git.oblomov.eu Git - linux-2.6/blob - net/ipv4/devinet.c

   1 /*
   2  *      NET3    IP device support routines.
   3  *
   4  *      Version: $Id: devinet.c,v 1.44 2001/10/31 21:55:54 davem Exp $
   5  *
   6  *              This program is free software; you can redistribute it and/or
   7  *              modify it under the terms of the GNU General Public License
   8  *              as published by the Free Software Foundation; either version
   9  *              2 of the License, or (at your option) any later version.
  10  *
  11  *      Derived from the IP parts of dev.c 1.0.19
  12  *              Authors:        Ross Biro
  13  *                              Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
  14  *                              Mark Evans, <evansmp@uhura.aston.ac.uk>
  15  *
  16  *      Additional Authors:
  17  *              Alan Cox, <gw4pts@gw4pts.ampr.org>
  18  *              Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
  19  *
  20  *      Changes:
  21  *              Alexey Kuznetsov:       pa_* fields are replaced with ifaddr
  22  *                                      lists.
  23  *              Cyrus Durgin:           updated for kmod
  24  *              Matthias Andree:        in devinet_ioctl, compare label and
  25  *                                      address (4.4BSD alias style support),
  26  *                                      fall back to comparing just the label
  27  *                                      if no match found.
  28  */
  29
  30
  31 #include <asm/uaccess.h>
  32 #include <asm/system.h>
  33 #include <linux/bitops.h>
  34 #include <linux/capability.h>
  35 #include <linux/module.h>
  36 #include <linux/types.h>
  37 #include <linux/kernel.h>
  38 #include <linux/string.h>
  39 #include <linux/mm.h>
  40 #include <linux/socket.h>
  41 #include <linux/sockios.h>
  42 #include <linux/in.h>
  43 #include <linux/errno.h>
  44 #include <linux/interrupt.h>
  45 #include <linux/if_addr.h>
  46 #include <linux/if_ether.h>
  47 #include <linux/inet.h>
  48 #include <linux/netdevice.h>
  49 #include <linux/etherdevice.h>
  50 #include <linux/skbuff.h>
  51 #include <linux/init.h>
  52 #include <linux/notifier.h>
  53 #include <linux/inetdevice.h>
  54 #include <linux/igmp.h>
  55 #ifdef CONFIG_SYSCTL
  56 #include <linux/sysctl.h>
  57 #endif
  58 #include <linux/kmod.h>
  59
  60 #include <net/arp.h>
  61 #include <net/ip.h>
  62 #include <net/route.h>
  63 #include <net/ip_fib.h>
  64 #include <net/rtnetlink.h>
  65
  66 struct ipv4_devconf ipv4_devconf = {
  67         .data = {
  68                 [NET_IPV4_CONF_ACCEPT_REDIRECTS - 1] = 1,
  69                 [NET_IPV4_CONF_SEND_REDIRECTS - 1] = 1,
  70                 [NET_IPV4_CONF_SECURE_REDIRECTS - 1] = 1,
  71                 [NET_IPV4_CONF_SHARED_MEDIA - 1] = 1,
  72         },
  73 };
  74
  75 static struct ipv4_devconf ipv4_devconf_dflt = {
  76         .data = {
  77                 [NET_IPV4_CONF_ACCEPT_REDIRECTS - 1] = 1,
  78                 [NET_IPV4_CONF_SEND_REDIRECTS - 1] = 1,
  79                 [NET_IPV4_CONF_SECURE_REDIRECTS - 1] = 1,
  80                 [NET_IPV4_CONF_SHARED_MEDIA - 1] = 1,
  81                 [NET_IPV4_CONF_ACCEPT_SOURCE_ROUTE - 1] = 1,
  82         },
  83 };
  84
  85 #define IPV4_DEVCONF_DFLT(attr) IPV4_DEVCONF(ipv4_devconf_dflt, attr)
  86
  87 static const struct nla_policy ifa_ipv4_policy[IFA_MAX+1] = {
  88         [IFA_LOCAL]             = { .type = NLA_U32 },
  89         [IFA_ADDRESS]           = { .type = NLA_U32 },
  90         [IFA_BROADCAST]         = { .type = NLA_U32 },
  91         [IFA_ANYCAST]           = { .type = NLA_U32 },
  92         [IFA_LABEL]             = { .type = NLA_STRING, .len = IFNAMSIZ - 1 },
  93 };
  94
  95 static void rtmsg_ifa(int event, struct in_ifaddr *, struct nlmsghdr *, u32);
  96
  97 static BLOCKING_NOTIFIER_HEAD(inetaddr_chain);
  98 static void inet_del_ifa(struct in_device *in_dev, struct in_ifaddr **ifap,
  99                          int destroy);
 100 #ifdef CONFIG_SYSCTL
 101 static void devinet_sysctl_register(struct in_device *idev);
 102 static void devinet_sysctl_unregister(struct ipv4_devconf *p);
 103 #endif
 104
 105 /* Locks all the inet devices. */
 106
 107 static struct in_ifaddr *inet_alloc_ifa(void)
 108 {
 109         struct in_ifaddr *ifa = kzalloc(sizeof(*ifa), GFP_KERNEL);
 110
 111         if (ifa) {
 112                 INIT_RCU_HEAD(&ifa->rcu_head);
 113         }
 114
 115         return ifa;
 116 }
 117
 118 static void inet_rcu_free_ifa(struct rcu_head *head)
 119 {
 120         struct in_ifaddr *ifa = container_of(head, struct in_ifaddr, rcu_head);
 121         if (ifa->ifa_dev)
 122                 in_dev_put(ifa->ifa_dev);
 123         kfree(ifa);
 124 }
 125
 126 static inline void inet_free_ifa(struct in_ifaddr *ifa)
 127 {
 128         call_rcu(&ifa->rcu_head, inet_rcu_free_ifa);
 129 }
 130
 131 void in_dev_finish_destroy(struct in_device *idev)
 132 {
 133         struct net_device *dev = idev->dev;
 134
 135         BUG_TRAP(!idev->ifa_list);
 136         BUG_TRAP(!idev->mc_list);
 137 #ifdef NET_REFCNT_DEBUG
 138         printk(KERN_DEBUG "in_dev_finish_destroy: %p=%s\n",
 139                idev, dev ? dev->name : "NIL");
 140 #endif
 141         dev_put(dev);
 142         if (!idev->dead)
 143                 printk("Freeing alive in_device %p\n", idev);
 144         else {
 145                 kfree(idev);
 146         }
 147 }
 148
 149 static struct in_device *inetdev_init(struct net_device *dev)
 150 {
 151         struct in_device *in_dev;
 152
 153         ASSERT_RTNL();
 154
 155         in_dev = kzalloc(sizeof(*in_dev), GFP_KERNEL);
 156         if (!in_dev)
 157                 goto out;
 158         INIT_RCU_HEAD(&in_dev->rcu_head);
 159         memcpy(&in_dev->cnf, &ipv4_devconf_dflt, sizeof(in_dev->cnf));
 160         in_dev->cnf.sysctl = NULL;
 161         in_dev->dev = dev;
 162         if ((in_dev->arp_parms = neigh_parms_alloc(dev, &arp_tbl)) == NULL)
 163                 goto out_kfree;
 164         /* Reference in_dev->dev */
 165         dev_hold(dev);
 166 #ifdef CONFIG_SYSCTL
 167         neigh_sysctl_register(dev, in_dev->arp_parms, NET_IPV4,
 168                               NET_IPV4_NEIGH, "ipv4", NULL, NULL);
 169 #endif
 170
 171         /* Account for reference dev->ip_ptr (below) */
 172         in_dev_hold(in_dev);
 173
 174 #ifdef CONFIG_SYSCTL
 175         devinet_sysctl_register(in_dev);
 176 #endif
 177         ip_mc_init_dev(in_dev);
 178         if (dev->flags & IFF_UP)
 179                 ip_mc_up(in_dev);
 180
 181         /* we can receive as soon as ip_ptr is set -- do this last */
 182         rcu_assign_pointer(dev->ip_ptr, in_dev);
 183 out:
 184         return in_dev;
 185 out_kfree:
 186         kfree(in_dev);
 187         in_dev = NULL;
 188         goto out;
 189 }
 190
 191 static void in_dev_rcu_put(struct rcu_head *head)
 192 {
 193         struct in_device *idev = container_of(head, struct in_device, rcu_head);
 194         in_dev_put(idev);
 195 }
 196
 197 static void inetdev_destroy(struct in_device *in_dev)
 198 {
 199         struct in_ifaddr *ifa;
 200         struct net_device *dev;
 201
 202         ASSERT_RTNL();
 203
 204         dev = in_dev->dev;
 205
 206         in_dev->dead = 1;
 207
 208         ip_mc_destroy_dev(in_dev);
 209
 210         while ((ifa = in_dev->ifa_list) != NULL) {
 211                 inet_del_ifa(in_dev, &in_dev->ifa_list, 0);
 212                 inet_free_ifa(ifa);
 213         }
 214
 215 #ifdef CONFIG_SYSCTL
 216         devinet_sysctl_unregister(&in_dev->cnf);
 217 #endif
 218
 219         dev->ip_ptr = NULL;
 220
 221 #ifdef CONFIG_SYSCTL
 222         neigh_sysctl_unregister(in_dev->arp_parms);
 223 #endif
 224         neigh_parms_release(&arp_tbl, in_dev->arp_parms);
 225         arp_ifdown(dev);
 226
 227         call_rcu(&in_dev->rcu_head, in_dev_rcu_put);
 228 }
 229
 230 int inet_addr_onlink(struct in_device *in_dev, __be32 a, __be32 b)
 231 {
 232         rcu_read_lock();
 233         for_primary_ifa(in_dev) {
 234                 if (inet_ifa_match(a, ifa)) {
 235                         if (!b || inet_ifa_match(b, ifa)) {
 236                                 rcu_read_unlock();
 237                                 return 1;
 238                         }
 239                 }
 240         } endfor_ifa(in_dev);
 241         rcu_read_unlock();
 242         return 0;
 243 }
 244
 245 static void __inet_del_ifa(struct in_device *in_dev, struct in_ifaddr **ifap,
 246                          int destroy, struct nlmsghdr *nlh, u32 pid)
 247 {
 248         struct in_ifaddr *promote = NULL;
 249         struct in_ifaddr *ifa, *ifa1 = *ifap;
 250         struct in_ifaddr *last_prim = in_dev->ifa_list;
 251         struct in_ifaddr *prev_prom = NULL;
 252         int do_promote = IN_DEV_PROMOTE_SECONDARIES(in_dev);
 253
 254         ASSERT_RTNL();
 255
 256         /* 1. Deleting primary ifaddr forces deletion all secondaries
 257          * unless alias promotion is set
 258          **/
 259
 260         if (!(ifa1->ifa_flags & IFA_F_SECONDARY)) {
 261                 struct in_ifaddr **ifap1 = &ifa1->ifa_next;
 262
 263                 while ((ifa = *ifap1) != NULL) {
 264                         if (!(ifa->ifa_flags & IFA_F_SECONDARY) &&
 265                             ifa1->ifa_scope <= ifa->ifa_scope)
 266                                 last_prim = ifa;
 267
 268                         if (!(ifa->ifa_flags & IFA_F_SECONDARY) ||
 269                             ifa1->ifa_mask != ifa->ifa_mask ||
 270                             !inet_ifa_match(ifa1->ifa_address, ifa)) {
 271                                 ifap1 = &ifa->ifa_next;
 272                                 prev_prom = ifa;
 273                                 continue;
 274                         }
 275
 276                         if (!do_promote) {
 277                                 *ifap1 = ifa->ifa_next;
 278
 279                                 rtmsg_ifa(RTM_DELADDR, ifa, nlh, pid);
 280                                 blocking_notifier_call_chain(&inetaddr_chain,
 281                                                 NETDEV_DOWN, ifa);
 282                                 inet_free_ifa(ifa);
 283                         } else {
 284                                 promote = ifa;
 285                                 break;
 286                         }
 287                 }
 288         }
 289
 290         /* 2. Unlink it */
 291
 292         *ifap = ifa1->ifa_next;
 293
 294         /* 3. Announce address deletion */
 295
 296         /* Send message first, then call notifier.
 297            At first sight, FIB update triggered by notifier
 298            will refer to already deleted ifaddr, that could confuse
 299            netlink listeners. It is not true: look, gated sees
 300            that route deleted and if it still thinks that ifaddr
 301            is valid, it will try to restore deleted routes... Grr.
 302            So that, this order is correct.
 303          */
 304         rtmsg_ifa(RTM_DELADDR, ifa1, nlh, pid);
 305         blocking_notifier_call_chain(&inetaddr_chain, NETDEV_DOWN, ifa1);
 306
 307         if (promote) {
 308
 309                 if (prev_prom) {
 310                         prev_prom->ifa_next = promote->ifa_next;
 311                         promote->ifa_next = last_prim->ifa_next;
 312                         last_prim->ifa_next = promote;
 313                 }
 314
 315                 promote->ifa_flags &= ~IFA_F_SECONDARY;
 316                 rtmsg_ifa(RTM_NEWADDR, promote, nlh, pid);
 317                 blocking_notifier_call_chain(&inetaddr_chain,
 318                                 NETDEV_UP, promote);
 319                 for (ifa = promote->ifa_next; ifa; ifa = ifa->ifa_next) {
 320                         if (ifa1->ifa_mask != ifa->ifa_mask ||
 321                             !inet_ifa_match(ifa1->ifa_address, ifa))
 322                                         continue;
 323                         fib_add_ifaddr(ifa);
 324                 }
 325
 326         }
 327         if (destroy)
 328                 inet_free_ifa(ifa1);
 329 }
 330
 331 static void inet_del_ifa(struct in_device *in_dev, struct in_ifaddr **ifap,
 332                          int destroy)
 333 {
 334         __inet_del_ifa(in_dev, ifap, destroy, NULL, 0);
 335 }
 336
 337 static int __inet_insert_ifa(struct in_ifaddr *ifa, struct nlmsghdr *nlh,
 338                              u32 pid)
 339 {
 340         struct in_device *in_dev = ifa->ifa_dev;
 341         struct in_ifaddr *ifa1, **ifap, **last_primary;
 342
 343         ASSERT_RTNL();
 344
 345         if (!ifa->ifa_local) {
 346                 inet_free_ifa(ifa);
 347                 return 0;
 348         }
 349
 350         ifa->ifa_flags &= ~IFA_F_SECONDARY;
 351         last_primary = &in_dev->ifa_list;
 352
 353         for (ifap = &in_dev->ifa_list; (ifa1 = *ifap) != NULL;
 354              ifap = &ifa1->ifa_next) {
 355                 if (!(ifa1->ifa_flags & IFA_F_SECONDARY) &&
 356                     ifa->ifa_scope <= ifa1->ifa_scope)
 357                         last_primary = &ifa1->ifa_next;
 358                 if (ifa1->ifa_mask == ifa->ifa_mask &&
 359                     inet_ifa_match(ifa1->ifa_address, ifa)) {
 360                         if (ifa1->ifa_local == ifa->ifa_local) {
 361                                 inet_free_ifa(ifa);
 362                                 return -EEXIST;
 363                         }
 364                         if (ifa1->ifa_scope != ifa->ifa_scope) {
 365                                 inet_free_ifa(ifa);
 366                                 return -EINVAL;
 367                         }
 368                         ifa->ifa_flags |= IFA_F_SECONDARY;
 369                 }
 370         }
 371
 372         if (!(ifa->ifa_flags & IFA_F_SECONDARY)) {
 373                 net_srandom(ifa->ifa_local);
 374                 ifap = last_primary;
 375         }
 376
 377         ifa->ifa_next = *ifap;
 378         *ifap = ifa;
 379
 380         /* Send message first, then call notifier.
 381            Notifier will trigger FIB update, so that
 382            listeners of netlink will know about new ifaddr */
 383         rtmsg_ifa(RTM_NEWADDR, ifa, nlh, pid);
 384         blocking_notifier_call_chain(&inetaddr_chain, NETDEV_UP, ifa);
 385
 386         return 0;
 387 }
 388
 389 static int inet_insert_ifa(struct in_ifaddr *ifa)
 390 {
 391         return __inet_insert_ifa(ifa, NULL, 0);
 392 }
 393
 394 static int inet_set_ifa(struct net_device *dev, struct in_ifaddr *ifa)
 395 {
 396         struct in_device *in_dev = __in_dev_get_rtnl(dev);
 397
 398         ASSERT_RTNL();
 399
 400         if (!in_dev) {
 401                 inet_free_ifa(ifa);
 402                 return -ENOBUFS;
 403         }
 404         ipv4_devconf_setall(in_dev);
 405         if (ifa->ifa_dev != in_dev) {
 406                 BUG_TRAP(!ifa->ifa_dev);
 407                 in_dev_hold(in_dev);
 408                 ifa->ifa_dev = in_dev;
 409         }
 410         if (LOOPBACK(ifa->ifa_local))
 411                 ifa->ifa_scope = RT_SCOPE_HOST;
 412         return inet_insert_ifa(ifa);
 413 }
 414
 415 struct in_device *inetdev_by_index(int ifindex)
 416 {
 417         struct net_device *dev;
 418         struct in_device *in_dev = NULL;
 419         read_lock(&dev_base_lock);
 420         dev = __dev_get_by_index(&init_net, ifindex);
 421         if (dev)
 422                 in_dev = in_dev_get(dev);
 423         read_unlock(&dev_base_lock);
 424         return in_dev;
 425 }
 426
 427 /* Called only from RTNL semaphored context. No locks. */
 428
 429 struct in_ifaddr *inet_ifa_byprefix(struct in_device *in_dev, __be32 prefix,
 430                                     __be32 mask)
 431 {
 432         ASSERT_RTNL();
 433
 434         for_primary_ifa(in_dev) {
 435                 if (ifa->ifa_mask == mask && inet_ifa_match(prefix, ifa))
 436                         return ifa;
 437         } endfor_ifa(in_dev);
 438         return NULL;
 439 }
 440
 441 static int inet_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
 442 {
 443         struct net *net = skb->sk->sk_net;
 444         struct nlattr *tb[IFA_MAX+1];
 445         struct in_device *in_dev;
 446         struct ifaddrmsg *ifm;
 447         struct in_ifaddr *ifa, **ifap;
 448         int err = -EINVAL;
 449
 450         ASSERT_RTNL();
 451
 452         if (net != &init_net)
 453                 return -EINVAL;
 454
 455         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv4_policy);
 456         if (err < 0)
 457                 goto errout;
 458
 459         ifm = nlmsg_data(nlh);
 460         in_dev = inetdev_by_index(ifm->ifa_index);
 461         if (in_dev == NULL) {
 462                 err = -ENODEV;
 463                 goto errout;
 464         }
 465
 466         __in_dev_put(in_dev);
 467
 468         for (ifap = &in_dev->ifa_list; (ifa = *ifap) != NULL;
 469              ifap = &ifa->ifa_next) {
 470                 if (tb[IFA_LOCAL] &&
 471                     ifa->ifa_local != nla_get_be32(tb[IFA_LOCAL]))
 472                         continue;
 473
 474                 if (tb[IFA_LABEL] && nla_strcmp(tb[IFA_LABEL], ifa->ifa_label))
 475                         continue;
 476
 477                 if (tb[IFA_ADDRESS] &&
 478                     (ifm->ifa_prefixlen != ifa->ifa_prefixlen ||
 479                     !inet_ifa_match(nla_get_be32(tb[IFA_ADDRESS]), ifa)))
 480                         continue;
 481
 482                 __inet_del_ifa(in_dev, ifap, 1, nlh, NETLINK_CB(skb).pid);
 483                 return 0;
 484         }
 485
 486         err = -EADDRNOTAVAIL;
 487 errout:
 488         return err;
 489 }
 490
 491 static struct in_ifaddr *rtm_to_ifaddr(struct nlmsghdr *nlh)
 492 {
 493         struct nlattr *tb[IFA_MAX+1];
 494         struct in_ifaddr *ifa;
 495         struct ifaddrmsg *ifm;
 496         struct net_device *dev;
 497         struct in_device *in_dev;
 498         int err = -EINVAL;
 499
 500         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv4_policy);
 501         if (err < 0)
 502                 goto errout;
 503
 504         ifm = nlmsg_data(nlh);
 505         if (ifm->ifa_prefixlen > 32 || tb[IFA_LOCAL] == NULL) {
 506                 err = -EINVAL;
 507                 goto errout;
 508         }
 509
 510         dev = __dev_get_by_index(&init_net, ifm->ifa_index);
 511         if (dev == NULL) {
 512                 err = -ENODEV;
 513                 goto errout;
 514         }
 515
 516         in_dev = __in_dev_get_rtnl(dev);
 517         if (in_dev == NULL) {
 518                 err = -ENOBUFS;
 519                 goto errout;
 520         }
 521
 522         ifa = inet_alloc_ifa();
 523         if (ifa == NULL) {
 524                 /*
 525                  * A potential indev allocation can be left alive, it stays
 526                  * assigned to its device and is destroy with it.
 527                  */
 528                 err = -ENOBUFS;
 529                 goto errout;
 530         }
 531
 532         ipv4_devconf_setall(in_dev);
 533         in_dev_hold(in_dev);
 534
 535         if (tb[IFA_ADDRESS] == NULL)
 536                 tb[IFA_ADDRESS] = tb[IFA_LOCAL];
 537
 538         ifa->ifa_prefixlen = ifm->ifa_prefixlen;
 539         ifa->ifa_mask = inet_make_mask(ifm->ifa_prefixlen);
 540         ifa->ifa_flags = ifm->ifa_flags;
 541         ifa->ifa_scope = ifm->ifa_scope;
 542         ifa->ifa_dev = in_dev;
 543
 544         ifa->ifa_local = nla_get_be32(tb[IFA_LOCAL]);
 545         ifa->ifa_address = nla_get_be32(tb[IFA_ADDRESS]);
 546
 547         if (tb[IFA_BROADCAST])
 548                 ifa->ifa_broadcast = nla_get_be32(tb[IFA_BROADCAST]);
 549
 550         if (tb[IFA_ANYCAST])
 551                 ifa->ifa_anycast = nla_get_be32(tb[IFA_ANYCAST]);
 552
 553         if (tb[IFA_LABEL])
 554                 nla_strlcpy(ifa->ifa_label, tb[IFA_LABEL], IFNAMSIZ);
 555         else
 556                 memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
 557
 558         return ifa;
 559
 560 errout:
 561         return ERR_PTR(err);
 562 }
 563
 564 static int inet_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
 565 {
 566         struct net *net = skb->sk->sk_net;
 567         struct in_ifaddr *ifa;
 568
 569         ASSERT_RTNL();
 570
 571         if (net != &init_net)
 572                 return -EINVAL;
 573
 574         ifa = rtm_to_ifaddr(nlh);
 575         if (IS_ERR(ifa))
 576                 return PTR_ERR(ifa);
 577
 578         return __inet_insert_ifa(ifa, nlh, NETLINK_CB(skb).pid);
 579 }
 580
 581 /*
 582  *      Determine a default network mask, based on the IP address.
 583  */
 584
 585 static __inline__ int inet_abc_len(__be32 addr)
 586 {
 587         int rc = -1;    /* Something else, probably a multicast. */
 588
 589         if (ZERONET(addr))
 590                 rc = 0;
 591         else {
 592                 __u32 haddr = ntohl(addr);
 593
 594                 if (IN_CLASSA(haddr))
 595                         rc = 8;
 596                 else if (IN_CLASSB(haddr))
 597                         rc = 16;
 598                 else if (IN_CLASSC(haddr))
 599                         rc = 24;
 600         }
 601
 602         return rc;
 603 }
 604
 605
 606 int devinet_ioctl(unsigned int cmd, void __user *arg)
 607 {
 608         struct ifreq ifr;
 609         struct sockaddr_in sin_orig;
 610         struct sockaddr_in *sin = (struct sockaddr_in *)&ifr.ifr_addr;
 611         struct in_device *in_dev;
 612         struct in_ifaddr **ifap = NULL;
 613         struct in_ifaddr *ifa = NULL;
 614         struct net_device *dev;
 615         char *colon;
 616         int ret = -EFAULT;
 617         int tryaddrmatch = 0;
 618
 619         /*
 620          *      Fetch the caller's info block into kernel space
 621          */
 622
 623         if (copy_from_user(&ifr, arg, sizeof(struct ifreq)))
 624                 goto out;
 625         ifr.ifr_name[IFNAMSIZ - 1] = 0;
 626
 627         /* save original address for comparison */
 628         memcpy(&sin_orig, sin, sizeof(*sin));
 629
 630         colon = strchr(ifr.ifr_name, ':');
 631         if (colon)
 632                 *colon = 0;
 633
 634 #ifdef CONFIG_KMOD
 635         dev_load(&init_net, ifr.ifr_name);
 636 #endif
 637
 638         switch (cmd) {
 639         case SIOCGIFADDR:       /* Get interface address */
 640         case SIOCGIFBRDADDR:    /* Get the broadcast address */
 641         case SIOCGIFDSTADDR:    /* Get the destination address */
 642         case SIOCGIFNETMASK:    /* Get the netmask for the interface */
 643                 /* Note that these ioctls will not sleep,
 644                    so that we do not impose a lock.
 645                    One day we will be forced to put shlock here (I mean SMP)
 646                  */
 647                 tryaddrmatch = (sin_orig.sin_family == AF_INET);
 648                 memset(sin, 0, sizeof(*sin));
 649                 sin->sin_family = AF_INET;
 650                 break;
 651
 652         case SIOCSIFFLAGS:
 653                 ret = -EACCES;
 654                 if (!capable(CAP_NET_ADMIN))
 655                         goto out;
 656                 break;
 657         case SIOCSIFADDR:       /* Set interface address (and family) */
 658         case SIOCSIFBRDADDR:    /* Set the broadcast address */
 659         case SIOCSIFDSTADDR:    /* Set the destination address */
 660         case SIOCSIFNETMASK:    /* Set the netmask for the interface */
 661                 ret = -EACCES;
 662                 if (!capable(CAP_NET_ADMIN))
 663                         goto out;
 664                 ret = -EINVAL;
 665                 if (sin->sin_family != AF_INET)
 666                         goto out;
 667                 break;
 668         default:
 669                 ret = -EINVAL;
 670                 goto out;
 671         }
 672
 673         rtnl_lock();
 674
 675         ret = -ENODEV;
 676         if ((dev = __dev_get_by_name(&init_net, ifr.ifr_name)) == NULL)
 677                 goto done;
 678
 679         if (colon)
 680                 *colon = ':';
 681
 682         if ((in_dev = __in_dev_get_rtnl(dev)) != NULL) {
 683                 if (tryaddrmatch) {
 684                         /* Matthias Andree */
 685                         /* compare label and address (4.4BSD style) */
 686                         /* note: we only do this for a limited set of ioctls
 687                            and only if the original address family was AF_INET.
 688                            This is checked above. */
 689                         for (ifap = &in_dev->ifa_list; (ifa = *ifap) != NULL;
 690                              ifap = &ifa->ifa_next) {
 691                                 if (!strcmp(ifr.ifr_name, ifa->ifa_label) &&
 692                                     sin_orig.sin_addr.s_addr ==
 693                                                         ifa->ifa_address) {
 694                                         break; /* found */
 695                                 }
 696                         }
 697                 }
 698                 /* we didn't get a match, maybe the application is
 699                    4.3BSD-style and passed in junk so we fall back to
 700                    comparing just the label */
 701                 if (!ifa) {
 702                         for (ifap = &in_dev->ifa_list; (ifa = *ifap) != NULL;
 703                              ifap = &ifa->ifa_next)
 704                                 if (!strcmp(ifr.ifr_name, ifa->ifa_label))
 705                                         break;
 706                 }
 707         }
 708
 709         ret = -EADDRNOTAVAIL;
 710         if (!ifa && cmd != SIOCSIFADDR && cmd != SIOCSIFFLAGS)
 711                 goto done;
 712
 713         switch (cmd) {
 714         case SIOCGIFADDR:       /* Get interface address */
 715                 sin->sin_addr.s_addr = ifa->ifa_local;
 716                 goto rarok;
 717
 718         case SIOCGIFBRDADDR:    /* Get the broadcast address */
 719                 sin->sin_addr.s_addr = ifa->ifa_broadcast;
 720                 goto rarok;
 721
 722         case SIOCGIFDSTADDR:    /* Get the destination address */
 723                 sin->sin_addr.s_addr = ifa->ifa_address;
 724                 goto rarok;
 725
 726         case SIOCGIFNETMASK:    /* Get the netmask for the interface */
 727                 sin->sin_addr.s_addr = ifa->ifa_mask;
 728                 goto rarok;
 729
 730         case SIOCSIFFLAGS:
 731                 if (colon) {
 732                         ret = -EADDRNOTAVAIL;
 733                         if (!ifa)
 734                                 break;
 735                         ret = 0;
 736                         if (!(ifr.ifr_flags & IFF_UP))
 737                                 inet_del_ifa(in_dev, ifap, 1);
 738                         break;
 739                 }
 740                 ret = dev_change_flags(dev, ifr.ifr_flags);
 741                 break;
 742
 743         case SIOCSIFADDR:       /* Set interface address (and family) */
 744                 ret = -EINVAL;
 745                 if (inet_abc_len(sin->sin_addr.s_addr) < 0)
 746                         break;
 747
 748                 if (!ifa) {
 749                         ret = -ENOBUFS;
 750                         if ((ifa = inet_alloc_ifa()) == NULL)
 751                                 break;
 752                         if (colon)
 753                                 memcpy(ifa->ifa_label, ifr.ifr_name, IFNAMSIZ);
 754                         else
 755                                 memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
 756                 } else {
 757                         ret = 0;
 758                         if (ifa->ifa_local == sin->sin_addr.s_addr)
 759                                 break;
 760                         inet_del_ifa(in_dev, ifap, 0);
 761                         ifa->ifa_broadcast = 0;
 762                         ifa->ifa_anycast = 0;
 763                 }
 764
 765                 ifa->ifa_address = ifa->ifa_local = sin->sin_addr.s_addr;
 766
 767                 if (!(dev->flags & IFF_POINTOPOINT)) {
 768                         ifa->ifa_prefixlen = inet_abc_len(ifa->ifa_address);
 769                         ifa->ifa_mask = inet_make_mask(ifa->ifa_prefixlen);
 770                         if ((dev->flags & IFF_BROADCAST) &&
 771                             ifa->ifa_prefixlen < 31)
 772                                 ifa->ifa_broadcast = ifa->ifa_address |
 773                                                      ~ifa->ifa_mask;
 774                 } else {
 775                         ifa->ifa_prefixlen = 32;
 776                         ifa->ifa_mask = inet_make_mask(32);
 777                 }
 778                 ret = inet_set_ifa(dev, ifa);
 779                 break;
 780
 781         case SIOCSIFBRDADDR:    /* Set the broadcast address */
 782                 ret = 0;
 783                 if (ifa->ifa_broadcast != sin->sin_addr.s_addr) {
 784                         inet_del_ifa(in_dev, ifap, 0);
 785                         ifa->ifa_broadcast = sin->sin_addr.s_addr;
 786                         inet_insert_ifa(ifa);
 787                 }
 788                 break;
 789
 790         case SIOCSIFDSTADDR:    /* Set the destination address */
 791                 ret = 0;
 792                 if (ifa->ifa_address == sin->sin_addr.s_addr)
 793                         break;
 794                 ret = -EINVAL;
 795                 if (inet_abc_len(sin->sin_addr.s_addr) < 0)
 796                         break;
 797                 ret = 0;
 798                 inet_del_ifa(in_dev, ifap, 0);
 799                 ifa->ifa_address = sin->sin_addr.s_addr;
 800                 inet_insert_ifa(ifa);
 801                 break;
 802
 803         case SIOCSIFNETMASK:    /* Set the netmask for the interface */
 804
 805                 /*
 806                  *      The mask we set must be legal.
 807                  */
 808                 ret = -EINVAL;
 809                 if (bad_mask(sin->sin_addr.s_addr, 0))
 810                         break;
 811                 ret = 0;
 812                 if (ifa->ifa_mask != sin->sin_addr.s_addr) {
 813                         __be32 old_mask = ifa->ifa_mask;
 814                         inet_del_ifa(in_dev, ifap, 0);
 815                         ifa->ifa_mask = sin->sin_addr.s_addr;
 816                         ifa->ifa_prefixlen = inet_mask_len(ifa->ifa_mask);
 817
 818                         /* See if current broadcast address matches
 819                          * with current netmask, then recalculate
 820                          * the broadcast address. Otherwise it's a
 821                          * funny address, so don't touch it since
 822                          * the user seems to know what (s)he's doing...
 823                          */
 824                         if ((dev->flags & IFF_BROADCAST) &&
 825                             (ifa->ifa_prefixlen < 31) &&
 826                             (ifa->ifa_broadcast ==
 827                              (ifa->ifa_local|~old_mask))) {
 828                                 ifa->ifa_broadcast = (ifa->ifa_local |
 829                                                       ~sin->sin_addr.s_addr);
 830                         }
 831                         inet_insert_ifa(ifa);
 832                 }
 833                 break;
 834         }
 835 done:
 836         rtnl_unlock();
 837 out:
 838         return ret;
 839 rarok:
 840         rtnl_unlock();
 841         ret = copy_to_user(arg, &ifr, sizeof(struct ifreq)) ? -EFAULT : 0;
 842         goto out;
 843 }
 844
 845 static int inet_gifconf(struct net_device *dev, char __user *buf, int len)
 846 {
 847         struct in_device *in_dev = __in_dev_get_rtnl(dev);
 848         struct in_ifaddr *ifa;
 849         struct ifreq ifr;
 850         int done = 0;
 851
 852         if (!in_dev || (ifa = in_dev->ifa_list) == NULL)
 853                 goto out;
 854
 855         for (; ifa; ifa = ifa->ifa_next) {
 856                 if (!buf) {
 857                         done += sizeof(ifr);
 858                         continue;
 859                 }
 860                 if (len < (int) sizeof(ifr))
 861                         break;
 862                 memset(&ifr, 0, sizeof(struct ifreq));
 863                 if (ifa->ifa_label)
 864                         strcpy(ifr.ifr_name, ifa->ifa_label);
 865                 else
 866                         strcpy(ifr.ifr_name, dev->name);
 867
 868                 (*(struct sockaddr_in *)&ifr.ifr_addr).sin_family = AF_INET;
 869                 (*(struct sockaddr_in *)&ifr.ifr_addr).sin_addr.s_addr =
 870                                                                 ifa->ifa_local;
 871
 872                 if (copy_to_user(buf, &ifr, sizeof(struct ifreq))) {
 873                         done = -EFAULT;
 874                         break;
 875                 }
 876                 buf  += sizeof(struct ifreq);
 877                 len  -= sizeof(struct ifreq);
 878                 done += sizeof(struct ifreq);
 879         }
 880 out:
 881         return done;
 882 }
 883
 884 __be32 inet_select_addr(const struct net_device *dev, __be32 dst, int scope)
 885 {
 886         __be32 addr = 0;
 887         struct in_device *in_dev;
 888
 889         rcu_read_lock();
 890         in_dev = __in_dev_get_rcu(dev);
 891         if (!in_dev)
 892                 goto no_in_dev;
 893
 894         for_primary_ifa(in_dev) {
 895                 if (ifa->ifa_scope > scope)
 896                         continue;
 897                 if (!dst || inet_ifa_match(dst, ifa)) {
 898                         addr = ifa->ifa_local;
 899                         break;
 900                 }
 901                 if (!addr)
 902                         addr = ifa->ifa_local;
 903         } endfor_ifa(in_dev);
 904 no_in_dev:
 905         rcu_read_unlock();
 906
 907         if (addr)
 908                 goto out;
 909
 910         /* Not loopback addresses on loopback should be preferred
 911            in this case. It is importnat that lo is the first interface
 912            in dev_base list.
 913          */
 914         read_lock(&dev_base_lock);
 915         rcu_read_lock();
 916         for_each_netdev(&init_net, dev) {
 917                 if ((in_dev = __in_dev_get_rcu(dev)) == NULL)
 918                         continue;
 919
 920                 for_primary_ifa(in_dev) {
 921                         if (ifa->ifa_scope != RT_SCOPE_LINK &&
 922                             ifa->ifa_scope <= scope) {
 923                                 addr = ifa->ifa_local;
 924                                 goto out_unlock_both;
 925                         }
 926                 } endfor_ifa(in_dev);
 927         }
 928 out_unlock_both:
 929         read_unlock(&dev_base_lock);
 930         rcu_read_unlock();
 931 out:
 932         return addr;
 933 }
 934
 935 static __be32 confirm_addr_indev(struct in_device *in_dev, __be32 dst,
 936                               __be32 local, int scope)
 937 {
 938         int same = 0;
 939         __be32 addr = 0;
 940
 941         for_ifa(in_dev) {
 942                 if (!addr &&
 943                     (local == ifa->ifa_local || !local) &&
 944                     ifa->ifa_scope <= scope) {
 945                         addr = ifa->ifa_local;
 946                         if (same)
 947                                 break;
 948                 }
 949                 if (!same) {
 950                         same = (!local || inet_ifa_match(local, ifa)) &&
 951                                 (!dst || inet_ifa_match(dst, ifa));
 952                         if (same && addr) {
 953                                 if (local || !dst)
 954                                         break;
 955                                 /* Is the selected addr into dst subnet? */
 956                                 if (inet_ifa_match(addr, ifa))
 957                                         break;
 958                                 /* No, then can we use new local src? */
 959                                 if (ifa->ifa_scope <= scope) {
 960                                         addr = ifa->ifa_local;
 961                                         break;
 962                                 }
 963                                 /* search for large dst subnet for addr */
 964                                 same = 0;
 965                         }
 966                 }
 967         } endfor_ifa(in_dev);
 968
 969         return same? addr : 0;
 970 }
 971
 972 /*
 973  * Confirm that local IP address exists using wildcards:
 974  * - dev: only on this interface, 0=any interface
 975  * - dst: only in the same subnet as dst, 0=any dst
 976  * - local: address, 0=autoselect the local address
 977  * - scope: maximum allowed scope value for the local address
 978  */
 979 __be32 inet_confirm_addr(const struct net_device *dev, __be32 dst, __be32 local, int scope)
 980 {
 981         __be32 addr = 0;
 982         struct in_device *in_dev;
 983
 984         if (dev) {
 985                 rcu_read_lock();
 986                 if ((in_dev = __in_dev_get_rcu(dev)))
 987                         addr = confirm_addr_indev(in_dev, dst, local, scope);
 988                 rcu_read_unlock();
 989
 990                 return addr;
 991         }
 992
 993         read_lock(&dev_base_lock);
 994         rcu_read_lock();
 995         for_each_netdev(&init_net, dev) {
 996                 if ((in_dev = __in_dev_get_rcu(dev))) {
 997                         addr = confirm_addr_indev(in_dev, dst, local, scope);
 998                         if (addr)
 999                                 break;
1000                 }
1001         }
1002         rcu_read_unlock();
1003         read_unlock(&dev_base_lock);
1004
1005         return addr;
1006 }
1007
1008 /*
1009  *      Device notifier
1010  */
1011
1012 int register_inetaddr_notifier(struct notifier_block *nb)
1013 {
1014         return blocking_notifier_chain_register(&inetaddr_chain, nb);
1015 }
1016
1017 int unregister_inetaddr_notifier(struct notifier_block *nb)
1018 {
1019         return blocking_notifier_chain_unregister(&inetaddr_chain, nb);
1020 }
1021
1022 /* Rename ifa_labels for a device name change. Make some effort to preserve existing
1023  * alias numbering and to create unique labels if possible.
1024 */
1025 static void inetdev_changename(struct net_device *dev, struct in_device *in_dev)
1026 {
1027         struct in_ifaddr *ifa;
1028         int named = 0;
1029
1030         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
1031                 char old[IFNAMSIZ], *dot;
1032
1033                 memcpy(old, ifa->ifa_label, IFNAMSIZ);
1034                 memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
1035                 if (named++ == 0)
1036                         continue;
1037                 dot = strchr(old, ':');
1038                 if (dot == NULL) {
1039                         sprintf(old, ":%d", named);
1040                         dot = old;
1041                 }
1042                 if (strlen(dot) + strlen(dev->name) < IFNAMSIZ) {
1043                         strcat(ifa->ifa_label, dot);
1044                 } else {
1045                         strcpy(ifa->ifa_label + (IFNAMSIZ - strlen(dot) - 1), dot);
1046                 }
1047         }
1048 }
1049
1050 /* Called only under RTNL semaphore */
1051
1052 static int inetdev_event(struct notifier_block *this, unsigned long event,
1053                          void *ptr)
1054 {
1055         struct net_device *dev = ptr;
1056         struct in_device *in_dev = __in_dev_get_rtnl(dev);
1057
1058         if (dev->nd_net != &init_net)
1059                 return NOTIFY_DONE;
1060
1061         ASSERT_RTNL();
1062
1063         if (!in_dev) {
1064                 if (event == NETDEV_REGISTER) {
1065                         in_dev = inetdev_init(dev);
1066                         if (!in_dev)
1067                                 return notifier_from_errno(-ENOMEM);
1068                         if (dev->flags & IFF_LOOPBACK) {
1069                                 IN_DEV_CONF_SET(in_dev, NOXFRM, 1);
1070                                 IN_DEV_CONF_SET(in_dev, NOPOLICY, 1);
1071                         }
1072                 }
1073                 goto out;
1074         }
1075
1076         switch (event) {
1077         case NETDEV_REGISTER:
1078                 printk(KERN_DEBUG "inetdev_event: bug\n");
1079                 dev->ip_ptr = NULL;
1080                 break;
1081         case NETDEV_UP:
1082                 if (dev->mtu < 68)
1083                         break;
1084                 if (dev->flags & IFF_LOOPBACK) {
1085                         struct in_ifaddr *ifa;
1086                         if ((ifa = inet_alloc_ifa()) != NULL) {
1087                                 ifa->ifa_local =
1088                                   ifa->ifa_address = htonl(INADDR_LOOPBACK);
1089                                 ifa->ifa_prefixlen = 8;
1090                                 ifa->ifa_mask = inet_make_mask(8);
1091                                 in_dev_hold(in_dev);
1092                                 ifa->ifa_dev = in_dev;
1093                                 ifa->ifa_scope = RT_SCOPE_HOST;
1094                                 memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
1095                                 inet_insert_ifa(ifa);
1096                         }
1097                 }
1098                 ip_mc_up(in_dev);
1099                 break;
1100         case NETDEV_DOWN:
1101                 ip_mc_down(in_dev);
1102                 break;
1103         case NETDEV_CHANGEMTU:
1104                 if (dev->mtu >= 68)
1105                         break;
1106                 /* MTU falled under 68, disable IP */
1107         case NETDEV_UNREGISTER:
1108                 inetdev_destroy(in_dev);
1109                 break;
1110         case NETDEV_CHANGENAME:
1111                 /* Do not notify about label change, this event is
1112                  * not interesting to applications using netlink.
1113                  */
1114                 inetdev_changename(dev, in_dev);
1115
1116 #ifdef CONFIG_SYSCTL
1117                 devinet_sysctl_unregister(&in_dev->cnf);
1118                 neigh_sysctl_unregister(in_dev->arp_parms);
1119                 neigh_sysctl_register(dev, in_dev->arp_parms, NET_IPV4,
1120                                       NET_IPV4_NEIGH, "ipv4", NULL, NULL);
1121                 devinet_sysctl_register(in_dev);
1122 #endif
1123                 break;
1124         }
1125 out:
1126         return NOTIFY_DONE;
1127 }
1128
1129 static struct notifier_block ip_netdev_notifier = {
1130         .notifier_call =inetdev_event,
1131 };
1132
1133 static inline size_t inet_nlmsg_size(void)
1134 {
1135         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
1136                + nla_total_size(4) /* IFA_ADDRESS */
1137                + nla_total_size(4) /* IFA_LOCAL */
1138                + nla_total_size(4) /* IFA_BROADCAST */
1139                + nla_total_size(4) /* IFA_ANYCAST */
1140                + nla_total_size(IFNAMSIZ); /* IFA_LABEL */
1141 }
1142
1143 static int inet_fill_ifaddr(struct sk_buff *skb, struct in_ifaddr *ifa,
1144                             u32 pid, u32 seq, int event, unsigned int flags)
1145 {
1146         struct ifaddrmsg *ifm;
1147         struct nlmsghdr  *nlh;
1148
1149         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*ifm), flags);
1150         if (nlh == NULL)
1151                 return -EMSGSIZE;
1152
1153         ifm = nlmsg_data(nlh);
1154         ifm->ifa_family = AF_INET;
1155         ifm->ifa_prefixlen = ifa->ifa_prefixlen;
1156         ifm->ifa_flags = ifa->ifa_flags|IFA_F_PERMANENT;
1157         ifm->ifa_scope = ifa->ifa_scope;
1158         ifm->ifa_index = ifa->ifa_dev->dev->ifindex;
1159
1160         if (ifa->ifa_address)
1161                 NLA_PUT_BE32(skb, IFA_ADDRESS, ifa->ifa_address);
1162
1163         if (ifa->ifa_local)
1164                 NLA_PUT_BE32(skb, IFA_LOCAL, ifa->ifa_local);
1165
1166         if (ifa->ifa_broadcast)
1167                 NLA_PUT_BE32(skb, IFA_BROADCAST, ifa->ifa_broadcast);
1168
1169         if (ifa->ifa_anycast)
1170                 NLA_PUT_BE32(skb, IFA_ANYCAST, ifa->ifa_anycast);
1171
1172         if (ifa->ifa_label[0])
1173                 NLA_PUT_STRING(skb, IFA_LABEL, ifa->ifa_label);
1174
1175         return nlmsg_end(skb, nlh);
1176
1177 nla_put_failure:
1178         nlmsg_cancel(skb, nlh);
1179         return -EMSGSIZE;
1180 }
1181
1182 static int inet_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
1183 {
1184         struct net *net = skb->sk->sk_net;
1185         int idx, ip_idx;
1186         struct net_device *dev;
1187         struct in_device *in_dev;
1188         struct in_ifaddr *ifa;
1189         int s_ip_idx, s_idx = cb->args[0];
1190
1191         if (net != &init_net)
1192                 return 0;
1193
1194         s_ip_idx = ip_idx = cb->args[1];
1195         idx = 0;
1196         for_each_netdev(&init_net, dev) {
1197                 if (idx < s_idx)
1198                         goto cont;
1199                 if (idx > s_idx)
1200                         s_ip_idx = 0;
1201                 if ((in_dev = __in_dev_get_rtnl(dev)) == NULL)
1202                         goto cont;
1203
1204                 for (ifa = in_dev->ifa_list, ip_idx = 0; ifa;
1205                      ifa = ifa->ifa_next, ip_idx++) {
1206                         if (ip_idx < s_ip_idx)
1207                                 continue;
1208                         if (inet_fill_ifaddr(skb, ifa, NETLINK_CB(cb->skb).pid,
1209                                              cb->nlh->nlmsg_seq,
1210                                              RTM_NEWADDR, NLM_F_MULTI) <= 0)
1211                                 goto done;
1212                 }
1213 cont:
1214                 idx++;
1215         }
1216
1217 done:
1218         cb->args[0] = idx;
1219         cb->args[1] = ip_idx;
1220
1221         return skb->len;
1222 }
1223
1224 static void rtmsg_ifa(int event, struct in_ifaddr* ifa, struct nlmsghdr *nlh,
1225                       u32 pid)
1226 {
1227         struct sk_buff *skb;
1228         u32 seq = nlh ? nlh->nlmsg_seq : 0;
1229         int err = -ENOBUFS;
1230
1231         skb = nlmsg_new(inet_nlmsg_size(), GFP_KERNEL);
1232         if (skb == NULL)
1233                 goto errout;
1234
1235         err = inet_fill_ifaddr(skb, ifa, pid, seq, event, 0);
1236         if (err < 0) {
1237                 /* -EMSGSIZE implies BUG in inet_nlmsg_size() */
1238                 WARN_ON(err == -EMSGSIZE);
1239                 kfree_skb(skb);
1240                 goto errout;
1241         }
1242         err = rtnl_notify(skb, &init_net, pid, RTNLGRP_IPV4_IFADDR, nlh, GFP_KERNEL);
1243 errout:
1244         if (err < 0)
1245                 rtnl_set_sk_err(&init_net, RTNLGRP_IPV4_IFADDR, err);
1246 }
1247
1248 #ifdef CONFIG_SYSCTL
1249
1250 static void devinet_copy_dflt_conf(int i)
1251 {
1252         struct net_device *dev;
1253
1254         read_lock(&dev_base_lock);
1255         for_each_netdev(&init_net, dev) {
1256                 struct in_device *in_dev;
1257                 rcu_read_lock();
1258                 in_dev = __in_dev_get_rcu(dev);
1259                 if (in_dev && !test_bit(i, in_dev->cnf.state))
1260                         in_dev->cnf.data[i] = ipv4_devconf_dflt.data[i];
1261                 rcu_read_unlock();
1262         }
1263         read_unlock(&dev_base_lock);
1264 }
1265
1266 static int devinet_conf_proc(ctl_table *ctl, int write,
1267                              struct file* filp, void __user *buffer,
1268                              size_t *lenp, loff_t *ppos)
1269 {
1270         int ret = proc_dointvec(ctl, write, filp, buffer, lenp, ppos);
1271
1272         if (write) {
1273                 struct ipv4_devconf *cnf = ctl->extra1;
1274                 int i = (int *)ctl->data - cnf->data;
1275
1276                 set_bit(i, cnf->state);
1277
1278                 if (cnf == &ipv4_devconf_dflt)
1279                         devinet_copy_dflt_conf(i);
1280         }
1281
1282         return ret;
1283 }
1284
1285 static int devinet_conf_sysctl(ctl_table *table, int __user *name, int nlen,
1286                                void __user *oldval, size_t __user *oldlenp,
1287                                void __user *newval, size_t newlen)
1288 {
1289         struct ipv4_devconf *cnf;
1290         int *valp = table->data;
1291         int new;
1292         int i;
1293
1294         if (!newval || !newlen)
1295                 return 0;
1296
1297         if (newlen != sizeof(int))
1298                 return -EINVAL;
1299
1300         if (get_user(new, (int __user *)newval))
1301                 return -EFAULT;
1302
1303         if (new == *valp)
1304                 return 0;
1305
1306         if (oldval && oldlenp) {
1307                 size_t len;
1308
1309                 if (get_user(len, oldlenp))
1310                         return -EFAULT;
1311
1312                 if (len) {
1313                         if (len > table->maxlen)
1314                                 len = table->maxlen;
1315                         if (copy_to_user(oldval, valp, len))
1316                                 return -EFAULT;
1317                         if (put_user(len, oldlenp))
1318                                 return -EFAULT;
1319                 }
1320         }
1321
1322         *valp = new;
1323
1324         cnf = table->extra1;
1325         i = (int *)table->data - cnf->data;
1326
1327         set_bit(i, cnf->state);
1328
1329         if (cnf == &ipv4_devconf_dflt)
1330                 devinet_copy_dflt_conf(i);
1331
1332         return 1;
1333 }
1334
1335 void inet_forward_change(void)
1336 {
1337         struct net_device *dev;
1338         int on = IPV4_DEVCONF_ALL(FORWARDING);
1339
1340         IPV4_DEVCONF_ALL(ACCEPT_REDIRECTS) = !on;
1341         IPV4_DEVCONF_DFLT(FORWARDING) = on;
1342
1343         read_lock(&dev_base_lock);
1344         for_each_netdev(&init_net, dev) {
1345                 struct in_device *in_dev;
1346                 rcu_read_lock();
1347                 in_dev = __in_dev_get_rcu(dev);
1348                 if (in_dev)
1349                         IN_DEV_CONF_SET(in_dev, FORWARDING, on);
1350                 rcu_read_unlock();
1351         }
1352         read_unlock(&dev_base_lock);
1353
1354         rt_cache_flush(0);
1355 }
1356
1357 static int devinet_sysctl_forward(ctl_table *ctl, int write,
1358                                   struct file* filp, void __user *buffer,
1359                                   size_t *lenp, loff_t *ppos)
1360 {
1361         int *valp = ctl->data;
1362         int val = *valp;
1363         int ret = proc_dointvec(ctl, write, filp, buffer, lenp, ppos);
1364
1365         if (write && *valp != val) {
1366                 if (valp == &IPV4_DEVCONF_ALL(FORWARDING))
1367                         inet_forward_change();
1368                 else if (valp != &IPV4_DEVCONF_DFLT(FORWARDING))
1369                         rt_cache_flush(0);
1370         }
1371
1372         return ret;
1373 }
1374
1375 int ipv4_doint_and_flush(ctl_table *ctl, int write,
1376                          struct file* filp, void __user *buffer,
1377                          size_t *lenp, loff_t *ppos)
1378 {
1379         int *valp = ctl->data;
1380         int val = *valp;
1381         int ret = proc_dointvec(ctl, write, filp, buffer, lenp, ppos);
1382
1383         if (write && *valp != val)
1384                 rt_cache_flush(0);
1385
1386         return ret;
1387 }
1388
1389 int ipv4_doint_and_flush_strategy(ctl_table *table, int __user *name, int nlen,
1390                                   void __user *oldval, size_t __user *oldlenp,
1391                                   void __user *newval, size_t newlen)
1392 {
1393         int ret = devinet_conf_sysctl(table, name, nlen, oldval, oldlenp,
1394                                       newval, newlen);
1395
1396         if (ret == 1)
1397                 rt_cache_flush(0);
1398
1399         return ret;
1400 }
1401
1402
1403 #define DEVINET_SYSCTL_ENTRY(attr, name, mval, proc, sysctl) \
1404         { \
1405                 .ctl_name       = NET_IPV4_CONF_ ## attr, \
1406                 .procname       = name, \
1407                 .data           = ipv4_devconf.data + \
1408                                   NET_IPV4_CONF_ ## attr - 1, \
1409                 .maxlen         = sizeof(int), \
1410                 .mode           = mval, \
1411                 .proc_handler   = proc, \
1412                 .strategy       = sysctl, \
1413                 .extra1         = &ipv4_devconf, \
1414         }
1415
1416 #define DEVINET_SYSCTL_RW_ENTRY(attr, name) \
1417         DEVINET_SYSCTL_ENTRY(attr, name, 0644, devinet_conf_proc, \
1418                              devinet_conf_sysctl)
1419
1420 #define DEVINET_SYSCTL_RO_ENTRY(attr, name) \
1421         DEVINET_SYSCTL_ENTRY(attr, name, 0444, devinet_conf_proc, \
1422                              devinet_conf_sysctl)
1423
1424 #define DEVINET_SYSCTL_COMPLEX_ENTRY(attr, name, proc, sysctl) \
1425         DEVINET_SYSCTL_ENTRY(attr, name, 0644, proc, sysctl)
1426
1427 #define DEVINET_SYSCTL_FLUSHING_ENTRY(attr, name) \
1428         DEVINET_SYSCTL_COMPLEX_ENTRY(attr, name, ipv4_doint_and_flush, \
1429                                      ipv4_doint_and_flush_strategy)
1430
1431 static struct devinet_sysctl_table {
1432         struct ctl_table_header *sysctl_header;
1433         struct ctl_table devinet_vars[__NET_IPV4_CONF_MAX];
1434         char *dev_name;
1435 } devinet_sysctl = {
1436         .devinet_vars = {
1437                 DEVINET_SYSCTL_COMPLEX_ENTRY(FORWARDING, "forwarding",
1438                                              devinet_sysctl_forward,
1439                                              devinet_conf_sysctl),
1440                 DEVINET_SYSCTL_RO_ENTRY(MC_FORWARDING, "mc_forwarding"),
1441
1442                 DEVINET_SYSCTL_RW_ENTRY(ACCEPT_REDIRECTS, "accept_redirects"),
1443                 DEVINET_SYSCTL_RW_ENTRY(SECURE_REDIRECTS, "secure_redirects"),
1444                 DEVINET_SYSCTL_RW_ENTRY(SHARED_MEDIA, "shared_media"),
1445                 DEVINET_SYSCTL_RW_ENTRY(RP_FILTER, "rp_filter"),
1446                 DEVINET_SYSCTL_RW_ENTRY(SEND_REDIRECTS, "send_redirects"),
1447                 DEVINET_SYSCTL_RW_ENTRY(ACCEPT_SOURCE_ROUTE,
1448                                         "accept_source_route"),
1449                 DEVINET_SYSCTL_RW_ENTRY(PROXY_ARP, "proxy_arp"),
1450                 DEVINET_SYSCTL_RW_ENTRY(MEDIUM_ID, "medium_id"),
1451                 DEVINET_SYSCTL_RW_ENTRY(BOOTP_RELAY, "bootp_relay"),
1452                 DEVINET_SYSCTL_RW_ENTRY(LOG_MARTIANS, "log_martians"),
1453                 DEVINET_SYSCTL_RW_ENTRY(TAG, "tag"),
1454                 DEVINET_SYSCTL_RW_ENTRY(ARPFILTER, "arp_filter"),
1455                 DEVINET_SYSCTL_RW_ENTRY(ARP_ANNOUNCE, "arp_announce"),
1456                 DEVINET_SYSCTL_RW_ENTRY(ARP_IGNORE, "arp_ignore"),
1457                 DEVINET_SYSCTL_RW_ENTRY(ARP_ACCEPT, "arp_accept"),
1458
1459                 DEVINET_SYSCTL_FLUSHING_ENTRY(NOXFRM, "disable_xfrm"),
1460                 DEVINET_SYSCTL_FLUSHING_ENTRY(NOPOLICY, "disable_policy"),
1461                 DEVINET_SYSCTL_FLUSHING_ENTRY(FORCE_IGMP_VERSION,
1462                                               "force_igmp_version"),
1463                 DEVINET_SYSCTL_FLUSHING_ENTRY(PROMOTE_SECONDARIES,
1464                                               "promote_secondaries"),
1465         },
1466 };
1467
1468 static void __devinet_sysctl_register(char *dev_name, int ctl_name,
1469                 struct ipv4_devconf *p)
1470 {
1471         int i;
1472         struct devinet_sysctl_table *t;
1473
1474 #define DEVINET_CTL_PATH_DEV    3
1475
1476         struct ctl_path devinet_ctl_path[] = {
1477                 { .procname = "net", .ctl_name = CTL_NET, },
1478                 { .procname = "ipv4", .ctl_name = NET_IPV4, },
1479                 { .procname = "conf", .ctl_name = NET_IPV4_CONF, },
1480                 { /* to be set */ },
1481                 { },
1482         };
1483
1484         t = kmemdup(&devinet_sysctl, sizeof(*t), GFP_KERNEL);
1485         if (!t)
1486                 goto out;
1487
1488         for (i = 0; i < ARRAY_SIZE(t->devinet_vars) - 1; i++) {
1489                 t->devinet_vars[i].data += (char *)p - (char *)&ipv4_devconf;
1490                 t->devinet_vars[i].extra1 = p;
1491         }
1492
1493         /*
1494          * Make a copy of dev_name, because '.procname' is regarded as const
1495          * by sysctl and we wouldn't want anyone to change it under our feet
1496          * (see SIOCSIFNAME).
1497          */
1498         t->dev_name = kstrdup(dev_name, GFP_KERNEL);
1499         if (!t->dev_name)
1500                 goto free;
1501
1502         devinet_ctl_path[DEVINET_CTL_PATH_DEV].procname = t->dev_name;
1503         devinet_ctl_path[DEVINET_CTL_PATH_DEV].ctl_name = ctl_name;
1504
1505         t->sysctl_header = register_sysctl_paths(devinet_ctl_path,
1506                         t->devinet_vars);
1507         if (!t->sysctl_header)
1508                 goto free_procname;
1509
1510         p->sysctl = t;
1511         return;
1512
1513 free_procname:
1514         kfree(t->dev_name);
1515 free:
1516         kfree(t);
1517 out:
1518         return;
1519 }
1520
1521 static void devinet_sysctl_register(struct in_device *idev)
1522 {
1523         return __devinet_sysctl_register(idev->dev->name, idev->dev->ifindex,
1524                         &idev->cnf);
1525 }
1526
1527 static void devinet_sysctl_unregister(struct ipv4_devconf *p)
1528 {
1529         if (p->sysctl) {
1530                 struct devinet_sysctl_table *t = p->sysctl;
1531                 p->sysctl = NULL;
1532                 unregister_sysctl_table(t->sysctl_header);
1533                 kfree(t->dev_name);
1534                 kfree(t);
1535         }
1536 }
1537 #endif
1538
1539 void __init devinet_init(void)
1540 {
1541         register_gifconf(PF_INET, inet_gifconf);
1542         register_netdevice_notifier(&ip_netdev_notifier);
1543
1544         rtnl_register(PF_INET, RTM_NEWADDR, inet_rtm_newaddr, NULL);
1545         rtnl_register(PF_INET, RTM_DELADDR, inet_rtm_deladdr, NULL);
1546         rtnl_register(PF_INET, RTM_GETADDR, NULL, inet_dump_ifaddr);
1547 #ifdef CONFIG_SYSCTL
1548         __devinet_sysctl_register("all", NET_PROTO_CONF_ALL,
1549                         &ipv4_devconf);
1550         __devinet_sysctl_register("default", NET_PROTO_CONF_DEFAULT,
1551                         &ipv4_devconf_dflt);
1552 #endif
1553 }
1554
1555 EXPORT_SYMBOL(in_dev_finish_destroy);
1556 EXPORT_SYMBOL(inet_select_addr);
1557 EXPORT_SYMBOL(inetdev_by_index);
1558 EXPORT_SYMBOL(register_inetaddr_notifier);
1559 EXPORT_SYMBOL(unregister_inetaddr_notifier);