifa->ifa_valid_lft * HZ;
 731
 732                                 if (!(ifa->ifa_flags & IFA_F_DEPRECATED))
 733                                         change_needed = true;
 734                         } else if (time_before(ifa->ifa_tstamp +
 735                                                ifa->ifa_preferred_lft * HZ,
 736                                                next)) {
 737                                 next = ifa->ifa_tstamp +
 738                                        ifa->ifa_preferred_lft * HZ;
 739                         }
 740                 }
 741                 rcu_read_unlock();
 742                 if (!change_needed)
 743                         continue;
 744                 rtnl_lock();
 745                 hlist_for_each_entry_safe(ifa, n, &inet_addr_lst[i], hash) {
 746                         unsigned long age;
 747
 748                         if (ifa->ifa_flags & IFA_F_PERMANENT)
 749                                 continue;
 750
 751                         /* We try to batch several events at once. */
 752                         age = (now - ifa->ifa_tstamp +
 753                                ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
 754
 755                         if (ifa->ifa_valid_lft != INFINITY_LIFE_TIME &&
 756                             age >= ifa->ifa_valid_lft) {
 757                                 struct in_ifaddr __rcu **ifap;
 758                                 struct in_ifaddr *tmp;
 759
 760                                 ifap = &ifa->ifa_dev->ifa_list;
 761                                 tmp = rtnl_dereference(*ifap);
 762                                 while (tmp) {
 763                                         if (tmp == ifa) {
 764                                                 inet_del_ifa(ifa->ifa_dev,
 765                                                              ifap, 1);
 766                                                 break;
 767                                         }
 768                                         ifap = &tmp->ifa_next;
 769                                         tmp = rtnl_dereference(*ifap);
 770                                 }
 771                         } else if (ifa->ifa_preferred_lft !=
 772                                    INFINITY_LIFE_TIME &&
 773                                    age >= ifa->ifa_preferred_lft &&
 774                                    !(ifa->ifa_flags & IFA_F_DEPRECATED)) {
 775                                 ifa->ifa_flags |= IFA_F_DEPRECATED;
 776                                 rtmsg_ifa(RTM_NEWADDR, ifa, NULL, 0);
 777                         }
 778                 }
 779                 rtnl_unlock();
 780         }
 781
 782         next_sec = round_jiffies_up(next);
 783         next_sched = next;
 784
 785         /* If rounded timeout is accurate enough, accept it. */
 786         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
 787                 next_sched = next_sec;
 788
 789         now = jiffies;
 790         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
 791         if (time_before(next_sched, now + ADDRCONF_TIMER_FUZZ_MAX))
 792                 next_sched = now + ADDRCONF_TIMER_FUZZ_MAX;
 793
 794         queue_delayed_work(system_power_efficient_wq, &check_lifetime_work,
 795                         next_sched - now);
 796 }
 797
 798 static void set_ifa_lifetime(struct in_ifaddr *ifa, __u32 valid_lft,
 799                              __u32 prefered_lft)
 800 {
 801         unsigned long timeout;
 802
 803         ifa->ifa_flags &= ~(IFA_F_PERMANENT | IFA_F_DEPRECATED);
 804
 805         timeout = addrconf_timeout_fixup(valid_lft, HZ);
 806         if (addrconf_finite_timeout(timeout))
 807                 ifa->ifa_valid_lft = timeout;
 808         else
 809                 ifa->ifa_flags |= IFA_F_PERMANENT;
 810
 811         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
 812         if (addrconf_finite_timeout(timeout)) {
 813                 if (timeout == 0)
 814                         ifa->ifa_flags |= IFA_F_DEPRECATED;
 815                 ifa->ifa_preferred_lft = timeout;
 816         }
 817         ifa->ifa_tstamp = jiffies;
 818         if (!ifa->ifa_cstamp)
 819                 ifa->ifa_cstamp = ifa->ifa_tstamp;
 820 }
 821
 822 static struct in_ifaddr *rtm_to_ifaddr(struct net *net, struct nlmsghdr *nlh,
 823                                        __u32 *pvalid_lft, __u32 *pprefered_lft,
 824                                        struct netlink_ext_ack *extack)
 825 {
 826         struct nlattr *tb[IFA_MAX+1];
 827         struct in_ifaddr *ifa;
 828         struct ifaddrmsg *ifm;
 829         struct net_device *dev;
 830         struct in_device *in_dev;
 831         int err;
 832
 833         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
 834                                      ifa_ipv4_policy, extack);
 835         if (err < 0)
 836                 goto errout;
 837
 838         ifm = nlmsg_data(nlh);
 839         err = -EINVAL;
 840         if (ifm->ifa_prefixlen > 32 || !tb[IFA_LOCAL])
 841                 goto errout;
 842
 843         dev = __dev_get_by_index(net, ifm->ifa_index);
 844         err = -ENODEV;
 845         if (!dev)
 846                 goto errout;
 847
 848         in_dev = __in_dev_get_rtnl(dev);
 849         err = -ENOBUFS;
 850         if (!in_dev)
 851                 goto errout;
 852
 853         ifa = inet_alloc_ifa();
 854         if (!ifa)
 855                 /*
 856                  * A potential indev allocation can be left alive, it stays
 857                  * assigned to its device and is destroy with it.
 858                  */
 859                 goto errout;
 860
 861         ipv4_devconf_setall(in_dev);
 862         neigh_parms_data_state_setall(in_dev->arp_parms);
 863         in_dev_hold(in_dev);
 864
 865         if (!tb[IFA_ADDRESS])
 866                 tb[IFA_ADDRESS] = tb[IFA_LOCAL];
 867
 868         INIT_HLIST_NODE(&ifa->hash);
 869         ifa->ifa_prefixlen = ifm->ifa_prefixlen;
 870         ifa->ifa_mask = inet_make_mask(ifm->ifa_prefixlen);
 871         ifa->ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) :
 872                                          ifm->ifa_flags;
 873         ifa->ifa_scope = ifm->ifa_scope;
 874         ifa->ifa_dev = in_dev;
 875
 876         ifa->ifa_local = nla_get_in_addr(tb[IFA_LOCAL]);
 877         ifa->ifa_address = nla_get_in_addr(tb[IFA_ADDRESS]);
 878
 879         if (tb[IFA_BROADCAST])
 880                 ifa->ifa_broadcast = nla_get_in_addr(tb[IFA_BROADCAST]);
 881
 882         if (tb[IFA_LABEL])
 883                 nla_strscpy(ifa->ifa_label, tb[IFA_LABEL], IFNAMSIZ);
 884         else
 885                 memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
 886
 887         if (tb[IFA_RT_PRIORITY])
 888                 ifa->ifa_rt_priority = nla_get_u32(tb[IFA_RT_PRIORITY]);
 889
 890         if (tb[IFA_CACHEINFO]) {
 891                 struct ifa_cacheinfo *ci;
 892
 893                 ci = nla_data(tb[IFA_CACHEINFO]);
 894                 if (!ci->ifa_valid || ci->ifa_prefered > ci->ifa_valid) {
 895                         err = -EINVAL;
 896                         goto errout_free;
 897                 }
 898                 *pvalid_lft = ci->ifa_valid;
 899                 *pprefered_lft = ci->ifa_prefered;
 900         }
 901
 902         return ifa;
 903
 904 errout_free:
 905         inet_free_ifa(ifa);
 906 errout:
 907         return ERR_PTR(err);
 908 }
 909
 910 static struct in_ifaddr *find_matching_ifa(struct in_ifaddr *ifa)
 911 {
 912         struct in_device *in_dev = ifa->ifa_dev;
 913         struct in_ifaddr *ifa1;
 914
 915         if (!ifa->ifa_local)
 916                 return NULL;
 917
 918         in_dev_for_each_ifa_rtnl(ifa1, in_dev) {
 919                 if (ifa1->ifa_mask == ifa->ifa_mask &&
 920                     inet_ifa_match(ifa1->ifa_address, ifa) &&
 921                     ifa1->ifa_local == ifa->ifa_local)
 922                         return ifa1;
 923         }
 924         return NULL;
 925 }
 926
 927 static int inet_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
 928                             struct netlink_ext_ack *extack)
 929 {
 930         struct net *net = sock_net(skb->sk);
 931         struct in_ifaddr *ifa;
 932         struct in_ifaddr *ifa_existing;
 933         __u32 valid_lft = INFINITY_LIFE_TIME;
 934         __u32 prefered_lft = INFINITY_LIFE_TIME;
 935
 936         ASSERT_RTNL();
 937
 938         ifa = rtm_to_ifaddr(net, nlh, &valid_lft, &prefered_lft, extack);
 939         if (IS_ERR(ifa))
 940                 return PTR_ERR(ifa);
 941
 942         ifa_existing = find_matching_ifa(ifa);
 943         if (!ifa_existing) {
 944                 /* It would be best to check for !NLM_F_CREATE here but
 945                  * userspace already relies on not having to provide this.
 946                  */
 947                 set_ifa_lifetime(ifa, valid_lft, prefered_lft);
 948                 if (ifa->ifa_flags & IFA_F_MCAUTOJOIN) {
 949                         int ret = ip_mc_autojoin_config(net, true, ifa);
 950
 951                         if (ret < 0) {
 952                                 inet_free_ifa(ifa);
 953                                 return ret;
 954                         }
 955                 }
 956                 return __inet_insert_ifa(ifa, nlh, NETLINK_CB(skb).portid,
 957                                          extack);
 958         } else {
 959                 u32 new_metric = ifa->ifa_rt_priority;
 960
 961                 inet_free_ifa(ifa);
 962
 963                 if (nlh->nlmsg_flags & NLM_F_EXCL ||
 964                     !(nlh->nlmsg_flags & NLM_F_REPLACE))
 965                         return -EEXIST;
 966                 ifa = ifa_existing;
 967
 968                 if (ifa->ifa_rt_priority != new_metric) {
 969                         fib_modify_prefix_metric(ifa, new_metric);
 970                         ifa->ifa_rt_priority = new_metric;
 971                 }
 972
 973                 set_ifa_lifetime(ifa, valid_lft, prefered_lft);
 974                 cancel_delayed_work(&check_lifetime_work);
 975                 queue_delayed_work(system_power_efficient_wq,
 976                                 &check_lifetime_work, 0);
 977                 rtmsg_ifa(RTM_NEWADDR, ifa, nlh, NETLINK_CB(skb).portid);
 978         }
 979         return 0;
 980 }
 981
 982 /*
 983  *      Determine a default network mask, based on the IP address.
 984  */
 985
 986 static int inet_abc_len(__be32 addr)
 987 {
 988         int rc = -1;    /* Something else, probably a multicast. */
 989
 990         if (ipv4_is_zeronet(addr) || ipv4_is_lbcast(addr))
 991                 rc = 0;
 992         else {
 993                 __u32 haddr = ntohl(addr);
 994                 if (IN_CLASSA(haddr))
 995                         rc = 8;
 996                 else if (IN_CLASSB(haddr))
 997                         rc = 16;
 998                 else if (IN_CLASSC(haddr))
 999                         rc = 24;
1000                 else if (IN_CLASSE(haddr))
1001                         rc = 32;
1002         }
1003
1004         return rc;
1005 }
1006
1007
1008 int devinet_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr)
1009 {
1010         struct sockaddr_in sin_orig;
1011         struct sockaddr_in *sin = (struct sockaddr_in *)&ifr->ifr_addr;
1012         struct in_ifaddr __rcu **ifap = NULL;
1013         struct in_device *in_dev;
1014         struct in_ifaddr *ifa = NULL;
1015         struct net_device *dev;
1016         char *colon;
1017         int ret = -EFAULT;
1018         int tryaddrmatch = 0;
1019
1020         ifr->ifr_name[IFNAMSIZ - 1] = 0;
1021
1022         /* save original address for comparison */
1023         memcpy(&sin_orig, sin, sizeof(*sin));
1024
1025         colon = strchr(ifr->ifr_name, ':');
1026         if (colon)
1027                 *colon = 0;
1028
1029         dev_load(net, ifr->ifr_name);
1030
1031         switch (cmd) {
1032         case SIOCGIFADDR:       /* Get interface address */
1033         case SIOCGIFBRDADDR:    /* Get the broadcast address */
1034         case SIOCGIFDSTADDR:    /* Get the destination address */
1035         case SIOCGIFNETMASK:    /* Get the netmask for the interface */
1036                 /* Note that these ioctls will not sleep,
1037                    so that we do not impose a lock.
1038                    One day we will be forced to put shlock here (I mean SMP)
1039                  */
1040                 tryaddrmatch = (sin_orig.sin_family == AF_INET);
1041                 memset(sin, 0, sizeof(*sin));
1042                 sin->sin_family = AF_INET;
1043                 break;
1044
1045         case SIOCSIFFLAGS:
1046                 ret = -EPERM;
1047                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1048                         goto out;
1049                 break;
1050         case SIOCSIFADDR:       /* Set interface address (and family) */
1051         case SIOCSIFBRDADDR:    /* Set the broadcast address */
1052         case SIOCSIFDSTADDR:    /* Set the destination address */
1053         case SIOCSIFNETMASK:    /* Set the netmask for the interface */
1054                 ret = -EPERM;
1055                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1056                         goto out;
1057                 ret = -EINVAL;
1058                 if (sin->sin_family != AF_INET)
1059                         goto out;
1060                 break;
1061         default:
1062                 ret = -EINVAL;
1063                 goto out;
1064         }
1065
1066         rtnl_lock();
1067
1068         ret = -ENODEV;
1069         dev = __dev_get_by_name(net, ifr->ifr_name);
1070         if (!dev)
1071                 goto done;
1072
1073         if (colon)
1074                 *colon = ':';
1075
1076         in_dev = __in_dev_get_rtnl(dev);
1077         if (in_dev) {
1078                 if (tryaddrmatch) {
1079                         /* Matthias Andree */
1080                         /* compare label and address (4.4BSD style) */
1081                         /* note: we only do this for a limited set of ioctls
1082                            and only if the original address family was AF_INET.
1083                            This is checked above. */
1084
1085                         for (ifap = &in_dev->ifa_list;
1086                              (ifa = rtnl_dereference(*ifap)) != NULL;
1087                              ifap = &ifa->ifa_next) {
1088                                 if (!strcmp(ifr->ifr_name, ifa->ifa_label) &&
1089                                     sin_orig.sin_addr.s_addr ==
1090                                                         ifa->ifa_local) {
1091                                         break; /* found */
1092                                 }
1093                         }
1094                 }
1095                 /* we didn't get a match, maybe the application is
1096                    4.3BSD-style and passed in junk so we fall back to
1097                    comparing just the label */
1098                 if (!ifa) {
1099                         for (ifap = &in_dev->ifa_list;
1100                              (ifa = rtnl_dereference(*ifap)) != NULL;
1101                              ifap = &ifa->ifa_next)
1102                                 if (!strcmp(ifr->ifr_name, ifa->ifa_label))
1103                                         break;
1104                 }
1105         }
1106
1107         ret = -EADDRNOTAVAIL;
1108         if (!ifa && cmd != SIOCSIFADDR && cmd != SIOCSIFFLAGS)
1109                 goto done;
1110
1111         switch (cmd) {
1112         case SIOCGIFADDR:       /* Get interface address */
1113                 ret = 0;
1114                 sin->sin_addr.s_addr = ifa->ifa_local;
1115                 break;
1116
1117         case SIOCGIFBRDADDR:    /* Get the broadcast address */
1118                 ret = 0;
1119                 sin->sin_addr.s_addr = ifa->ifa_broadcast;
1120                 break;
1121
1122         case SIOCGIFDSTADDR:    /* Get the destination address */
1123                 ret = 0;
1124                 sin->sin_addr.s_addr = ifa->ifa_address;
1125                 break;
1126
1127         case SIOCGIFNETMASK:    /* Get the netmask for the interface */
1128                 ret = 0;
1129                 sin->sin_addr.s_addr = ifa->ifa_mask;
1130                 break;
1131
1132         case SIOCSIFFLAGS:
1133                 if (colon) {
1134                         ret = -EADDRNOTAVAIL;
1135                         if (!ifa)
1136                                 break;
1137                         ret = 0;
1138                         if (!(ifr->ifr_flags & IFF_UP))
1139                                 inet_del_ifa(in_dev, ifap, 1);
1140                         break;
1141                 }
1142                 ret = dev_change_flags(dev, ifr->ifr_flags, NULL);
1143                 break;
1144
1145         case SIOCSIFADDR:       /* Set interface address (and family) */
1146                 ret = -EINVAL;
1147                 if (inet_abc_len(sin->sin_addr.s_addr) < 0)
1148                         break;
1149
1150                 if (!ifa) {
1151                         ret = -ENOBUFS;
1152                         ifa = inet_alloc_ifa();
1153                         if (!ifa)
1154                                 break;
1155                         INIT_HLIST_NODE(&ifa->hash);
1156                         if (colon)
1157                                 memcpy(ifa->ifa_label, ifr->ifr_name, IFNAMSIZ);
1158                         else
1159                                 memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
1160                 } else {
1161                         ret = 0;
1162                         if (ifa->ifa_local == sin->sin_addr.s_addr)
1163                                 break;
1164                         inet_del_ifa(in_dev, ifap, 0);
1165                         ifa->ifa_broadcast = 0;
1166                         ifa->ifa_scope = 0;
1167                 }
1168
1169                 ifa->ifa_address = ifa->ifa_local = sin->sin_addr.s_addr;
1170
1171                 if (!(dev->flags & IFF_POINTOPOINT)) {
1172                         ifa->ifa_prefixlen = inet_abc_len(ifa->ifa_address);
1173                         ifa->ifa_mask = inet_make_mask(ifa->ifa_prefixlen);
1174                         if ((dev->flags & IFF_BROADCAST) &&
1175                             ifa->ifa_prefixlen < 31)
1176                                 ifa->ifa_broadcast = ifa->ifa_address |
1177                                                      ~ifa->ifa_mask;
1178                 } else {
1179                         ifa->ifa_prefixlen = 32;
1180                         ifa->ifa_mask = inet_make_mask(32);
1181                 }
1182                 set_ifa_lifetime(ifa, INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
1183                 ret = inet_set_ifa(dev, ifa);
1184                 break;
1185
1186         case SIOCSIFBRDADDR:    /* Set the broadcast address */
1187                 ret = 0;
1188                 if (ifa->ifa_broadcast != sin->sin_addr.s_addr) {
1189                         inet_del_ifa(in_dev, ifap, 0);
1190                         ifa->ifa_broadcast = sin->sin_addr.s_addr;
1191                         inet_insert_ifa(ifa);
1192                 }
1193                 break;
1194
1195         case SIOCSIFDSTADDR:    /* Set the destination address */
1196                 ret = 0;
1197                 if (ifa->ifa_address == sin->sin_addr.s_addr)
1198                         break;
1199                 ret = -EINVAL;
1200                 if (inet_abc_len(sin->sin_addr.s_addr) < 0)
1201                         break;
1202                 ret = 0;
1203                 inet_del_ifa(in_dev, ifap, 0);
1204                 ifa->ifa_address = sin->sin_addr.s_addr;
1205                 inet_insert_ifa(ifa);
1206                 break;
1207
1208         case SIOCSIFNETMASK:    /* Set the netmask for the interface */
1209
1210                 /*
1211                  *      The mask we set must be legal.
1212                  */
1213                 ret = -EINVAL;
1214                 if (bad_mask(sin->sin_addr.s_addr, 0))
1215                         break;
1216                 ret = 0;
1217                 if (ifa->ifa_mask != sin->sin_addr.s_addr) {
1218                         __be32 old_mask = ifa->ifa_mask;
1219                         inet_del_ifa(in_dev, ifap, 0);
1220                         ifa->ifa_mask = sin->sin_addr.s_addr;
1221                         ifa->ifa_prefixlen = inet_mask_len(ifa->ifa_mask);
1222
1223                         /* See if current broadcast address matches
1224                          * with current netmask, then recalculate
1225                          * the broadcast address. Otherwise it's a
1226                          * funny address, so don't touch it since
1227                          * the user seems to know what (s)he's doing...
1228                          */
1229                         if ((dev->flags & IFF_BROADCAST) &&
1230                             (ifa->ifa_prefixlen < 31) &&
1231                             (ifa->ifa_broadcast ==
1232                              (ifa->ifa_local|~old_mask))) {
1233                                 ifa->ifa_broadcast = (ifa->ifa_local |
1234                                                       ~sin->sin_addr.s_addr);
1235                         }
1236                         inet_insert_ifa(ifa);
1237                 }
1238                 break;
1239         }
1240 done:
1241         rtnl_unlock();
1242 out:
1243         return ret;
1244 }
1245
1246 int inet_gifconf(struct net_device *dev, char __user *buf, int len, int size)
1247 {
1248         struct in_device *in_dev = __in_dev_get_rtnl(dev);
1249         const struct in_ifaddr *ifa;
1250         struct ifreq ifr;
1251         int done = 0;
1252
1253         if (WARN_ON(size > sizeof(struct ifreq)))
1254                 goto out;
1255
1256         if (!in_dev)
1257                 goto out;
1258
1259         in_dev_for_each_ifa_rtnl(ifa, in_dev) {
1260                 if (!buf) {
1261                         done += size;
1262                         continue;
1263                 }
1264                 if (len < size)
1265                         break;
1266                 memset(&ifr, 0, sizeof(struct ifreq));
1267                 strcpy(ifr.ifr_name, ifa->ifa_label);
1268
1269                 (*(struct sockaddr_in *)&ifr.ifr_addr).sin_family = AF_INET;
1270                 (*(struct sockaddr_in *)&ifr.ifr_addr).sin_addr.s_addr =
1271                                                                 ifa->ifa_local;
1272
1273                 if (copy_to_user(buf + done, &ifr, size)) {
1274                         done = -EFAULT;
1275                         break;
1276                 }
1277                 len  -= size;
1278                 done += size;
1279         }
1280 out:
1281         return done;
1282 }
1283
1284 static __be32 in_dev_select_addr(const struct in_device *in_dev,
1285                                  int scope)
1286 {
1287         const struct in_ifaddr *ifa;
1288
1289         in_dev_for_each_ifa_rcu(ifa, in_dev) {
1290                 if (ifa->ifa_flags & IFA_F_SECONDARY)
1291                         continue;
1292                 if (ifa->ifa_scope != RT_SCOPE_LINK &&
1293                     ifa->ifa_scope <= scope)
1294                         return ifa->ifa_local;
1295         }
1296
1297         return 0;
1298 }
1299
1300 __be32 inet_select_addr(const struct net_device *dev, __be32 dst, int scope)
1301 {
1302         const struct in_ifaddr *ifa;
1303         __be32 addr = 0;
1304         unsigned char localnet_scope = RT_SCOPE_HOST;
1305         struct in_device *in_dev;
1306         struct net *net = dev_net(dev);
1307         int master_idx;
1308
1309         rcu_read_lock();
1310         in_dev = __in_dev_get_rcu(dev);
1311         if (!in_dev)
1312                 goto no_in_dev;
1313
1314         if (unlikely(IN_DEV_ROUTE_LOCALNET(in_dev)))
1315                 localnet_scope = RT_SCOPE_LINK;
1316
1317         in_dev_for_each_ifa_rcu(ifa, in_dev) {
1318                 if (ifa->ifa_flags & IFA_F_SECONDARY)
1319                         continue;
1320                 if (min(ifa->ifa_scope, localnet_scope) > scope)
1321                         continue;
1322                 if (!dst || inet_ifa_match(dst, ifa)) {
1323                         addr = ifa->ifa_local;
1324                         break;
1325                 }
1326                 if (!addr)
1327                         addr = ifa->ifa_local;
1328         }
1329
1330         if (addr)
1331                 goto out_unlock;
1332 no_in_dev:
1333         master_idx = l3mdev_master_ifindex_rcu(dev);
1334
1335         /* For VRFs, the VRF device takes the place of the loopback device,
1336          * with addresses on it being preferred.  Note in such cases the
1337          * loopback device will be among the devices that fail the master_idx
1338          * equality check in the loop below.
1339          */
1340         if (master_idx &&
1341             (dev = dev_get_by_index_rcu(net, master_idx)) &&
1342             (in_dev = __in_dev_get_rcu(dev))) {
1343                 addr = in_dev_select_addr(in_dev, scope);
1344                 if (addr)
1345                         goto out_unlock;
1346         }
1347
1348         /* Not loopback addresses on loopback should be preferred
1349            in this case. It is important that lo is the first interface
1350            in dev_base list.
1351          */
1352         for_each_netdev_rcu(net, dev) {
1353                 if (l3mdev_master_ifindex_rcu(dev) != master_idx)
1354                         continue;
1355
1356                 in_dev = __in_dev_get_rcu(dev);
1357                 if (!in_dev)
1358                         continue;
1359
1360                 addr = in_dev_select_addr(in_dev, scope);
1361                 if (addr)
1362                         goto out_unlock;
1363         }
1364 out_unlock:
1365         rcu_read_unlock();
1366         return addr;
1367 }
1368 EXPORT_SYMBOL(inet_select_addr);
1369
1370 static __be32 confirm_addr_indev(struct in_device *in_dev, __be32 dst,
1371                               __be32 local, int scope)
1372 {
1373         unsigned char localnet_scope = RT_SCOPE_HOST;
1374         const struct in_ifaddr *ifa;
1375         __be32 addr = 0;
1376         int same = 0;
1377
1378         if (unlikely(IN_DEV_ROUTE_LOCALNET(in_dev)))
1379                 localnet_scope = RT_SCOPE_LINK;
1380
1381         in_dev_for_each_ifa_rcu(ifa, in_dev) {
1382                 unsigned char min_scope = min(ifa->ifa_scope, localnet_scope);
1383
1384                 if (!addr &&
1385                     (local == ifa->ifa_local || !local) &&
1386                     min_scope <= scope) {
1387                         addr = ifa->ifa_local;
1388                         if (same)
1389                                 break;
1390                 }
1391                 if (!same) {
1392                         same = (!local || inet_ifa_match(local, ifa)) &&
1393                                 (!dst || inet_ifa_match(dst, ifa));
1394                         if (same && addr) {
1395                                 if (local || !dst)
1396                                         break;
1397                                 /* Is the selected addr into dst subnet? */
1398                                 if (inet_ifa_match(addr, ifa))
1399                                         break;
1400                                 /* No, then can we use new local src? */
1401                                 if (min_scope <= scope) {
1402                                         addr = ifa->ifa_local;
1403                                         break;
1404                                 }
1405                                 /* search for large dst subnet for addr */
1406                                 same = 0;
1407                         }
1408                 }
1409         }
1410
1411         return same ? addr : 0;
1412 }
1413
1414 /*
1415  * Confirm that local IP address exists using wildcards:
1416  * - net: netns to check, cannot be NULL
1417  * - in_dev: only on this interface, NULL=any interface
1418  * - dst: only in the same subnet as dst, 0=any dst
1419  * - local: address, 0=autoselect the local address
1420  * - scope: maximum allowed scope value for the local address
1421  */
1422 __be32 inet_confirm_addr(struct net *net, struct in_device *in_dev,
1423                          __be32 dst, __be32 local, int scope)
1424 {
1425         __be32 addr = 0;
1426         struct net_device *dev;
1427
1428         if (in_dev)
1429                 return confirm_addr_indev(in_dev, dst, local, scope);
1430
1431         rcu_read_lock();
1432         for_each_netdev_rcu(net, dev) {
1433                 in_dev = __in_dev_get_rcu(dev);
1434                 if (in_dev) {
1435                         addr = confirm_addr_indev(in_dev, dst, local, scope);
1436                         if (addr)
1437                                 break;
1438                 }
1439         }
1440         rcu_read_unlock();
1441
1442         return addr;
1443 }
1444 EXPORT_SYMBOL(inet_confirm_addr);
1445
1446 /*
1447  *      Device notifier
1448  */
1449
1450 int register_inetaddr_notifier(struct notifier_block *nb)
1451 {
1452         return blocking_notifier_chain_register(&inetaddr_chain, nb);
1453 }
1454 EXPORT_SYMBOL(register_inetaddr_notifier);
1455
1456 int unregister_inetaddr_notifier(struct notifier_block *nb)
1457 {
1458         return blocking_notifier_chain_unregister(&inetaddr_chain, nb);
1459 }
1460 EXPORT_SYMBOL(unregister_inetaddr_notifier);
1461
1462 int register_inetaddr_validator_notifier(struct notifier_block *nb)
1463 {
1464         return blocking_notifier_chain_register(&inetaddr_validator_chain, nb);
1465 }
1466 EXPORT_SYMBOL(register_inetaddr_validator_notifier);
1467
1468 int unregister_inetaddr_validator_notifier(struct notifier_block *nb)
1469 {
1470         return blocking_notifier_chain_unregister(&inetaddr_validator_chain,
1471             nb);
1472 }
1473 EXPORT_SYMBOL(unregister_inetaddr_validator_notifier);
1474
1475 /* Rename ifa_labels for a device name change. Make some effort to preserve
1476  * existing alias numbering and to create unique labels if possible.
1477 */
1478 static void inetdev_changename(struct net_device *dev, struct in_device *in_dev)
1479 {
1480         struct in_ifaddr *ifa;
1481         int named = 0;
1482
1483         in_dev_for_each_ifa_rtnl(ifa, in_dev) {
1484                 char old[IFNAMSIZ], *dot;
1485
1486                 memcpy(old, ifa->ifa_label, IFNAMSIZ);
1487                 memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
1488                 if (named++ == 0)
1489                         goto skip;
1490                 dot = strchr(old, ':');
1491                 if (!dot) {
1492                         sprintf(old, ":%d", named);
1493                         dot = old;
1494                 }
1495                 if (strlen(dot) + strlen(dev->name) < IFNAMSIZ)
1496                         strcat(ifa->ifa_label, dot);
1497                 else
1498                         strcpy(ifa->ifa_label + (IFNAMSIZ - strlen(dot) - 1), dot);
1499 skip:
1500                 rtmsg_ifa(RTM_NEWADDR, ifa, NULL, 0);
1501         }
1502 }
1503
1504 static void inetdev_send_gratuitous_arp(struct net_device *dev,
1505                                         struct in_device *in_dev)
1506
1507 {
1508         const struct in_ifaddr *ifa;
1509
1510         in_dev_for_each_ifa_rtnl(ifa, in_dev) {
1511                 arp_send(ARPOP_REQUEST, ETH_P_ARP,
1512                          ifa->ifa_local, dev,
1513                          ifa->ifa_local, NULL,
1514                          dev->dev_addr, NULL);
1515         }
1516 }
1517
1518 /* Called only under RTNL semaphore */
1519
1520 static int inetdev_event(struct notifier_block *this, unsigned long event,
1521                          void *ptr)
1522 {
1523         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1524         struct in_device *in_dev = __in_dev_get_rtnl(dev);
1525
1526         ASSERT_RTNL();
1527
1528         if (!in_dev) {
1529                 if (event == NETDEV_REGISTER) {
1530                         in_dev = inetdev_init(dev);
1531                         if (IS_ERR(in_dev))
1532                                 return notifier_from_errno(PTR_ERR(in_dev));
1533                         if (dev->flags & IFF_LOOPBACK) {
1534                                 IN_DEV_CONF_SET(in_dev, NOXFRM, 1);
1535                                 IN_DEV_CONF_SET(in_dev, NOPOLICY, 1);
1536                         }
1537                 } else if (event == NETDEV_CHANGEMTU) {
1538                         /* Re-enabling IP */
1539                         if (inetdev_valid_mtu(dev->mtu))
1540                                 in_dev = inetdev_init(dev);
1541                 }
1542                 goto out;
1543         }
1544
1545         switch (event) {
1546         case NETDEV_REGISTER:
1547                 pr_debug("%s: bug\n", __func__);
1548                 RCU_INIT_POINTER(dev->ip_ptr, NULL);
1549                 break;
1550         case NETDEV_UP:
1551                 if (!inetdev_valid_mtu(dev->mtu))
1552                         break;
1553                 if (dev->flags & IFF_LOOPBACK) {
1554                         struct in_ifaddr *ifa = inet_alloc_ifa();
1555
1556                         if (ifa) {
1557                                 INIT_HLIST_NODE(&ifa->hash);
1558                                 ifa->ifa_local =
1559                                   ifa->ifa_address = htonl(INADDR_LOOPBACK);
1560                                 ifa->ifa_prefixlen = 8;
1561                                 ifa->ifa_mask = inet_make_mask(8);
1562                                 in_dev_hold(in_dev);
1563                                 ifa->ifa_dev = in_dev;
1564                                 ifa->ifa_scope = RT_SCOPE_HOST;
1565                                 memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
1566                                 set_ifa_lifetime(ifa, INFINITY_LIFE_TIME,
1567                                                  INFINITY_LIFE_TIME);
1568                                 ipv4_devconf_setall(in_dev);
1569                                 neigh_parms_data_state_setall(in_dev->arp_parms);
1570                                 inet_insert_ifa(ifa);
1571                         }
1572                 }
1573                 ip_mc_up(in_dev);
1574                 fallthrough;
1575         case NETDEV_CHANGEADDR:
1576                 if (!IN_DEV_ARP_NOTIFY(in_dev))
1577                         break;
1578                 fallthrough;
1579         case NETDEV_NOTIFY_PEERS:
1580                 /* Send gratuitous ARP to notify of link change */
1581                 inetdev_send_gratuitous_arp(dev, in_dev);
1582                 break;
1583         case NETDEV_DOWN:
1584                 ip_mc_down(in_dev);
1585                 break;
1586         case NETDEV_PRE_TYPE_CHANGE:
1587                 ip_mc_unmap(in_dev);
1588                 break;
1589         case NETDEV_POST_TYPE_CHANGE:
1590                 ip_mc_remap(in_dev);
1591                 break;
1592         case NETDEV_CHANGEMTU:
1593                 if (inetdev_valid_mtu(dev->mtu))
1594                         break;
1595                 /* disable IP when MTU is not enough */
1596                 fallthrough;
1597         case NETDEV_UNREGISTER:
1598                 inetdev_destroy(in_dev);
1599                 break;
1600         case NETDEV_CHANGENAME:
1601                 /* Do not notify about label change, this event is
1602                  * not interesting to applications using netlink.
1603                  */
1604                 inetdev_changename(dev, in_dev);
1605
1606                 devinet_sysctl_unregister(in_dev);
1607                 devinet_sysctl_register(in_dev);
1608                 break;
1609         }
1610 out:
1611         return NOTIFY_DONE;
1612 }
1613
1614 static struct notifier_block ip_netdev_notifier = {
1615         .notifier_call = inetdev_event,
1616 };
1617
1618 static size_t inet_nlmsg_size(void)
1619 {
1620         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
1621                + nla_total_size(4) /* IFA_ADDRESS */
1622                + nla_total_size(4) /* IFA_LOCAL */
1623                + nla_total_size(4) /* IFA_BROADCAST */
1624                + nla_total_size(IFNAMSIZ) /* IFA_LABEL */
1625                + nla_total_size(4)  /* IFA_FLAGS */
1626                + nla_total_size(4)  /* IFA_RT_PRIORITY */
1627                + nla_total_size(sizeof(struct ifa_cacheinfo)); /* IFA_CACHEINFO */
1628 }
1629
1630 static inline u32 cstamp_delta(unsigned long cstamp)
1631 {
1632         return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
1633 }
1634
1635 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
1636                          unsigned long tstamp, u32 preferred, u32 valid)
1637 {
1638         struct ifa_cacheinfo ci;
1639
1640         ci.cstamp = cstamp_delta(cstamp);
1641         ci.tstamp = cstamp_delta(tstamp);
1642         ci.ifa_prefered = preferred;
1643         ci.ifa_valid = valid;
1644
1645         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
1646 }
1647
1648 static int inet_fill_ifaddr(struct sk_buff *skb, struct in_ifaddr *ifa,
1649                             struct inet_fill_args *args)
1650 {
1651         struct ifaddrmsg *ifm;
1652         struct nlmsghdr  *nlh;
1653         u32 preferred, valid;
1654
1655         nlh = nlmsg_put(skb, args->portid, args->seq, args->event, sizeof(*ifm),
1656                         args->flags);
1657         if (!nlh)
1658                 return -EMSGSIZE;
1659
1660         ifm = nlmsg_data(nlh);
1661         ifm->ifa_family = AF_INET;
1662         ifm->ifa_prefixlen = ifa->ifa_prefixlen;
1663         ifm->ifa_flags = ifa->ifa_flags;
1664         ifm->ifa_scope = ifa->ifa_scope;
1665         ifm->ifa_index = ifa->ifa_dev->dev->ifindex;
1666
1667         if (args->netnsid >= 0 &&
1668             nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
1669                 goto nla_put_failure;
1670
1671         if (!(ifm->ifa_flags & IFA_F_PERMANENT)) {
1672                 preferred = ifa->ifa_preferred_lft;
1673                 valid = ifa->ifa_valid_lft;
1674                 if (preferred != INFINITY_LIFE_TIME) {
1675                         long tval = (jiffies - ifa->ifa_tstamp) / HZ;
1676
1677                         if (preferred > tval)
1678                                 preferred -= tval;
1679                         else
1680                                 preferred = 0;
1681                         if (valid != INFINITY_LIFE_TIME) {
1682                                 if (valid > tval)
1683                                         valid -= tval;
1684                                 else
1685                                         valid = 0;
1686                         }
1687                 }
1688         } else {
1689                 preferred = INFINITY_LIFE_TIME;
1690                 valid = INFINITY_LIFE_TIME;
1691         }
1692         if ((ifa->ifa_address &&
1693              nla_put_in_addr(skb, IFA_ADDRESS, ifa->ifa_address)) ||
1694             (ifa->ifa_local &&
1695              nla_put_in_addr(skb, IFA_LOCAL, ifa->ifa_local)) ||
1696             (ifa->ifa_broadcast &&
1697              nla_put_in_addr(skb, IFA_BROADCAST, ifa->ifa_broadcast)) ||
1698             (ifa->ifa_label[0] &&
1699              nla_put_string(skb, IFA_LABEL, ifa->ifa_label)) ||
1700             nla_put_u32(skb, IFA_FLAGS, ifa->ifa_flags) ||
1701             (ifa->ifa_rt_priority &&
1702              nla_put_u32(skb, IFA_RT_PRIORITY, ifa->ifa_rt_priority)) ||
1703             put_cacheinfo(skb, ifa->ifa_cstamp, ifa->ifa_tstamp,
1704                           preferred, valid))
1705                 goto nla_put_failure;
1706
1707         nlmsg_end(skb, nlh);
1708         return 0;
1709
1710 nla_put_failure:
1711         nlmsg_cancel(skb, nlh);
1712         return -EMSGSIZE;
1713 }
1714
1715 static int inet_valid_dump_ifaddr_req(const struct nlmsghdr *nlh,
1716                                       struct inet_fill_args *fillargs,
1717                                       struct net **tgt_net, struct sock *sk,
1718                                       struct netlink_callback *cb)
1719 {
1720         struct netlink_ext_ack *extack = cb->extack;
1721         struct nlattr *tb[IFA_MAX+1];
1722         struct ifaddrmsg *ifm;
1723         int err, i;
1724
1725         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
1726                 NL_SET_ERR_MSG(extack, "ipv4: Invalid header for address dump request");
1727                 return -EINVAL;
1728         }
1729
1730         ifm = nlmsg_data(nlh);
1731         if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
1732                 NL_SET_ERR_MSG(extack, "ipv4: Invalid values in header for address dump request");
1733                 return -EINVAL;
1734         }
1735
1736         fillargs->ifindex = ifm->ifa_index;
1737         if (fillargs->ifindex) {
1738                 cb->answer_flags |= NLM_F_DUMP_FILTERED;
1739                 fillargs->flags |= NLM_F_DUMP_FILTERED;
1740         }
1741
1742         err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
1743                                             ifa_ipv4_policy, extack);
1744         if (err < 0)
1745                 return err;
1746
1747         for (i = 0; i <= IFA_MAX; ++i) {
1748                 if (!tb[i])
1749                         continue;
1750
1751                 if (i == IFA_TARGET_NETNSID) {
1752                         struct net *net;
1753
1754                         fillargs->netnsid = nla_get_s32(tb[i]);
1755
1756                         net = rtnl_get_net_ns_capable(sk, fillargs->netnsid);
1757                         if (IS_ERR(net)) {
1758                                 fillargs->netnsid = -1;
1759                                 NL_SET_ERR_MSG(extack, "ipv4: Invalid target network namespace id");
1760                                 return PTR_ERR(net);
1761                         }
1762                         *tgt_net = net;
1763                 } else {
1764                         NL_SET_ERR_MSG(extack, "ipv4: Unsupported attribute in dump request");
1765                         return -EINVAL;
1766                 }
1767         }
1768
1769         return 0;
1770 }
1771
1772 static int in_dev_dump_addr(struct in_device *in_dev, struct sk_buff *skb,
1773                             struct netlink_callback *cb, int s_ip_idx,
1774                             struct inet_fill_args *fillargs)
1775 {
1776         struct in_ifaddr *ifa;
1777         int ip_idx = 0;
1778         int err;
1779
1780         in_dev_for_each_ifa_rtnl(ifa, in_dev) {
1781                 if (ip_idx < s_ip_idx) {
1782                         ip_idx++;
1783                         continue;
1784                 }
1785                 err = inet_fill_ifaddr(skb, ifa, fillargs);
1786                 if (err < 0)
1787                         goto done;
1788
1789                 nl_dump_check_consistent(cb, nlmsg_hdr(skb));
1790                 ip_idx++;
1791         }
1792         err = 0;
1793
1794 done:
1795         cb->args[2] = ip_idx;
1796
1797         return err;
1798 }
1799
1800 static int inet_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
1801 {
1802         const struct nlmsghdr *nlh = cb->nlh;
1803         struct inet_fill_args fillargs = {
1804                 .portid = NETLINK_CB(cb->skb).portid,
1805                 .seq = nlh->nlmsg_seq,
1806                 .event = RTM_NEWADDR,
1807                 .flags = NLM_F_MULTI,
1808                 .netnsid = -1,
1809         };
1810         struct net *net = sock_net(skb->sk);
1811         struct net *tgt_net = net;
1812         int h, s_h;
1813         int idx, s_idx;
1814         int s_ip_idx;
1815         struct net_device *dev;
1816         struct in_device *in_dev;
1817         struct hlist_head *head;
1818         int err = 0;
1819
1820         s_h = cb->args[0];
1821         s_idx = idx = cb->args[1];
1822         s_ip_idx = cb->args[2];
1823
1824         if (cb->strict_check) {
1825                 err = inet_valid_dump_ifaddr_req(nlh, &fillargs, &tgt_net,
1826                                                  skb->sk, cb);
1827                 if (err < 0)
1828                         goto put_tgt_net;
1829
1830                 err = 0;
1831                 if (fillargs.ifindex) {
1832                         dev = __dev_get_by_index(tgt_net, fillargs.ifindex);
1833                         if (!dev) {
1834                                 err = -ENODEV;
1835                                 goto put_tgt_net;
1836                         }
1837
1838                         in_dev = __in_dev_get_rtnl(dev);
1839                         if (in_dev) {
1840                                 err = in_dev_dump_addr(in_dev, skb, cb, s_ip_idx,
1841                                                        &fillargs);
1842                         }
1843                         goto put_tgt_net;
1844                 }
1845         }
1846
1847         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
1848                 idx = 0;
1849                 head = &tgt_net->dev_index_head[h];
1850                 rcu_read_lock();
1851                 cb->seq = atomic_read(&tgt_net->ipv4.dev_addr_genid) ^
1852                           tgt_net->dev_base_seq;
1853                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
1854                         if (idx < s_idx)
1855                                 goto cont;
1856                         if (h > s_h || idx > s_idx)
1857                                 s_ip_idx = 0;
1858                         in_dev = __in_dev_get_rcu(dev);
1859                         if (!in_dev)
1860                                 goto cont;
1861
1862                         err = in_dev_dump_addr(in_dev, skb, cb, s_ip_idx,
1863                                                &fillargs);
1864                         if (err < 0) {
1865                                 rcu_read_unlock();
1866                                 goto done;
1867                         }
1868 cont:
1869                         idx++;
1870                 }
1871                 rcu_read_unlock();
1872         }
1873
1874 done:
1875         cb->args[0] = h;
1876         cb->args[1] = idx;
1877 put_tgt_net:
1878         if (fillargs.netnsid >= 0)
1879                 put_net(tgt_net);
1880
1881         return skb->len ? : err;
1882 }
1883
1884 static void rtmsg_ifa(int event, struct in_ifaddr *ifa, struct nlmsghdr *nlh,
1885                       u32 portid)
1886 {
1887         struct inet_fill_args fillargs = {
1888                 .portid = portid,
1889                 .seq = nlh ? nlh->nlmsg_seq : 0,
1890                 .event = event,
1891                 .flags = 0,
1892                 .netnsid = -1,
1893         };
1894         struct sk_buff *skb;
1895         int err = -ENOBUFS;
1896         struct net *net;
1897
1898         net = dev_net(ifa->ifa_dev->dev);
1899         skb = nlmsg_new(inet_nlmsg_size(), GFP_KERNEL);
1900         if (!skb)
1901                 goto errout;
1902
1903         err = inet_fill_ifaddr(skb, ifa, &fillargs);
1904         if (err < 0) {
1905                 /* -EMSGSIZE implies BUG in inet_nlmsg_size() */
1906                 WARN_ON(err == -EMSGSIZE);
1907                 kfree_skb(skb);
1908                 goto errout;
1909         }
1910         rtnl_notify(skb, net, portid, RTNLGRP_IPV4_IFADDR, nlh, GFP_KERNEL);
1911         return;
1912 errout:
1913         if (err < 0)
1914                 rtnl_set_sk_err(net, RTNLGRP_IPV4_IFADDR, err);
1915 }
1916
1917 static size_t inet_get_link_af_size(const struct net_device *dev,
1918                                     u32 ext_filter_mask)
1919 {
1920         struct in_device *in_dev = rcu_dereference_rtnl(dev->ip_ptr);
1921
1922         if (!in_dev)
1923                 return 0;
1924
1925         return nla_total_size(IPV4_DEVCONF_MAX * 4); /* IFLA_INET_CONF */
1926 }
1927
1928 static int inet_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
1929                              u32 ext_filter_mask)
1930 {
1931         struct in_device *in_dev = rcu_dereference_rtnl(dev->ip_ptr);
1932         struct nlattr *nla;
1933         int i;
1934
1935         if (!in_dev)
1936                 return -ENODATA;
1937
1938         nla = nla_reserve(skb, IFLA_INET_CONF, IPV4_DEVCONF_MAX * 4);
1939         if (!nla)
1940                 return -EMSGSIZE;
1941
1942         for (i = 0; i < IPV4_DEVCONF_MAX; i++)
1943                 ((u32 *) nla_data(nla))[i] = in_dev->cnf.data[i];
1944
1945         return 0;
1946 }
1947
1948 static const struct nla_policy inet_af_policy[IFLA_INET_MAX+1] = {
1949         [IFLA_INET_CONF]        = { .type = NLA_NESTED },
1950 };
1951
1952 static int inet_validate_link_af(const struct net_device *dev,
1953                                  const struct nlattr *nla,
1954                                  struct netlink_ext_ack *extack)
1955 {
1956         struct nlattr *a, *tb[IFLA_INET_MAX+1];
1957         int err, rem;
1958
1959         if (dev && !__in_dev_get_rtnl(dev))
1960                 return -EAFNOSUPPORT;
1961
1962         err = nla_parse_nested_deprecated(tb, IFLA_INET_MAX, nla,
1963                                           inet_af_policy, extack);
1964         if (err < 0)
1965                 return err;
1966
1967         if (tb[IFLA_INET_CONF]) {
1968                 nla_for_each_nested(a, tb[IFLA_INET_CONF], rem) {
1969                         int cfgid = nla_type(a);
1970
1971                         if (nla_len(a) < 4)
1972                                 return -EINVAL;
1973
1974                         if (cfgid <= 0 || cfgid > IPV4_DEVCONF_MAX)
1975                                 return -EINVAL;
1976                 }
1977         }
1978
1979         return 0;
1980 }
1981
1982 static int inet_set_link_af(struct net_device *dev, const struct nlattr *nla,
1983                             struct netlink_ext_ack *extack)
1984 {
1985         struct in_device *in_dev = __in_dev_get_rtnl(dev);
1986         struct nlattr *a, *tb[IFLA_INET_MAX+1];
1987         int rem;
1988
1989         if (!in_dev)
1990                 return -EAFNOSUPPORT;
1991
1992         if (nla_parse_nested_deprecated(tb, IFLA_INET_MAX, nla, NULL, NULL) < 0)
1993                 return -EINVAL;
1994
1995         if (tb[IFLA_INET_CONF]) {
1996                 nla_for_each_nested(a, tb[IFLA_INET_CONF], rem)
1997                         ipv4_devconf_set(in_dev, nla_type(a), nla_get_u32(a));
1998         }
1999
2000         return 0;
2001 }
2002
2003 static int inet_netconf_msgsize_devconf(int type)
2004 {
2005         int size = NLMSG_ALIGN(sizeof(struct netconfmsg))
2006                    + nla_total_size(4); /* NETCONFA_IFINDEX */
2007         bool all = false;
2008
2009         if (type == NETCONFA_ALL)
2010                 all = true;
2011
2012         if (all || type == NETCONFA_FORWARDING)
2013                 size += nla_total_size(4);
2014         if (all || type == NETCONFA_RP_FILTER)
2015                 size += nla_total_size(4);
2016         if (all || type == NETCONFA_MC_FORWARDING)
2017                 size += nla_total_size(4);
2018         if (all || type == NETCONFA_BC_FORWARDING)
2019                 size += nla_total_size(4);
2020         if (all || type == NETCONFA_PROXY_NEIGH)
2021                 size += nla_total_size(4);
2022         if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
2023                 size += nla_total_size(4);
2024
2025         return size;
2026 }
2027
2028 static int inet_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
2029                                      struct ipv4_devconf *devconf, u32 portid,
2030                                      u32 seq, int event, unsigned int flags,
2031                                      int type)
2032 {
2033         struct nlmsghdr  *nlh;
2034         struct netconfmsg *ncm;
2035         bool all = false;
2036
2037         nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
2038                         flags);
2039         if (!nlh)
2040                 return -EMSGSIZE;
2041
2042         if (type == NETCONFA_ALL)
2043                 all = true;
2044
2045         ncm = nlmsg_data(nlh);
2046         ncm->ncm_family = AF_INET;
2047
2048         if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
2049                 goto nla_put_failure;
2050
2051         if (!devconf)
2052                 goto out;
2053
2054         if ((all || type == NETCONFA_FORWARDING) &&
2055             nla_put_s32(skb, NETCONFA_FORWARDING,
2056                         IPV4_DEVCONF(*devconf, FORWARDING)) < 0)
2057                 goto nla_put_failure;
2058         if ((all || type == NETCONFA_RP_FILTER) &&
2059             nla_put_s32(skb, NETCONFA_RP_FILTER,
2060                         IPV4_DEVCONF(*devconf, RP_FILTER)) < 0)
2061                 goto nla_put_failure;
2062         if ((all || type == NETCONFA_MC_FORWARDING) &&
2063             nla_put_s32(skb, NETCONFA_MC_FORWARDING,
2064                         IPV4_DEVCONF(*devconf, MC_FORWARDING)) < 0)
2065                 goto nla_put_failure;
2066         if ((all || type == NETCONFA_BC_FORWARDING) &&
2067             nla_put_s32(skb, NETCONFA_BC_FORWARDING,
2068                         IPV4_DEVCONF(*devconf, BC_FORWARDING)) < 0)
2069                 goto nla_put_failure;
2070         if ((all || type == NETCONFA_PROXY_NEIGH) &&
2071             nla_put_s32(skb, NETCONFA_PROXY_NEIGH,
2072                         IPV4_DEVCONF(*devconf, PROXY_ARP)) < 0)
2073                 goto nla_put_failure;
2074         if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
2075             nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
2076                         IPV4_DEVCONF(*devconf, IGNORE_ROUTES_WITH_LINKDOWN)) < 0)
2077                 goto nla_put_failure;
2078
2079 out:
2080         nlmsg_end(skb, nlh);
2081         return 0;
2082
2083 nla_put_failure:
2084         nlmsg_cancel(skb, nlh);
2085         return -EMSGSIZE;
2086 }
2087
2088 void inet_netconf_notify_devconf(struct net *net, int event, int type,
2089                                  int ifindex, struct ipv4_devconf *devconf)
2090 {
2091         struct sk_buff *skb;
2092         int err = -ENOBUFS;
2093
2094         skb = nlmsg_new(inet_netconf_msgsize_devconf(type), GFP_KERNEL);
2095         if (!skb)
2096                 goto errout;
2097
2098         err = inet_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
2099                                         event, 0, type);
2100         if (err < 0) {
2101                 /* -EMSGSIZE implies BUG in inet_netconf_msgsize_devconf() */
2102                 WARN_ON(err == -EMSGSIZE);
2103                 kfree_skb(skb);
2104                 goto errout;
2105         }
2106         rtnl_notify(skb, net, 0, RTNLGRP_IPV4_NETCONF, NULL, GFP_KERNEL);
2107         return;
2108 errout:
2109         if (err < 0)
2110                 rtnl_set_sk_err(net, RTNLGRP_IPV4_NETCONF, err);
2111 }
2112
2113 static const struct nla_policy devconf_ipv4_policy[NETCONFA_MAX+1] = {
2114         [NETCONFA_IFINDEX]      = { .len = sizeof(int) },
2115         [NETCONFA_FORWARDING]   = { .len = sizeof(int) },
2116         [NETCONFA_RP_FILTER]    = { .len = sizeof(int) },
2117         [NETCONFA_PROXY_NEIGH]  = { .len = sizeof(int) },
2118         [NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN]  = { .len = sizeof(int) },
2119 };
2120
2121 static int inet_netconf_valid_get_req(struct sk_buff *skb,
2122                                       const struct nlmsghdr *nlh,
2123                                       struct nlattr **tb,
2124                                       struct netlink_ext_ack *extack)
2125 {
2126         int i, err;
2127
2128         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(struct netconfmsg))) {
2129                 NL_SET_ERR_MSG(extack, "ipv4: Invalid header for netconf get request");
2130                 return -EINVAL;
2131         }
2132
2133         if (!netlink_strict_get_check(skb))
2134                 return nlmsg_parse_deprecated(nlh, sizeof(struct netconfmsg),
2135                                               tb, NETCONFA_MAX,
2136                                               devconf_ipv4_policy, extack);
2137
2138         err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct netconfmsg),
2139                                             tb, NETCONFA_MAX,
2140                                             devconf_ipv4_policy, extack);
2141         if (err)
2142                 return err;
2143
2144         for (i = 0; i <= NETCONFA_MAX; i++) {
2145                 if (!tb[i])
2146                         continue;
2147
2148                 switch (i) {
2149                 case NETCONFA_IFINDEX:
2150                         break;
2151                 default:
2152                         NL_SET_ERR_MSG(extack, "ipv4: Unsupported attribute in netconf get request");
2153                         return -EINVAL;
2154                 }
2155         }
2156
2157         return 0;
2158 }
2159
2160 static int inet_netconf_get_devconf(struct sk_buff *in_skb,
2161                                     struct nlmsghdr *nlh,
2162                                     struct netlink_ext_ack *extack)
2163 {
2164         struct net *net = sock_net(in_skb->sk);
2165         struct nlattr *tb[NETCONFA_MAX+1];
2166         struct sk_buff *skb;
2167         struct ipv4_devconf *devconf;
2168         struct in_device *in_dev;
2169         struct net_device *dev;
2170         int ifindex;
2171         int err;
2172
2173         err = inet_netconf_valid_get_req(in_skb, nlh, tb, extack);
2174         if (err)
2175                 goto errout;
2176
2177         err = -EINVAL;
2178         if (!tb[NETCONFA_IFINDEX])
2179                 goto errout;
2180
2181         ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
2182         switch (ifindex) {
2183         case NETCONFA_IFINDEX_ALL:
2184                 devconf = net->ipv4.devconf_all;
2185                 break;
2186         case NETCONFA_IFINDEX_DEFAULT:
2187                 devconf = net->ipv4.devconf_dflt;
2188                 break;
2189         default:
2190                 dev = __dev_get_by_index(net, ifindex);
2191                 if (!dev)
2192                         goto errout;
2193                 in_dev = __in_dev_get_rtnl(dev);
2194                 if (!in_dev)
2195                         goto errout;
2196                 devconf = &in_dev->cnf;
2197                 break;
2198         }
2199
2200         err = -ENOBUFS;
2201         skb = nlmsg_new(inet_netconf_msgsize_devconf(NETCONFA_ALL), GFP_KERNEL);
2202         if (!skb)
2203                 goto errout;
2204
2205         err = inet_netconf_fill_devconf(skb, ifindex, devconf,
2206                                         NETLINK_CB(in_skb).portid,
2207                                         nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
2208                                         NETCONFA_ALL);
2209         if (err < 0) {
2210                 /* -EMSGSIZE implies BUG in inet_netconf_msgsize_devconf() */
2211                 WARN_ON(err == -EMSGSIZE);
2212                 kfree_skb(skb);
2213                 goto errout;
2214         }
2215         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
2216 errout:
2217         return err;
2218 }
2219
2220 static int inet_netconf_dump_devconf(struct sk_buff *skb,
2221                                      struct netlink_callback *cb)
2222 {
2223         const struct nlmsghdr *nlh = cb->nlh;
2224         struct net *net = sock_net(skb->sk);
2225         int h, s_h;
2226         int idx, s_idx;
2227         struct net_device *dev;
2228         struct in_device *in_dev;
2229         struct hlist_head *head;
2230
2231         if (cb->strict_check) {
2232                 struct netlink_ext_ack *extack = cb->extack;
2233                 struct netconfmsg *ncm;
2234
2235                 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ncm))) {
2236                         NL_SET_ERR_MSG(extack, "ipv4: Invalid header for netconf dump request");
2237                         return -EINVAL;
2238                 }
2239
2240                 if (nlmsg_attrlen(nlh, sizeof(*ncm))) {
2241                         NL_SET_ERR_MSG(extack, "ipv4: Invalid data after header in netconf dump request");
2242                         return -EINVAL;
2243                 }
2244         }
2245
2246         s_h = cb->args[0];
2247         s_idx = idx = cb->args[1];
2248
2249         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
2250                 idx = 0;
2251                 head = &net->dev_index_head[h];
2252                 rcu_read_lock();
2253                 cb->seq = atomic_read(&net->ipv4.dev_addr_genid) ^
2254                           net->dev_base_seq;
2255                 hlist_for_each_entry_rcu(dev, head, index_hlist) {
2256                         if (idx < s_idx)
2257                                 goto cont;
2258                         in_dev = __in_dev_get_rcu(dev);
2259                         if (!in_dev)
2260                                 goto cont;
2261
2262                         if (inet_netconf_fill_devconf(skb, dev->ifindex,
2263                                                       &in_dev->cnf,
2264                                                       NETLINK_CB(cb->skb).portid,
2265                                                       nlh->nlmsg_seq,
2266                                                       RTM_NEWNETCONF,
2267                                                       NLM_F_MULTI,
2268                                                       NETCONFA_ALL) < 0) {
2269                                 rcu_read_unlock();
2270                                 goto done;
2271                         }
2272                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
2273 cont:
2274                         idx++;
2275                 }
2276                 rcu_read_unlock();
2277         }
2278         if (h == NETDEV_HASHENTRIES) {
2279                 if (inet_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
2280                                               net->ipv4.devconf_all,
2281                                               NETLINK_CB(cb->skb).portid,
2282                                               nlh->nlmsg_seq,
2283                                               RTM_NEWNETCONF, NLM_F_MULTI,
2284                                               NETCONFA_ALL) < 0)
2285                         goto done;
2286                 else
2287                         h++;
2288         }
2289         if (h == NETDEV_HASHENTRIES + 1) {
2290                 if (inet_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
2291                                               net->ipv4.devconf_dflt,
2292                                               NETLINK_CB(cb->skb).portid,
2293                                               nlh->nlmsg_seq,
2294                                               RTM_NEWNETCONF, NLM_F_MULTI,
2295                                               NETCONFA_ALL) < 0)
2296                         goto done;
2297                 else
2298                         h++;
2299         }
2300 done:
2301         cb->args[0] = h;
2302         cb->args[1] = idx;
2303
2304         return skb->len;
2305 }
2306
2307 #ifdef CONFIG_SYSCTL
2308
2309 static void devinet_copy_dflt_conf(struct net *net, int i)
2310 {
2311         struct net_device *dev;
2312
2313         rcu_read_lock();
2314         for_each_netdev_rcu(net, dev) {
2315                 struct in_device *in_dev;
2316
2317                 in_dev = __in_dev_get_rcu(dev);
2318                 if (in_dev && !test_bit(i, in_dev->cnf.state))
2319                         in_dev->cnf.data[i] = net->ipv4.devconf_dflt->data[i];
2320         }
2321         rcu_read_unlock();
2322 }
2323
2324 /* called with RTNL locked */
2325 static void inet_forward_change(struct net *net)
2326 {
2327         struct net_device *dev;
2328         int on = IPV4_DEVCONF_ALL(net, FORWARDING);
2329
2330         IPV4_DEVCONF_ALL(net, ACCEPT_REDIRECTS) = !on;
2331         IPV4_DEVCONF_DFLT(net, FORWARDING) = on;
2332         inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
2333                                     NETCONFA_FORWARDING,
2334                                     NETCONFA_IFINDEX_ALL,
2335                                     net->ipv4.devconf_all);
2336         inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
2337                                     NETCONFA_FORWARDING,
2338                                     NETCONFA_IFINDEX_DEFAULT,
2339                                     net->ipv4.devconf_dflt);
2340
2341         for_each_netdev(net, dev) {
2342                 struct in_device *in_dev;
2343
2344                 if (on)
2345                         dev_disable_lro(dev);
2346
2347                 in_dev = __in_dev_get_rtnl(dev);
2348                 if (in_dev) {
2349                         IN_DEV_CONF_SET(in_dev, FORWARDING, on);
2350                         inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
2351                                                     NETCONFA_FORWARDING,
2352                                                     dev->ifindex, &in_dev->cnf);
2353                 }
2354         }
2355 }
2356
2357 static int devinet_conf_ifindex(struct net *net, struct ipv4_devconf *cnf)
2358 {
2359         if (cnf == net->ipv4.devconf_dflt)
2360                 return NETCONFA_IFINDEX_DEFAULT;
2361         else if (cnf == net->ipv4.devconf_all)
2362                 return NETCONFA_IFINDEX_ALL;
2363         else {
2364                 struct in_device *idev
2365                         = container_of(cnf, struct in_device, cnf);
2366                 return idev->dev->ifindex;
2367         }
2368 }
2369
2370 static int devinet_conf_proc(struct ctl_table *ctl, int write,
2371                              void *buffer, size_t *lenp, loff_t *ppos)
2372 {
2373         int old_value = *(int *)ctl->data;
2374         int ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
2375         int new_value = *(int *)ctl->data;
2376
2377         if (write) {
2378                 struct ipv4_devconf *cnf = ctl->extra1;
2379                 struct net *net = ctl->extra2;
2380                 int i = (int *)ctl->data - cnf->data;
2381                 int ifindex;
2382
2383                 set_bit(i, cnf->state);
2384
2385                 if (cnf == net->ipv4.devconf_dflt)
2386                         devinet_copy_dflt_conf(net, i);
2387                 if (i == IPV4_DEVCONF_ACCEPT_LOCAL - 1 ||
2388                     i == IPV4_DEVCONF_ROUTE_LOCALNET - 1)
2389                         if ((new_value == 0) && (old_value != 0))
2390                                 rt_cache_flush(net);
2391
2392                 if (i == IPV4_DEVCONF_BC_FORWARDING - 1 &&
2393                     new_value != old_value)
2394                         rt_cache_flush(net);
2395
2396                 if (i == IPV4_DEVCONF_RP_FILTER - 1 &&
2397                     new_value != old_value) {
2398                         ifindex = devinet_conf_ifindex(net, cnf);
2399                         inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
2400                                                     NETCONFA_RP_FILTER,
2401                                                     ifindex, cnf);
2402                 }
2403                 if (i == IPV4_DEVCONF_PROXY_ARP - 1 &&
2404                     new_value != old_value) {
2405                         ifindex = devinet_conf_ifindex(net, cnf);
2406                         inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
2407                                                     NETCONFA_PROXY_NEIGH,
2408                                                     ifindex, cnf);
2409                 }
2410                 if (i == IPV4_DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN - 1 &&
2411                     new_value != old_value) {
2412                         ifindex = devinet_conf_ifindex(net, cnf);
2413                         inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
2414                                                     NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
2415                                                     ifindex, cnf);
2416                 }
2417         }
2418
2419         return ret;
2420 }
2421
2422 static int devinet_sysctl_forward(struct ctl_table *ctl, int write,
2423                                   void *buffer, size_t *lenp, loff_t *ppos)
2424 {
2425         int *valp = ctl->data;
2426         int val = *valp;
2427         loff_t pos = *ppos;
2428         struct net *net = ctl->extra2;
2429         int ret;
2430
2431         if (write && !ns_capable(net->user_ns, CAP_NET_ADMIN))
2432                 return -EPERM;
2433
2434         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
2435
2436         if (write && *valp != val) {
2437                 if (valp != &IPV4_DEVCONF_DFLT(net, FORWARDING)) {
2438                         if (!rtnl_trylock()) {
2439                                 /* Restore the original values before restarting */
2440                                 *valp = val;
2441                                 *ppos = pos;
2442                                 return restart_syscall();
2443                         }
2444                         if (valp == &IPV4_DEVCONF_ALL(net, FORWARDING)) {
2445                                 inet_forward_change(net);
2446                         } else {
2447                                 struct ipv4_devconf *cnf = ctl->extra1;
2448                                 struct in_device *idev =
2449                                         container_of(cnf, struct in_device, cnf);
2450                                 if (*valp)
2451                                         dev_disable_lro(idev->dev);
2452                                 inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
2453                                                             NETCONFA_FORWARDING,
2454                                                             idev->dev->ifindex,
2455                                                             cnf);
2456                         }
2457                         rtnl_unlock();
2458                         rt_cache_flush(net);
2459                 } else
2460                         inet_netconf_notify_devconf(net, RTM_NEWNETCONF,
2461                                                     NETCONFA_FORWARDING,
2462                                                     NETCONFA_IFINDEX_DEFAULT,
2463                                                     net->ipv4.devconf_dflt);
2464         }
2465
2466         return ret;
2467 }
2468
2469 static int ipv4_doint_and_flush(struct ctl_table *ctl, int write,
2470                                 void *buffer, size_t *lenp, loff_t *ppos)
2471 {
2472         int *valp = ctl->data;
2473         int val = *valp;
2474         int ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
2475         struct net *net = ctl->extra2;
2476
2477         if (write && *valp != val)
2478                 rt_cache_flush(net);
2479
2480         return ret;
2481 }
2482
2483 #define DEVINET_SYSCTL_ENTRY(attr, name, mval, proc) \
2484         { \
2485                 .procname       = name, \
2486                 .data           = ipv4_devconf.data + \
2487                                   IPV4_DEVCONF_ ## attr - 1, \
2488                 .maxlen         = sizeof(int), \
2489                 .mode           = mval, \
2490                 .proc_handler   = proc, \
2491                 .extra1         = &ipv4_devconf, \
2492         }
2493
2494 #define DEVINET_SYSCTL_RW_ENTRY(attr, name) \
2495         DEVINET_SYSCTL_ENTRY(attr, name, 0644, devinet_conf_proc)
2496
2497 #define DEVINET_SYSCTL_RO_ENTRY(attr, name) \
2498         DEVINET_SYSCTL_ENTRY(attr, name, 0444, devinet_conf_proc)
2499
2500 #define DEVINET_SYSCTL_COMPLEX_ENTRY(attr, name, proc) \
2501         DEVINET_SYSCTL_ENTRY(attr, name, 0644, proc)
2502
2503 #define DEVINET_SYSCTL_FLUSHING_ENTRY(attr, name) \
2504         DEVINET_SYSCTL_COMPLEX_ENTRY(attr, name, ipv4_doint_and_flush)
2505
2506 static struct devinet_sysctl_table {
2507         struct ctl_table_header *sysctl_header;
2508         struct ctl_table devinet_vars[__IPV4_DEVCONF_MAX];
2509 } devinet_sysctl = {
2510         .devinet_vars = {
2511                 DEVINET_SYSCTL_COMPLEX_ENTRY(FORWARDING, "forwarding",
2512                                              devinet_sysctl_forward),
2513                 DEVINET_SYSCTL_RO_ENTRY(MC_FORWARDING, "mc_forwarding"),
2514                 DEVINET_SYSCTL_RW_ENTRY(BC_FORWARDING, "bc_forwarding"),
2515
2516                 DEVINET_SYSCTL_RW_ENTRY(ACCEPT_REDIRECTS, "accept_redirects"),
2517                 DEVINET_SYSCTL_RW_ENTRY(SECURE_REDIRECTS, "secure_redirects"),
2518                 DEVINET_SYSCTL_RW_ENTRY(SHARED_MEDIA, "shared_media"),
2519                 DEVINET_SYSCTL_RW_ENTRY(RP_FILTER, "rp_filter"),
2520                 DEVINET_SYSCTL_RW_ENTRY(SEND_REDIRECTS, "send_redirects"),
2521                 DEVINET_SYSCTL_RW_ENTRY(ACCEPT_SOURCE_ROUTE,
2522                                         "accept_source_route"),
2523                 DEVINET_SYSCTL_RW_ENTRY(ACCEPT_LOCAL, "accept_local"),
2524                 DEVINET_SYSCTL_RW_ENTRY(SRC_VMARK, "src_valid_mark"),
2525                 DEVINET_SYSCTL_RW_ENTRY(PROXY_ARP, "proxy_arp"),
2526                 DEVINET_SYSCTL_RW_ENTRY(MEDIUM_ID, "medium_id"),
2527                 DEVINET_SYSCTL_RW_ENTRY(BOOTP_RELAY, "bootp_relay"),
2528                 DEVINET_SYSCTL_RW_ENTRY(LOG_MARTIANS, "log_martians"),
2529                 DEVINET_SYSCTL_RW_ENTRY(TAG, "tag"),
2530                 DEVINET_SYSCTL_RW_ENTRY(ARPFILTER, "arp_filter"),
2531                 DEVINET_SYSCTL_RW_ENTRY(ARP_ANNOUNCE, "arp_announce"),
2532                 DEVINET_SYSCTL_RW_ENTRY(ARP_IGNORE, "arp_ignore"),
2533                 DEVINET_SYSCTL_RW_ENTRY(ARP_ACCEPT, "arp_accept"),
2534                 DEVINET_SYSCTL_RW_ENTRY(ARP_NOTIFY, "arp_notify"),
2535                 DEVINET_SYSCTL_RW_ENTRY(PROXY_ARP_PVLAN, "proxy_arp_pvlan"),
2536                 DEVINET_SYSCTL_RW_ENTRY(FORCE_IGMP_VERSION,
2537                                         "force_igmp_version"),
2538                 DEVINET_SYSCTL_RW_ENTRY(IGMPV2_UNSOLICITED_REPORT_INTERVAL,
2539                                         "igmpv2_unsolicited_report_interval"),
2540                 DEVINET_SYSCTL_RW_ENTRY(IGMPV3_UNSOLICITED_REPORT_INTERVAL,
2541                                         "igmpv3_unsolicited_report_interval"),
2542                 DEVINET_SYSCTL_RW_ENTRY(IGNORE_ROUTES_WITH_LINKDOWN,
2543                                         "ignore_routes_with_linkdown"),
2544                 DEVINET_SYSCTL_RW_ENTRY(DROP_GRATUITOUS_ARP,
2545                                         "drop_gratuitous_arp"),
2546
2547                 DEVINET_SYSCTL_FLUSHING_ENTRY(NOXFRM, "disable_xfrm"),
2548                 DEVINET_SYSCTL_FLUSHING_ENTRY(NOPOLICY, "disable_policy"),
2549                 DEVINET_SYSCTL_FLUSHING_ENTRY(PROMOTE_SECONDARIES,
2550                                               "promote_secondaries"),
2551                 DEVINET_SYSCTL_FLUSHING_ENTRY(ROUTE_LOCALNET,
2552                                               "route_localnet"),
2553                 DEVINET_SYSCTL_FLUSHING_ENTRY(DROP_UNICAST_IN_L2_MULTICAST,
2554                                               "drop_unicast_in_l2_multicast"),
2555         },
2556 };
2557
2558 static int __devinet_sysctl_register(struct net *net, char *dev_name,
2559                                      int ifindex, struct ipv4_devconf *p)
2560 {
2561         int i;
2562         struct devinet_sysctl_table *t;
2563         char path[sizeof("net/ipv4/conf/") + IFNAMSIZ];
2564
2565         t = kmemdup(&devinet_sysctl, sizeof(*t), GFP_KERNEL);
2566         if (!t)
2567                 goto out;
2568
2569         for (i = 0; i < ARRAY_SIZE(t->devinet_vars) - 1; i++) {
2570                 t->devinet_vars[i].data += (char *)p - (char *)&ipv4_devconf;
2571                 t->devinet_vars[i].extra1 = p;
2572                 t->devinet_vars[i].extra2 = net;
2573         }
2574
2575         snprintf(path, sizeof(path), "net/ipv4/conf/%s", dev_name);
2576
2577         t->sysctl_header = register_net_sysctl(net, path, t->devinet_vars);
2578         if (!t->sysctl_header)
2579                 goto free;
2580
2581         p->sysctl = t;
2582
2583         inet_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
2584                                     ifindex, p);
2585         return 0;
2586
2587 free:
2588         kfree(t);
2589 out:
2590         return -ENOBUFS;
2591 }
2592
2593 static void __devinet_sysctl_unregister(struct net *net,
2594                                         struct ipv4_devconf *cnf, int ifindex)
2595 {
2596         struct devinet_sysctl_table *t = cnf->sysctl;
2597
2598         if (t) {
2599                 cnf->sysctl = NULL;
2600                 unregister_net_sysctl_table(t->sysctl_header);
2601                 kfree(t);
2602         }
2603
2604         inet_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
2605 }
2606
2607 static int devinet_sysctl_register(struct in_device *idev)
2608 {
2609         int err;
2610
2611         if (!sysctl_dev_name_is_allowed(idev->dev->name))
2612                 return -EINVAL;
2613
2614         err = neigh_sysctl_register(idev->dev, idev->arp_parms, NULL);
2615         if (err)
2616                 return err;
2617         err = __devinet_sysctl_register(dev_net(idev->dev), idev->dev->name,
2618                                         idev->dev->ifindex, &idev->cnf);
2619         if (err)
2620                 neigh_sysctl_unregister(idev->arp_parms);
2621         return err;
2622 }
2623
2624 static void devinet_sysctl_unregister(struct in_device *idev)
2625 {
2626         struct net *net = dev_net(idev->dev);
2627
2628         __devinet_sysctl_unregister(net, &idev->cnf, idev->dev->ifindex);
2629         neigh_sysctl_unregister(idev->arp_parms);
2630 }
2631
2632 static struct ctl_table ctl_forward_entry[] = {
2633         {
2634                 .procname       = "ip_forward",
2635                 .data           = &ipv4_devconf.data[
2636                                         IPV4_DEVCONF_FORWARDING - 1],
2637                 .maxlen         = sizeof(int),
2638                 .mode           = 0644,
2639                 .proc_handler   = devinet_sysctl_forward,
2640                 .extra1         = &ipv4_devconf,
2641                 .extra2         = &init_net,
2642         },
2643         { },
2644 };
2645 #endif
2646
2647 static __net_init int devinet_init_net(struct net *net)
2648 {
2649         int err;
2650         struct ipv4_devconf *all, *dflt;
2651 #ifdef CONFIG_SYSCTL
2652         struct ctl_table *tbl;
2653         struct ctl_table_header *forw_hdr;
2654 #endif
2655
2656         err = -ENOMEM;
2657         all = kmemdup(&ipv4_devconf, sizeof(ipv4_devconf), GFP_KERNEL);
2658         if (!all)
2659                 goto err_alloc_all;
2660
2661         dflt = kmemdup(&ipv4_devconf_dflt, sizeof(ipv4_devconf_dflt), GFP_KERNEL);
2662         if (!dflt)
2663                 goto err_alloc_dflt;
2664
2665 #ifdef CONFIG_SYSCTL
2666         tbl = kmemdup(ctl_forward_entry, sizeof(ctl_forward_entry), GFP_KERNEL);
2667         if (!tbl)
2668                 goto err_alloc_ctl;
2669
2670         tbl[0].data = &all->data[IPV4_DEVCONF_FORWARDING - 1];
2671         tbl[0].extra1 = all;
2672         tbl[0].extra2 = net;
2673 #endif
2674
2675         if (!net_eq(net, &init_net)) {
2676                 if (IS_ENABLED(CONFIG_SYSCTL) &&
2677                     sysctl_devconf_inherit_init_net == 3) {
2678                         /* copy from the current netns */
2679                         memcpy(all, current->nsproxy->net_ns->ipv4.devconf_all,
2680                                sizeof(ipv4_devconf));
2681                         memcpy(dflt,
2682                                current->nsproxy->net_ns->ipv4.devconf_dflt,
2683                                sizeof(ipv4_devconf_dflt));
2684                 } else if (!IS_ENABLED(CONFIG_SYSCTL) ||
2685                            sysctl_devconf_inherit_init_net != 2) {
2686                         /* inherit == 0 or 1: copy from init_net */
2687                         memcpy(all, init_net.ipv4.devconf_all,
2688                                sizeof(ipv4_devconf));
2689                         memcpy(dflt, init_net.ipv4.devconf_dflt,
2690                                sizeof(ipv4_devconf_dflt));
2691                 }
2692                 /* else inherit == 2: use compiled values */
2693         }
2694
2695 #ifdef CONFIG_SYSCTL
2696         err = __devinet_sysctl_register(net, "all", NETCONFA_IFINDEX_ALL, all);
2697         if (err < 0)
2698                 goto err_reg_all;
2699
2700         err = __devinet_sysctl_register(net, "default",
2701                                         NETCONFA_IFINDEX_DEFAULT, dflt);
2702         if (err < 0)
2703                 goto err_reg_dflt;
2704
2705         err = -ENOMEM;
2706         forw_hdr = register_net_sysctl(net, "net/ipv4", tbl);
2707         if (!forw_hdr)
2708                 goto err_reg_ctl;
2709         net->ipv4.forw_hdr = forw_hdr;
2710 #endif
2711
2712         net->ipv4.devconf_all = all;
2713         net->ipv4.devconf_dflt = dflt;
2714         return 0;
2715
2716 #ifdef CONFIG_SYSCTL
2717 err_reg_ctl:
2718         __devinet_sysctl_unregister(net, dflt, NETCONFA_IFINDEX_DEFAULT);
2719 err_reg_dflt:
2720         __devinet_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
2721 err_reg_all:
2722         kfree(tbl);
2723 err_alloc_ctl:
2724 #endif
2725         kfree(dflt);
2726 err_alloc_dflt:
2727         kfree(all);
2728 err_alloc_all:
2729         return err;
2730 }
2731
2732 static __net_exit void devinet_exit_net(struct net *net)
2733 {
2734 #ifdef CONFIG_SYSCTL
2735         struct ctl_table *tbl;
2736
2737         tbl = net->ipv4.forw_hdr->ctl_table_arg;
2738         unregister_net_sysctl_table(net->ipv4.forw_hdr);
2739         __devinet_sysctl_unregister(net, net->ipv4.devconf_dflt,
2740                                     NETCONFA_IFINDEX_DEFAULT);
2741         __devinet_sysctl_unregister(net, net->ipv4.devconf_all,
2742                                     NETCONFA_IFINDEX_ALL);
2743         kfree(tbl);
2744 #endif
2745         kfree(net->ipv4.devconf_dflt);
2746         kfree(net->ipv4.devconf_all);
2747 }
2748
2749 static __net_initdata struct pernet_operations devinet_ops = {
2750         .init = devinet_init_net,
2751         .exit = devinet_exit_net,
2752 };
2753
2754 static struct rtnl_af_ops inet_af_ops __read_mostly = {
2755         .family           = AF_INET,
2756         .fill_link_af     = inet_fill_link_af,
2757         .get_link_af_size = inet_get_link_af_size,
2758         .validate_link_af = inet_validate_link_af,
2759         .set_link_af      = inet_set_link_af,
2760 };
2761
2762 void __init devinet_init(void)
2763 {
2764         int i;
2765
2766         for (i = 0; i < IN4_ADDR_HSIZE; i++)
2767                 INIT_HLIST_HEAD(&inet_addr_lst[i]);
2768
2769         register_pernet_subsys(&devinet_ops);
2770         register_netdevice_notifier(&ip_netdev_notifier);
2771
2772         queue_delayed_work(system_power_efficient_wq, &check_lifetime_work, 0);
2773
2774         rtnl_af_register(&inet_af_ops);
2775
2776         rtnl_register(PF_INET, RTM_NEWADDR, inet_rtm_newaddr, NULL, 0);
2777         rtnl_register(PF_INET, RTM_DELADDR, inet_rtm_deladdr, NULL, 0);
2778         rtnl_register(PF_INET, RTM_GETADDR, NULL, inet_dump_ifaddr, 0);
2779         rtnl_register(PF_INET, RTM_GETNETCONF, inet_netconf_get_devconf,
2780                       inet_netconf_dump_devconf, 0);
2781 }