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 #include <linux/config.h>
  31
  32 #include <asm/uaccess.h>
  33 #include <asm/system.h>
  34 #include <linux/bitops.h>
  35 #include <linux/capability.h>
  36 #include <linux/module.h>
  37 #include <linux/types.h>
  38 #include <linux/kernel.h>
  39 #include <linux/sched.h>
  40 #include <linux/string.h>
  41 #include <linux/mm.h>
  42 #include <linux/socket.h>
  43 #include <linux/sockios.h>
  44 #include <linux/in.h>
  45 #include <linux/errno.h>
  46 #include <linux/interrupt.h>
  47 #include <linux/if_ether.h>
  48 #include <linux/inet.h>
  49 #include <linux/netdevice.h>
  50 #include <linux/etherdevice.h>
  51 #include <linux/skbuff.h>
  52 #include <linux/rtnetlink.h>
  53 #include <linux/init.h>
  54 #include <linux/notifier.h>
  55 #include <linux/inetdevice.h>
  56 #include <linux/igmp.h>
  57 #ifdef CONFIG_SYSCTL
  58 #include <linux/sysctl.h>
  59 #endif
  60 #include <linux/kmod.h>
  61
  62 #include <net/arp.h>
  63 #include <net/ip.h>
  64 #include <net/route.h>
  65 #include <net/ip_fib.h>
  66
  67 struct ipv4_devconf ipv4_devconf = {
  68         .accept_redirects = 1,
  69         .send_redirects =  1,
  70         .secure_redirects = 1,
  71         .shared_media =   1,
  72 };
  73
  74 static struct ipv4_devconf ipv4_devconf_dflt = {
  75         .accept_redirects =  1,
  76         .send_redirects =    1,
  77         .secure_redirects =  1,
  78         .shared_media =      1,
  79         .accept_source_route = 1,
  80 };
  81
  82 static void rtmsg_ifa(int event, struct in_ifaddr *);
  83
  84 static BLOCKING_NOTIFIER_HEAD(inetaddr_chain);
  85 static void inet_del_ifa(struct in_device *in_dev, struct in_ifaddr **ifap,
  86                          int destroy);
  87 #ifdef CONFIG_SYSCTL
  88 static void devinet_sysctl_register(struct in_device *in_dev,
  89                                     struct ipv4_devconf *p);
  90 static void devinet_sysctl_unregister(struct ipv4_devconf *p);
  91 #endif
  92
  93 /* Locks all the inet devices. */
  94
  95 static struct in_ifaddr *inet_alloc_ifa(void)
  96 {
  97         struct in_ifaddr *ifa = kmalloc(sizeof(*ifa), GFP_KERNEL);
  98
  99         if (ifa) {
 100                 memset(ifa, 0, sizeof(*ifa));
 101                 INIT_RCU_HEAD(&ifa->rcu_head);
 102         }
 103
 104         return ifa;
 105 }
 106
 107 static void inet_rcu_free_ifa(struct rcu_head *head)
 108 {
 109         struct in_ifaddr *ifa = container_of(head, struct in_ifaddr, rcu_head);
 110         if (ifa->ifa_dev)
 111                 in_dev_put(ifa->ifa_dev);
 112         kfree(ifa);
 113 }
 114
 115 static inline void inet_free_ifa(struct in_ifaddr *ifa)
 116 {
 117         call_rcu(&ifa->rcu_head, inet_rcu_free_ifa);
 118 }
 119
 120 void in_dev_finish_destroy(struct in_device *idev)
 121 {
 122         struct net_device *dev = idev->dev;
 123
 124         BUG_TRAP(!idev->ifa_list);
 125         BUG_TRAP(!idev->mc_list);
 126 #ifdef NET_REFCNT_DEBUG
 127         printk(KERN_DEBUG "in_dev_finish_destroy: %p=%s\n",
 128                idev, dev ? dev->name : "NIL");
 129 #endif
 130         dev_put(dev);
 131         if (!idev->dead)
 132                 printk("Freeing alive in_device %p\n", idev);
 133         else {
 134                 kfree(idev);
 135         }
 136 }
 137
 138 struct in_device *inetdev_init(struct net_device *dev)
 139 {
 140         struct in_device *in_dev;
 141
 142         ASSERT_RTNL();
 143
 144         in_dev = kmalloc(sizeof(*in_dev), GFP_KERNEL);
 145         if (!in_dev)
 146                 goto out;
 147         memset(in_dev, 0, sizeof(*in_dev));
 148         INIT_RCU_HEAD(&in_dev->rcu_head);
 149         memcpy(&in_dev->cnf, &ipv4_devconf_dflt, sizeof(in_dev->cnf));
 150         in_dev->cnf.sysctl = NULL;
 151         in_dev->dev = dev;
 152         if ((in_dev->arp_parms = neigh_parms_alloc(dev, &arp_tbl)) == NULL)
 153                 goto out_kfree;
 154         /* Reference in_dev->dev */
 155         dev_hold(dev);
 156 #ifdef CONFIG_SYSCTL
 157         neigh_sysctl_register(dev, in_dev->arp_parms, NET_IPV4,
 158                               NET_IPV4_NEIGH, "ipv4", NULL, NULL);
 159 #endif
 160
 161         /* Account for reference dev->ip_ptr */
 162         in_dev_hold(in_dev);
 163         rcu_assign_pointer(dev->ip_ptr, in_dev);
 164
 165 #ifdef CONFIG_SYSCTL
 166         devinet_sysctl_register(in_dev, &in_dev->cnf);
 167 #endif
 168         ip_mc_init_dev(in_dev);
 169         if (dev->flags & IFF_UP)
 170                 ip_mc_up(in_dev);
 171 out:
 172         return in_dev;
 173 out_kfree:
 174         kfree(in_dev);
 175         in_dev = NULL;
 176         goto out;
 177 }
 178
 179 static void in_dev_rcu_put(struct rcu_head *head)
 180 {
 181         struct in_device *idev = container_of(head, struct in_device, rcu_head);
 182         in_dev_put(idev);
 183 }
 184
 185 static void inetdev_destroy(struct in_device *in_dev)
 186 {
 187         struct in_ifaddr *ifa;
 188         struct net_device *dev;
 189
 190         ASSERT_RTNL();
 191
 192         dev = in_dev->dev;
 193         if (dev == &loopback_dev)
 194                 return;
 195
 196         in_dev->dead = 1;
 197
 198         ip_mc_destroy_dev(in_dev);
 199
 200         while ((ifa = in_dev->ifa_list) != NULL) {
 201                 inet_del_ifa(in_dev, &in_dev->ifa_list, 0);
 202                 inet_free_ifa(ifa);
 203         }
 204
 205 #ifdef CONFIG_SYSCTL
 206         devinet_sysctl_unregister(&in_dev->cnf);
 207 #endif
 208
 209         dev->ip_ptr = NULL;
 210
 211 #ifdef CONFIG_SYSCTL
 212         neigh_sysctl_unregister(in_dev->arp_parms);
 213 #endif
 214         neigh_parms_release(&arp_tbl, in_dev->arp_parms);
 215         arp_ifdown(dev);
 216
 217         call_rcu(&in_dev->rcu_head, in_dev_rcu_put);
 218 }
 219
 220 int inet_addr_onlink(struct in_device *in_dev, u32 a, u32 b)
 221 {
 222         rcu_read_lock();
 223         for_primary_ifa(in_dev) {
 224                 if (inet_ifa_match(a, ifa)) {
 225                         if (!b || inet_ifa_match(b, ifa)) {
 226                                 rcu_read_unlock();
 227                                 return 1;
 228                         }
 229                 }
 230         } endfor_ifa(in_dev);
 231         rcu_read_unlock();
 232         return 0;
 233 }
 234
 235 static void inet_del_ifa(struct in_device *in_dev, struct in_ifaddr **ifap,
 236                          int destroy)
 237 {
 238         struct in_ifaddr *promote = NULL;
 239         struct in_ifaddr *ifa, *ifa1 = *ifap;
 240         struct in_ifaddr *last_prim = in_dev->ifa_list;
 241         struct in_ifaddr *prev_prom = NULL;
 242         int do_promote = IN_DEV_PROMOTE_SECONDARIES(in_dev);
 243
 244         ASSERT_RTNL();
 245
 246         /* 1. Deleting primary ifaddr forces deletion all secondaries
 247          * unless alias promotion is set
 248          **/
 249
 250         if (!(ifa1->ifa_flags & IFA_F_SECONDARY)) {
 251                 struct in_ifaddr **ifap1 = &ifa1->ifa_next;
 252
 253                 while ((ifa = *ifap1) != NULL) {
 254                         if (!(ifa->ifa_flags & IFA_F_SECONDARY) &&
 255                             ifa1->ifa_scope <= ifa->ifa_scope)
 256                                 last_prim = ifa;
 257
 258                         if (!(ifa->ifa_flags & IFA_F_SECONDARY) ||
 259                             ifa1->ifa_mask != ifa->ifa_mask ||
 260                             !inet_ifa_match(ifa1->ifa_address, ifa)) {
 261                                 ifap1 = &ifa->ifa_next;
 262                                 prev_prom = ifa;
 263                                 continue;
 264                         }
 265
 266                         if (!do_promote) {
 267                                 *ifap1 = ifa->ifa_next;
 268
 269                                 rtmsg_ifa(RTM_DELADDR, ifa);
 270                                 blocking_notifier_call_chain(&inetaddr_chain,
 271                                                 NETDEV_DOWN, ifa);
 272                                 inet_free_ifa(ifa);
 273                         } else {
 274                                 promote = ifa;
 275                                 break;
 276                         }
 277                 }
 278         }
 279
 280         /* 2. Unlink it */
 281
 282         *ifap = ifa1->ifa_next;
 283
 284         /* 3. Announce address deletion */
 285
 286         /* Send message first, then call notifier.
 287            At first sight, FIB update triggered by notifier
 288            will refer to already deleted ifaddr, that could confuse
 289            netlink listeners. It is not true: look, gated sees
 290            that route deleted and if it still thinks that ifaddr
 291            is valid, it will try to restore deleted routes... Grr.
 292            So that, this order is correct.
 293          */
 294         rtmsg_ifa(RTM_DELADDR, ifa1);
 295         blocking_notifier_call_chain(&inetaddr_chain, NETDEV_DOWN, ifa1);
 296
 297         if (promote) {
 298
 299                 if (prev_prom) {
 300                         prev_prom->ifa_next = promote->ifa_next;
 301                         promote->ifa_next = last_prim->ifa_next;
 302                         last_prim->ifa_next = promote;
 303                 }
 304
 305                 promote->ifa_flags &= ~IFA_F_SECONDARY;
 306                 rtmsg_ifa(RTM_NEWADDR, promote);
 307                 blocking_notifier_call_chain(&inetaddr_chain,
 308                                 NETDEV_UP, promote);
 309                 for (ifa = promote->ifa_next; ifa; ifa = ifa->ifa_next) {
 310                         if (ifa1->ifa_mask != ifa->ifa_mask ||
 311                             !inet_ifa_match(ifa1->ifa_address, ifa))
 312                                         continue;
 313                         fib_add_ifaddr(ifa);
 314                 }
 315
 316         }
 317         if (destroy) {
 318                 inet_free_ifa(ifa1);
 319
 320                 if (!in_dev->ifa_list)
 321                         inetdev_destroy(in_dev);
 322         }
 323 }
 324
 325 static int inet_insert_ifa(struct in_ifaddr *ifa)
 326 {
 327         struct in_device *in_dev = ifa->ifa_dev;
 328         struct in_ifaddr *ifa1, **ifap, **last_primary;
 329
 330         ASSERT_RTNL();
 331
 332         if (!ifa->ifa_local) {
 333                 inet_free_ifa(ifa);
 334                 return 0;
 335         }
 336
 337         ifa->ifa_flags &= ~IFA_F_SECONDARY;
 338         last_primary = &in_dev->ifa_list;
 339
 340         for (ifap = &in_dev->ifa_list; (ifa1 = *ifap) != NULL;
 341              ifap = &ifa1->ifa_next) {
 342                 if (!(ifa1->ifa_flags & IFA_F_SECONDARY) &&
 343                     ifa->ifa_scope <= ifa1->ifa_scope)
 344                         last_primary = &ifa1->ifa_next;
 345                 if (ifa1->ifa_mask == ifa->ifa_mask &&
 346                     inet_ifa_match(ifa1->ifa_address, ifa)) {
 347                         if (ifa1->ifa_local == ifa->ifa_local) {
 348                                 inet_free_ifa(ifa);
 349                                 return -EEXIST;
 350                         }
 351                         if (ifa1->ifa_scope != ifa->ifa_scope) {
 352                                 inet_free_ifa(ifa);
 353                                 return -EINVAL;
 354                         }
 355                         ifa->ifa_flags |= IFA_F_SECONDARY;
 356                 }
 357         }
 358
 359         if (!(ifa->ifa_flags & IFA_F_SECONDARY)) {
 360                 net_srandom(ifa->ifa_local);
 361                 ifap = last_primary;
 362         }
 363
 364         ifa->ifa_next = *ifap;
 365         *ifap = ifa;
 366
 367         /* Send message first, then call notifier.
 368            Notifier will trigger FIB update, so that
 369            listeners of netlink will know about new ifaddr */
 370         rtmsg_ifa(RTM_NEWADDR, ifa);
 371         blocking_notifier_call_chain(&inetaddr_chain, NETDEV_UP, ifa);
 372
 373         return 0;
 374 }
 375
 376 static int inet_set_ifa(struct net_device *dev, struct in_ifaddr *ifa)
 377 {
 378         struct in_device *in_dev = __in_dev_get_rtnl(dev);
 379
 380         ASSERT_RTNL();
 381
 382         if (!in_dev) {
 383                 in_dev = inetdev_init(dev);
 384                 if (!in_dev) {
 385                         inet_free_ifa(ifa);
 386                         return -ENOBUFS;
 387                 }
 388         }
 389         if (ifa->ifa_dev != in_dev) {
 390                 BUG_TRAP(!ifa->ifa_dev);
 391                 in_dev_hold(in_dev);
 392                 ifa->ifa_dev = in_dev;
 393         }
 394         if (LOOPBACK(ifa->ifa_local))
 395                 ifa->ifa_scope = RT_SCOPE_HOST;
 396         return inet_insert_ifa(ifa);
 397 }
 398
 399 struct in_device *inetdev_by_index(int ifindex)
 400 {
 401         struct net_device *dev;
 402         struct in_device *in_dev = NULL;
 403         read_lock(&dev_base_lock);
 404         dev = __dev_get_by_index(ifindex);
 405         if (dev)
 406                 in_dev = in_dev_get(dev);
 407         read_unlock(&dev_base_lock);
 408         return in_dev;
 409 }
 410
 411 /* Called only from RTNL semaphored context. No locks. */
 412
 413 struct in_ifaddr *inet_ifa_byprefix(struct in_device *in_dev, u32 prefix,
 414                                     u32 mask)
 415 {
 416         ASSERT_RTNL();
 417
 418         for_primary_ifa(in_dev) {
 419                 if (ifa->ifa_mask == mask && inet_ifa_match(prefix, ifa))
 420                         return ifa;
 421         } endfor_ifa(in_dev);
 422         return NULL;
 423 }
 424
 425 static int inet_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
 426 {
 427         struct rtattr **rta = arg;
 428         struct in_device *in_dev;
 429         struct ifaddrmsg *ifm = NLMSG_DATA(nlh);
 430         struct in_ifaddr *ifa, **ifap;
 431
 432         ASSERT_RTNL();
 433
 434         if ((in_dev = inetdev_by_index(ifm->ifa_index)) == NULL)
 435                 goto out;
 436         __in_dev_put(in_dev);
 437
 438         for (ifap = &in_dev->ifa_list; (ifa = *ifap) != NULL;
 439              ifap = &ifa->ifa_next) {
 440                 if ((rta[IFA_LOCAL - 1] &&
 441                      memcmp(RTA_DATA(rta[IFA_LOCAL - 1]),
 442                             &ifa->ifa_local, 4)) ||
 443                     (rta[IFA_LABEL - 1] &&
 444                      rtattr_strcmp(rta[IFA_LABEL - 1], ifa->ifa_label)) ||
 445                     (rta[IFA_ADDRESS - 1] &&
 446                      (ifm->ifa_prefixlen != ifa->ifa_prefixlen ||
 447                       !inet_ifa_match(*(u32*)RTA_DATA(rta[IFA_ADDRESS - 1]),
 448                                       ifa))))
 449                         continue;
 450                 inet_del_ifa(in_dev, ifap, 1);
 451                 return 0;
 452         }
 453 out:
 454         return -EADDRNOTAVAIL;
 455 }
 456
 457 static int inet_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
 458 {
 459         struct rtattr **rta = arg;
 460         struct net_device *dev;
 461         struct in_device *in_dev;
 462         struct ifaddrmsg *ifm = NLMSG_DATA(nlh);
 463         struct in_ifaddr *ifa;
 464         int rc = -EINVAL;
 465
 466         ASSERT_RTNL();
 467
 468         if (ifm->ifa_prefixlen > 32 || !rta[IFA_LOCAL - 1])
 469                 goto out;
 470
 471         rc = -ENODEV;
 472         if ((dev = __dev_get_by_index(ifm->ifa_index)) == NULL)
 473                 goto out;
 474
 475         rc = -ENOBUFS;
 476         if ((in_dev = __in_dev_get_rtnl(dev)) == NULL) {
 477                 in_dev = inetdev_init(dev);
 478                 if (!in_dev)
 479                         goto out;
 480         }
 481
 482         if ((ifa = inet_alloc_ifa()) == NULL)
 483                 goto out;
 484
 485         if (!rta[IFA_ADDRESS - 1])
 486                 rta[IFA_ADDRESS - 1] = rta[IFA_LOCAL - 1];
 487         memcpy(&ifa->ifa_local, RTA_DATA(rta[IFA_LOCAL - 1]), 4);
 488         memcpy(&ifa->ifa_address, RTA_DATA(rta[IFA_ADDRESS - 1]), 4);
 489         ifa->ifa_prefixlen = ifm->ifa_prefixlen;
 490         ifa->ifa_mask = inet_make_mask(ifm->ifa_prefixlen);
 491         if (rta[IFA_BROADCAST - 1])
 492                 memcpy(&ifa->ifa_broadcast,
 493                        RTA_DATA(rta[IFA_BROADCAST - 1]), 4);
 494         if (rta[IFA_ANYCAST - 1])
 495                 memcpy(&ifa->ifa_anycast, RTA_DATA(rta[IFA_ANYCAST - 1]), 4);
 496         ifa->ifa_flags = ifm->ifa_flags;
 497         ifa->ifa_scope = ifm->ifa_scope;
 498         in_dev_hold(in_dev);
 499         ifa->ifa_dev   = in_dev;
 500         if (rta[IFA_LABEL - 1])
 501                 rtattr_strlcpy(ifa->ifa_label, rta[IFA_LABEL - 1], IFNAMSIZ);
 502         else
 503                 memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
 504
 505         rc = inet_insert_ifa(ifa);
 506 out:
 507         return rc;
 508 }
 509
 510 /*
 511  *      Determine a default network mask, based on the IP address.
 512  */
 513
 514 static __inline__ int inet_abc_len(u32 addr)
 515 {
 516         int rc = -1;    /* Something else, probably a multicast. */
 517
 518         if (ZERONET(addr))
 519                 rc = 0;
 520         else {
 521                 addr = ntohl(addr);
 522
 523                 if (IN_CLASSA(addr))
 524                         rc = 8;
 525                 else if (IN_CLASSB(addr))
 526                         rc = 16;
 527                 else if (IN_CLASSC(addr))
 528                         rc = 24;
 529         }
 530
 531         return rc;
 532 }
 533
 534
 535 int devinet_ioctl(unsigned int cmd, void __user *arg)
 536 {
 537         struct ifreq ifr;
 538         struct sockaddr_in sin_orig;
 539         struct sockaddr_in *sin = (struct sockaddr_in *)&ifr.ifr_addr;
 540         struct in_device *in_dev;
 541         struct in_ifaddr **ifap = NULL;
 542         struct in_ifaddr *ifa = NULL;
 543         struct net_device *dev;
 544         char *colon;
 545         int ret = -EFAULT;
 546         int tryaddrmatch = 0;
 547
 548         /*
 549          *      Fetch the caller's info block into kernel space
 550          */
 551
 552         if (copy_from_user(&ifr, arg, sizeof(struct ifreq)))
 553                 goto out;
 554         ifr.ifr_name[IFNAMSIZ - 1] = 0;
 555
 556         /* save original address for comparison */
 557         memcpy(&sin_orig, sin, sizeof(*sin));
 558
 559         colon = strchr(ifr.ifr_name, ':');
 560         if (colon)
 561                 *colon = 0;
 562
 563 #ifdef CONFIG_KMOD
 564         dev_load(ifr.ifr_name);
 565 #endif
 566
 567         switch(cmd) {
 568         case SIOCGIFADDR:       /* Get interface address */
 569         case SIOCGIFBRDADDR:    /* Get the broadcast address */
 570         case SIOCGIFDSTADDR:    /* Get the destination address */
 571         case SIOCGIFNETMASK:    /* Get the netmask for the interface */
 572                 /* Note that these ioctls will not sleep,
 573                    so that we do not impose a lock.
 574                    One day we will be forced to put shlock here (I mean SMP)
 575                  */
 576                 tryaddrmatch = (sin_orig.sin_family == AF_INET);
 577                 memset(sin, 0, sizeof(*sin));
 578                 sin->sin_family = AF_INET;
 579                 break;
 580
 581         case SIOCSIFFLAGS:
 582                 ret = -EACCES;
 583                 if (!capable(CAP_NET_ADMIN))
 584                         goto out;
 585                 break;
 586         case SIOCSIFADDR:       /* Set interface address (and family) */
 587         case SIOCSIFBRDADDR:    /* Set the broadcast address */
 588         case SIOCSIFDSTADDR:    /* Set the destination address */
 589         case SIOCSIFNETMASK:    /* Set the netmask for the interface */
 590                 ret = -EACCES;
 591                 if (!capable(CAP_NET_ADMIN))
 592                         goto out;
 593                 ret = -EINVAL;
 594                 if (sin->sin_family != AF_INET)
 595                         goto out;
 596                 break;
 597         default:
 598                 ret = -EINVAL;
 599                 goto out;
 600         }
 601
 602         rtnl_lock();
 603
 604         ret = -ENODEV;
 605         if ((dev = __dev_get_by_name(ifr.ifr_name)) == NULL)
 606                 goto done;
 607
 608         if (colon)
 609                 *colon = ':';
 610
 611         if ((in_dev = __in_dev_get_rtnl(dev)) != NULL) {
 612                 if (tryaddrmatch) {
 613                         /* Matthias Andree */
 614                         /* compare label and address (4.4BSD style) */
 615                         /* note: we only do this for a limited set of ioctls
 616                            and only if the original address family was AF_INET.
 617                            This is checked above. */
 618                         for (ifap = &in_dev->ifa_list; (ifa = *ifap) != NULL;
 619                              ifap = &ifa->ifa_next) {
 620                                 if (!strcmp(ifr.ifr_name, ifa->ifa_label) &&
 621                                     sin_orig.sin_addr.s_addr ==
 622                                                         ifa->ifa_address) {
 623                                         break; /* found */
 624                                 }
 625                         }
 626                 }
 627                 /* we didn't get a match, maybe the application is
 628                    4.3BSD-style and passed in junk so we fall back to
 629                    comparing just the label */
 630                 if (!ifa) {
 631                         for (ifap = &in_dev->ifa_list; (ifa = *ifap) != NULL;
 632                              ifap = &ifa->ifa_next)
 633                                 if (!strcmp(ifr.ifr_name, ifa->ifa_label))
 634                                         break;
 635                 }
 636         }
 637
 638         ret = -EADDRNOTAVAIL;
 639         if (!ifa && cmd != SIOCSIFADDR && cmd != SIOCSIFFLAGS)
 640                 goto done;
 641
 642         switch(cmd) {
 643         case SIOCGIFADDR:       /* Get interface address */
 644                 sin->sin_addr.s_addr = ifa->ifa_local;
 645                 goto rarok;
 646
 647         case SIOCGIFBRDADDR:    /* Get the broadcast address */
 648                 sin->sin_addr.s_addr = ifa->ifa_broadcast;
 649                 goto rarok;
 650
 651         case SIOCGIFDSTADDR:    /* Get the destination address */
 652                 sin->sin_addr.s_addr = ifa->ifa_address;
 653                 goto rarok;
 654
 655         case SIOCGIFNETMASK:    /* Get the netmask for the interface */
 656                 sin->sin_addr.s_addr = ifa->ifa_mask;
 657                 goto rarok;
 658
 659         case SIOCSIFFLAGS:
 660                 if (colon) {
 661                         ret = -EADDRNOTAVAIL;
 662                         if (!ifa)
 663                                 break;
 664                         ret = 0;
 665                         if (!(ifr.ifr_flags & IFF_UP))
 666                                 inet_del_ifa(in_dev, ifap, 1);
 667                         break;
 668                 }
 669                 ret = dev_change_flags(dev, ifr.ifr_flags);
 670                 break;
 671
 672         case SIOCSIFADDR:       /* Set interface address (and family) */
 673                 ret = -EINVAL;
 674                 if (inet_abc_len(sin->sin_addr.s_addr) < 0)
 675                         break;
 676
 677                 if (!ifa) {
 678                         ret = -ENOBUFS;
 679                         if ((ifa = inet_alloc_ifa()) == NULL)
 680                                 break;
 681                         if (colon)
 682                                 memcpy(ifa->ifa_label, ifr.ifr_name, IFNAMSIZ);
 683                         else
 684                                 memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
 685                 } else {
 686                         ret = 0;
 687                         if (ifa->ifa_local == sin->sin_addr.s_addr)
 688                                 break;
 689                         inet_del_ifa(in_dev, ifap, 0);
 690                         ifa->ifa_broadcast = 0;
 691                         ifa->ifa_anycast = 0;
 692                 }
 693
 694                 ifa->ifa_address = ifa->ifa_local = sin->sin_addr.s_addr;
 695
 696                 if (!(dev->flags & IFF_POINTOPOINT)) {
 697                         ifa->ifa_prefixlen = inet_abc_len(ifa->ifa_address);
 698                         ifa->ifa_mask = inet_make_mask(ifa->ifa_prefixlen);
 699                         if ((dev->flags & IFF_BROADCAST) &&
 700                             ifa->ifa_prefixlen < 31)
 701                                 ifa->ifa_broadcast = ifa->ifa_address |
 702                                                      ~ifa->ifa_mask;
 703                 } else {
 704                         ifa->ifa_prefixlen = 32;
 705                         ifa->ifa_mask = inet_make_mask(32);
 706                 }
 707                 ret = inet_set_ifa(dev, ifa);
 708                 break;
 709
 710         case SIOCSIFBRDADDR:    /* Set the broadcast address */
 711                 ret = 0;
 712                 if (ifa->ifa_broadcast != sin->sin_addr.s_addr) {
 713                         inet_del_ifa(in_dev, ifap, 0);
 714                         ifa->ifa_broadcast = sin->sin_addr.s_addr;
 715                         inet_insert_ifa(ifa);
 716                 }
 717                 break;
 718
 719         case SIOCSIFDSTADDR:    /* Set the destination address */
 720                 ret = 0;
 721                 if (ifa->ifa_address == sin->sin_addr.s_addr)
 722                         break;
 723                 ret = -EINVAL;
 724                 if (inet_abc_len(sin->sin_addr.s_addr) < 0)
 725                         break;
 726                 ret = 0;
 727                 inet_del_ifa(in_dev, ifap, 0);
 728                 ifa->ifa_address = sin->sin_addr.s_addr;
 729                 inet_insert_ifa(ifa);
 730                 break;
 731
 732         case SIOCSIFNETMASK:    /* Set the netmask for the interface */
 733
 734                 /*
 735                  *      The mask we set must be legal.
 736                  */
 737                 ret = -EINVAL;
 738                 if (bad_mask(sin->sin_addr.s_addr, 0))
 739                         break;
 740                 ret = 0;
 741                 if (ifa->ifa_mask != sin->sin_addr.s_addr) {
 742                         u32 old_mask = ifa->ifa_mask;
 743                         inet_del_ifa(in_dev, ifap, 0);
 744                         ifa->ifa_mask = sin->sin_addr.s_addr;
 745                         ifa->ifa_prefixlen = inet_mask_len(ifa->ifa_mask);
 746
 747                         /* See if current broadcast address matches
 748                          * with current netmask, then recalculate
 749                          * the broadcast address. Otherwise it's a
 750                          * funny address, so don't touch it since
 751                          * the user seems to know what (s)he's doing...
 752                          */
 753                         if ((dev->flags & IFF_BROADCAST) &&
 754                             (ifa->ifa_prefixlen < 31) &&
 755                             (ifa->ifa_broadcast ==
 756                              (ifa->ifa_local|~old_mask))) {
 757                                 ifa->ifa_broadcast = (ifa->ifa_local |
 758                                                       ~sin->sin_addr.s_addr);
 759                         }
 760                         inet_insert_ifa(ifa);
 761                 }
 762                 break;
 763         }
 764 done:
 765         rtnl_unlock();
 766 out:
 767         return ret;
 768 rarok:
 769         rtnl_unlock();
 770         ret = copy_to_user(arg, &ifr, sizeof(struct ifreq)) ? -EFAULT : 0;
 771         goto out;
 772 }
 773
 774 static int inet_gifconf(struct net_device *dev, char __user *buf, int len)
 775 {
 776         struct in_device *in_dev = __in_dev_get_rtnl(dev);
 777         struct in_ifaddr *ifa;
 778         struct ifreq ifr;
 779         int done = 0;
 780
 781         if (!in_dev || (ifa = in_dev->ifa_list) == NULL)
 782                 goto out;
 783
 784         for (; ifa; ifa = ifa->ifa_next) {
 785                 if (!buf) {
 786                         done += sizeof(ifr);
 787                         continue;
 788                 }
 789                 if (len < (int) sizeof(ifr))
 790                         break;
 791                 memset(&ifr, 0, sizeof(struct ifreq));
 792                 if (ifa->ifa_label)
 793                         strcpy(ifr.ifr_name, ifa->ifa_label);
 794                 else
 795                         strcpy(ifr.ifr_name, dev->name);
 796
 797                 (*(struct sockaddr_in *)&ifr.ifr_addr).sin_family = AF_INET;
 798                 (*(struct sockaddr_in *)&ifr.ifr_addr).sin_addr.s_addr =
 799                                                                 ifa->ifa_local;
 800
 801                 if (copy_to_user(buf, &ifr, sizeof(struct ifreq))) {
 802                         done = -EFAULT;
 803                         break;
 804                 }
 805                 buf  += sizeof(struct ifreq);
 806                 len  -= sizeof(struct ifreq);
 807                 done += sizeof(struct ifreq);
 808         }
 809 out:
 810         return done;
 811 }
 812
 813 u32 inet_select_addr(const struct net_device *dev, u32 dst, int scope)
 814 {
 815         u32 addr = 0;
 816         struct in_device *in_dev;
 817
 818         rcu_read_lock();
 819         in_dev = __in_dev_get_rcu(dev);
 820         if (!in_dev)
 821                 goto no_in_dev;
 822
 823         for_primary_ifa(in_dev) {
 824                 if (ifa->ifa_scope > scope)
 825                         continue;
 826                 if (!dst || inet_ifa_match(dst, ifa)) {
 827                         addr = ifa->ifa_local;
 828                         break;
 829                 }
 830                 if (!addr)
 831                         addr = ifa->ifa_local;
 832         } endfor_ifa(in_dev);
 833 no_in_dev:
 834         rcu_read_unlock();
 835
 836         if (addr)
 837                 goto out;
 838
 839         /* Not loopback addresses on loopback should be preferred
 840            in this case. It is importnat that lo is the first interface
 841            in dev_base list.
 842          */
 843         read_lock(&dev_base_lock);
 844         rcu_read_lock();
 845         for (dev = dev_base; dev; dev = dev->next) {
 846                 if ((in_dev = __in_dev_get_rcu(dev)) == NULL)
 847                         continue;
 848
 849                 for_primary_ifa(in_dev) {
 850                         if (ifa->ifa_scope != RT_SCOPE_LINK &&
 851                             ifa->ifa_scope <= scope) {
 852                                 addr = ifa->ifa_local;
 853                                 goto out_unlock_both;
 854                         }
 855                 } endfor_ifa(in_dev);
 856         }
 857 out_unlock_both:
 858         read_unlock(&dev_base_lock);
 859         rcu_read_unlock();
 860 out:
 861         return addr;
 862 }
 863
 864 static u32 confirm_addr_indev(struct in_device *in_dev, u32 dst,
 865                               u32 local, int scope)
 866 {
 867         int same = 0;
 868         u32 addr = 0;
 869
 870         for_ifa(in_dev) {
 871                 if (!addr &&
 872                     (local == ifa->ifa_local || !local) &&
 873                     ifa->ifa_scope <= scope) {
 874                         addr = ifa->ifa_local;
 875                         if (same)
 876                                 break;
 877                 }
 878                 if (!same) {
 879                         same = (!local || inet_ifa_match(local, ifa)) &&
 880                                 (!dst || inet_ifa_match(dst, ifa));
 881                         if (same && addr) {
 882                                 if (local || !dst)
 883                                         break;
 884                                 /* Is the selected addr into dst subnet? */
 885                                 if (inet_ifa_match(addr, ifa))
 886                                         break;
 887                                 /* No, then can we use new local src? */
 888                                 if (ifa->ifa_scope <= scope) {
 889                                         addr = ifa->ifa_local;
 890                                         break;
 891                                 }
 892                                 /* search for large dst subnet for addr */
 893                                 same = 0;
 894                         }
 895                 }
 896         } endfor_ifa(in_dev);
 897
 898         return same? addr : 0;
 899 }
 900
 901 /*
 902  * Confirm that local IP address exists using wildcards:
 903  * - dev: only on this interface, 0=any interface
 904  * - dst: only in the same subnet as dst, 0=any dst
 905  * - local: address, 0=autoselect the local address
 906  * - scope: maximum allowed scope value for the local address
 907  */
 908 u32 inet_confirm_addr(const struct net_device *dev, u32 dst, u32 local, int scope)
 909 {
 910         u32 addr = 0;
 911         struct in_device *in_dev;
 912
 913         if (dev) {
 914                 rcu_read_lock();
 915                 if ((in_dev = __in_dev_get_rcu(dev)))
 916                         addr = confirm_addr_indev(in_dev, dst, local, scope);
 917                 rcu_read_unlock();
 918
 919                 return addr;
 920         }
 921
 922         read_lock(&dev_base_lock);
 923         rcu_read_lock();
 924         for (dev = dev_base; dev; dev = dev->next) {
 925                 if ((in_dev = __in_dev_get_rcu(dev))) {
 926                         addr = confirm_addr_indev(in_dev, dst, local, scope);
 927                         if (addr)
 928                                 break;
 929                 }
 930         }
 931         rcu_read_unlock();
 932         read_unlock(&dev_base_lock);
 933
 934         return addr;
 935 }
 936
 937 /*
 938  *      Device notifier
 939  */
 940
 941 int register_inetaddr_notifier(struct notifier_block *nb)
 942 {
 943         return blocking_notifier_chain_register(&inetaddr_chain, nb);
 944 }
 945
 946 int unregister_inetaddr_notifier(struct notifier_block *nb)
 947 {
 948         return blocking_notifier_chain_unregister(&inetaddr_chain, nb);
 949 }
 950
 951 /* Rename ifa_labels for a device name change. Make some effort to preserve existing
 952  * alias numbering and to create unique labels if possible.
 953 */
 954 static void inetdev_changename(struct net_device *dev, struct in_device *in_dev)
 955 {
 956         struct in_ifaddr *ifa;
 957         int named = 0;
 958
 959         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
 960                 char old[IFNAMSIZ], *dot;
 961
 962                 memcpy(old, ifa->ifa_label, IFNAMSIZ);
 963                 memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
 964                 if (named++ == 0)
 965                         continue;
 966                 dot = strchr(ifa->ifa_label, ':');
 967                 if (dot == NULL) {
 968                         sprintf(old, ":%d", named);
 969                         dot = old;
 970                 }
 971                 if (strlen(dot) + strlen(dev->name) < IFNAMSIZ) {
 972                         strcat(ifa->ifa_label, dot);
 973                 } else {
 974                         strcpy(ifa->ifa_label + (IFNAMSIZ - strlen(dot) - 1), dot);
 975                 }
 976         }
 977 }
 978
 979 /* Called only under RTNL semaphore */
 980
 981 static int inetdev_event(struct notifier_block *this, unsigned long event,
 982                          void *ptr)
 983 {
 984         struct net_device *dev = ptr;
 985         struct in_device *in_dev = __in_dev_get_rtnl(dev);
 986
 987         ASSERT_RTNL();
 988
 989         if (!in_dev) {
 990                 if (event == NETDEV_REGISTER && dev == &loopback_dev) {
 991                         in_dev = inetdev_init(dev);
 992                         if (!in_dev)
 993                                 panic("devinet: Failed to create loopback\n");
 994                         in_dev->cnf.no_xfrm = 1;
 995                         in_dev->cnf.no_policy = 1;
 996                 }
 997                 goto out;
 998         }
 999
1000         switch (event) {
1001         case NETDEV_REGISTER:
1002                 printk(KERN_DEBUG "inetdev_event: bug\n");
1003                 dev->ip_ptr = NULL;
1004                 break;
1005         case NETDEV_UP:
1006                 if (dev->mtu < 68)
1007                         break;
1008                 if (dev == &loopback_dev) {
1009                         struct in_ifaddr *ifa;
1010                         if ((ifa = inet_alloc_ifa()) != NULL) {
1011                                 ifa->ifa_local =
1012                                   ifa->ifa_address = htonl(INADDR_LOOPBACK);
1013                                 ifa->ifa_prefixlen = 8;
1014                                 ifa->ifa_mask = inet_make_mask(8);
1015                                 in_dev_hold(in_dev);
1016                                 ifa->ifa_dev = in_dev;
1017                                 ifa->ifa_scope = RT_SCOPE_HOST;
1018                                 memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
1019                                 inet_insert_ifa(ifa);
1020                         }
1021                 }
1022                 ip_mc_up(in_dev);
1023                 break;
1024         case NETDEV_DOWN:
1025                 ip_mc_down(in_dev);
1026                 break;
1027         case NETDEV_CHANGEMTU:
1028                 if (dev->mtu >= 68)
1029                         break;
1030                 /* MTU falled under 68, disable IP */
1031         case NETDEV_UNREGISTER:
1032                 inetdev_destroy(in_dev);
1033                 break;
1034         case NETDEV_CHANGENAME:
1035                 /* Do not notify about label change, this event is
1036                  * not interesting to applications using netlink.
1037                  */
1038                 inetdev_changename(dev, in_dev);
1039
1040 #ifdef CONFIG_SYSCTL
1041                 devinet_sysctl_unregister(&in_dev->cnf);
1042                 neigh_sysctl_unregister(in_dev->arp_parms);
1043                 neigh_sysctl_register(dev, in_dev->arp_parms, NET_IPV4,
1044                                       NET_IPV4_NEIGH, "ipv4", NULL, NULL);
1045                 devinet_sysctl_register(in_dev, &in_dev->cnf);
1046 #endif
1047                 break;
1048         }
1049 out:
1050         return NOTIFY_DONE;
1051 }
1052
1053 static struct notifier_block ip_netdev_notifier = {
1054         .notifier_call =inetdev_event,
1055 };
1056
1057 static int inet_fill_ifaddr(struct sk_buff *skb, struct in_ifaddr *ifa,
1058                             u32 pid, u32 seq, int event, unsigned int flags)
1059 {
1060         struct ifaddrmsg *ifm;
1061         struct nlmsghdr  *nlh;
1062         unsigned char    *b = skb->tail;
1063
1064         nlh = NLMSG_NEW(skb, pid, seq, event, sizeof(*ifm), flags);
1065         ifm = NLMSG_DATA(nlh);
1066         ifm->ifa_family = AF_INET;
1067         ifm->ifa_prefixlen = ifa->ifa_prefixlen;
1068         ifm->ifa_flags = ifa->ifa_flags|IFA_F_PERMANENT;
1069         ifm->ifa_scope = ifa->ifa_scope;
1070         ifm->ifa_index = ifa->ifa_dev->dev->ifindex;
1071         if (ifa->ifa_address)
1072                 RTA_PUT(skb, IFA_ADDRESS, 4, &ifa->ifa_address);
1073         if (ifa->ifa_local)
1074                 RTA_PUT(skb, IFA_LOCAL, 4, &ifa->ifa_local);
1075         if (ifa->ifa_broadcast)
1076                 RTA_PUT(skb, IFA_BROADCAST, 4, &ifa->ifa_broadcast);
1077         if (ifa->ifa_anycast)
1078                 RTA_PUT(skb, IFA_ANYCAST, 4, &ifa->ifa_anycast);
1079         if (ifa->ifa_label[0])
1080                 RTA_PUT(skb, IFA_LABEL, IFNAMSIZ, &ifa->ifa_label);
1081         nlh->nlmsg_len = skb->tail - b;
1082         return skb->len;
1083
1084 nlmsg_failure:
1085 rtattr_failure:
1086         skb_trim(skb, b - skb->data);
1087         return -1;
1088 }
1089
1090 static int inet_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
1091 {
1092         int idx, ip_idx;
1093         struct net_device *dev;
1094         struct in_device *in_dev;
1095         struct in_ifaddr *ifa;
1096         int s_ip_idx, s_idx = cb->args[0];
1097
1098         s_ip_idx = ip_idx = cb->args[1];
1099         read_lock(&dev_base_lock);
1100         for (dev = dev_base, idx = 0; dev; dev = dev->next, idx++) {
1101                 if (idx < s_idx)
1102                         continue;
1103                 if (idx > s_idx)
1104                         s_ip_idx = 0;
1105                 rcu_read_lock();
1106                 if ((in_dev = __in_dev_get_rcu(dev)) == NULL) {
1107                         rcu_read_unlock();
1108                         continue;
1109                 }
1110
1111                 for (ifa = in_dev->ifa_list, ip_idx = 0; ifa;
1112                      ifa = ifa->ifa_next, ip_idx++) {
1113                         if (ip_idx < s_ip_idx)
1114                                 continue;
1115                         if (inet_fill_ifaddr(skb, ifa, NETLINK_CB(cb->skb).pid,
1116                                              cb->nlh->nlmsg_seq,
1117                                              RTM_NEWADDR, NLM_F_MULTI) <= 0) {
1118                                 rcu_read_unlock();
1119                                 goto done;
1120                         }
1121                 }
1122                 rcu_read_unlock();
1123         }
1124
1125 done:
1126         read_unlock(&dev_base_lock);
1127         cb->args[0] = idx;
1128         cb->args[1] = ip_idx;
1129
1130         return skb->len;
1131 }
1132
1133 static void rtmsg_ifa(int event, struct in_ifaddr* ifa)
1134 {
1135         int size = NLMSG_SPACE(sizeof(struct ifaddrmsg) + 128);
1136         struct sk_buff *skb = alloc_skb(size, GFP_KERNEL);
1137
1138         if (!skb)
1139                 netlink_set_err(rtnl, 0, RTNLGRP_IPV4_IFADDR, ENOBUFS);
1140         else if (inet_fill_ifaddr(skb, ifa, 0, 0, event, 0) < 0) {
1141                 kfree_skb(skb);
1142                 netlink_set_err(rtnl, 0, RTNLGRP_IPV4_IFADDR, EINVAL);
1143         } else {
1144                 netlink_broadcast(rtnl, skb, 0, RTNLGRP_IPV4_IFADDR, GFP_KERNEL);
1145         }
1146 }
1147
1148 static struct rtnetlink_link inet_rtnetlink_table[RTM_NR_MSGTYPES] = {
1149         [RTM_NEWADDR  - RTM_BASE] = { .doit     = inet_rtm_newaddr,     },
1150         [RTM_DELADDR  - RTM_BASE] = { .doit     = inet_rtm_deladdr,     },
1151         [RTM_GETADDR  - RTM_BASE] = { .dumpit   = inet_dump_ifaddr,     },
1152         [RTM_NEWROUTE - RTM_BASE] = { .doit     = inet_rtm_newroute,    },
1153         [RTM_DELROUTE - RTM_BASE] = { .doit     = inet_rtm_delroute,    },
1154         [RTM_GETROUTE - RTM_BASE] = { .doit     = inet_rtm_getroute,
1155                                       .dumpit   = inet_dump_fib,        },
1156 #ifdef CONFIG_IP_MULTIPLE_TABLES
1157         [RTM_NEWRULE  - RTM_BASE] = { .doit     = inet_rtm_newrule,     },
1158         [RTM_DELRULE  - RTM_BASE] = { .doit     = inet_rtm_delrule,     },
1159         [RTM_GETRULE  - RTM_BASE] = { .dumpit   = inet_dump_rules,      },
1160 #endif
1161 };
1162
1163 #ifdef CONFIG_SYSCTL
1164
1165 void inet_forward_change(void)
1166 {
1167         struct net_device *dev;
1168         int on = ipv4_devconf.forwarding;
1169
1170         ipv4_devconf.accept_redirects = !on;
1171         ipv4_devconf_dflt.forwarding = on;
1172
1173         read_lock(&dev_base_lock);
1174         for (dev = dev_base; dev; dev = dev->next) {
1175                 struct in_device *in_dev;
1176                 rcu_read_lock();
1177                 in_dev = __in_dev_get_rcu(dev);
1178                 if (in_dev)
1179                         in_dev->cnf.forwarding = on;
1180                 rcu_read_unlock();
1181         }
1182         read_unlock(&dev_base_lock);
1183
1184         rt_cache_flush(0);
1185 }
1186
1187 static int devinet_sysctl_forward(ctl_table *ctl, int write,
1188                                   struct file* filp, void __user *buffer,
1189                                   size_t *lenp, loff_t *ppos)
1190 {
1191         int *valp = ctl->data;
1192         int val = *valp;
1193         int ret = proc_dointvec(ctl, write, filp, buffer, lenp, ppos);
1194
1195         if (write && *valp != val) {
1196                 if (valp == &ipv4_devconf.forwarding)
1197                         inet_forward_change();
1198                 else if (valp != &ipv4_devconf_dflt.forwarding)
1199                         rt_cache_flush(0);
1200         }
1201
1202         return ret;
1203 }
1204
1205 int ipv4_doint_and_flush(ctl_table *ctl, int write,
1206                          struct file* filp, void __user *buffer,
1207                          size_t *lenp, loff_t *ppos)
1208 {
1209         int *valp = ctl->data;
1210         int val = *valp;
1211         int ret = proc_dointvec(ctl, write, filp, buffer, lenp, ppos);
1212
1213         if (write && *valp != val)
1214                 rt_cache_flush(0);
1215
1216         return ret;
1217 }
1218
1219 int ipv4_doint_and_flush_strategy(ctl_table *table, int __user *name, int nlen,
1220                                   void __user *oldval, size_t __user *oldlenp,
1221                                   void __user *newval, size_t newlen,
1222                                   void **context)
1223 {
1224         int *valp = table->data;
1225         int new;
1226
1227         if (!newval || !newlen)
1228                 return 0;
1229
1230         if (newlen != sizeof(int))
1231                 return -EINVAL;
1232
1233         if (get_user(new, (int __user *)newval))
1234                 return -EFAULT;
1235
1236         if (new == *valp)
1237                 return 0;
1238
1239         if (oldval && oldlenp) {
1240                 size_t len;
1241
1242                 if (get_user(len, oldlenp))
1243                         return -EFAULT;
1244
1245                 if (len) {
1246                         if (len > table->maxlen)
1247                                 len = table->maxlen;
1248                         if (copy_to_user(oldval, valp, len))
1249                                 return -EFAULT;
1250                         if (put_user(len, oldlenp))
1251                                 return -EFAULT;
1252                 }
1253         }
1254
1255         *valp = new;
1256         rt_cache_flush(0);
1257         return 1;
1258 }
1259
1260
1261 static struct devinet_sysctl_table {
1262         struct ctl_table_header *sysctl_header;
1263         ctl_table               devinet_vars[__NET_IPV4_CONF_MAX];
1264         ctl_table               devinet_dev[2];
1265         ctl_table               devinet_conf_dir[2];
1266         ctl_table               devinet_proto_dir[2];
1267         ctl_table               devinet_root_dir[2];
1268 } devinet_sysctl = {
1269         .devinet_vars = {
1270                 {
1271                         .ctl_name       = NET_IPV4_CONF_FORWARDING,
1272                         .procname       = "forwarding",
1273                         .data           = &ipv4_devconf.forwarding,
1274                         .maxlen         = sizeof(int),
1275                         .mode           = 0644,
1276                         .proc_handler   = &devinet_sysctl_forward,
1277                 },
1278                 {
1279                         .ctl_name       = NET_IPV4_CONF_MC_FORWARDING,
1280                         .procname       = "mc_forwarding",
1281                         .data           = &ipv4_devconf.mc_forwarding,
1282                         .maxlen         = sizeof(int),
1283                         .mode           = 0444,
1284                         .proc_handler   = &proc_dointvec,
1285                 },
1286                 {
1287                         .ctl_name       = NET_IPV4_CONF_ACCEPT_REDIRECTS,
1288                         .procname       = "accept_redirects",
1289                         .data           = &ipv4_devconf.accept_redirects,
1290                         .maxlen         = sizeof(int),
1291                         .mode           = 0644,
1292                         .proc_handler   = &proc_dointvec,
1293                 },
1294                 {
1295                         .ctl_name       = NET_IPV4_CONF_SECURE_REDIRECTS,
1296                         .procname       = "secure_redirects",
1297                         .data           = &ipv4_devconf.secure_redirects,
1298                         .maxlen         = sizeof(int),
1299                         .mode           = 0644,
1300                         .proc_handler   = &proc_dointvec,
1301                 },
1302                 {
1303                         .ctl_name       = NET_IPV4_CONF_SHARED_MEDIA,
1304                         .procname       = "shared_media",
1305                         .data           = &ipv4_devconf.shared_media,
1306                         .maxlen         = sizeof(int),
1307                         .mode           = 0644,
1308                         .proc_handler   = &proc_dointvec,
1309                 },
1310                 {
1311                         .ctl_name       = NET_IPV4_CONF_RP_FILTER,
1312                         .procname       = "rp_filter",
1313                         .data           = &ipv4_devconf.rp_filter,
1314                         .maxlen         = sizeof(int),
1315                         .mode           = 0644,
1316                         .proc_handler   = &proc_dointvec,
1317                 },
1318                 {
1319                         .ctl_name       = NET_IPV4_CONF_SEND_REDIRECTS,
1320                         .procname       = "send_redirects",
1321                         .data           = &ipv4_devconf.send_redirects,
1322                         .maxlen         = sizeof(int),
1323                         .mode           = 0644,
1324                         .proc_handler   = &proc_dointvec,
1325                 },
1326                 {
1327                         .ctl_name       = NET_IPV4_CONF_ACCEPT_SOURCE_ROUTE,
1328                         .procname       = "accept_source_route",
1329                         .data           = &ipv4_devconf.accept_source_route,
1330                         .maxlen         = sizeof(int),
1331                         .mode           = 0644,
1332                         .proc_handler   = &proc_dointvec,
1333                 },
1334                 {
1335                         .ctl_name       = NET_IPV4_CONF_PROXY_ARP,
1336                         .procname       = "proxy_arp",
1337                         .data           = &ipv4_devconf.proxy_arp,
1338                         .maxlen         = sizeof(int),
1339                         .mode           = 0644,
1340                         .proc_handler   = &proc_dointvec,
1341                 },
1342                 {
1343                         .ctl_name       = NET_IPV4_CONF_MEDIUM_ID,
1344                         .procname       = "medium_id",
1345                         .data           = &ipv4_devconf.medium_id,
1346                         .maxlen         = sizeof(int),
1347                         .mode           = 0644,
1348                         .proc_handler   = &proc_dointvec,
1349                 },
1350                 {
1351                         .ctl_name       = NET_IPV4_CONF_BOOTP_RELAY,
1352                         .procname       = "bootp_relay",
1353                         .data           = &ipv4_devconf.bootp_relay,
1354                         .maxlen         = sizeof(int),
1355                         .mode           = 0644,
1356                         .proc_handler   = &proc_dointvec,
1357                 },
1358                 {
1359                         .ctl_name       = NET_IPV4_CONF_LOG_MARTIANS,
1360                         .procname       = "log_martians",
1361                         .data           = &ipv4_devconf.log_martians,
1362                         .maxlen         = sizeof(int),
1363                         .mode           = 0644,
1364                         .proc_handler   = &proc_dointvec,
1365                 },
1366                 {
1367                         .ctl_name       = NET_IPV4_CONF_TAG,
1368                         .procname       = "tag",
1369                         .data           = &ipv4_devconf.tag,
1370                         .maxlen         = sizeof(int),
1371                         .mode           = 0644,
1372                         .proc_handler   = &proc_dointvec,
1373                 },
1374                 {
1375                         .ctl_name       = NET_IPV4_CONF_ARPFILTER,
1376                         .procname       = "arp_filter",
1377                         .data           = &ipv4_devconf.arp_filter,
1378                         .maxlen         = sizeof(int),
1379                         .mode           = 0644,
1380                         .proc_handler   = &proc_dointvec,
1381                 },
1382                 {
1383                         .ctl_name       = NET_IPV4_CONF_ARP_ANNOUNCE,
1384                         .procname       = "arp_announce",
1385                         .data           = &ipv4_devconf.arp_announce,
1386                         .maxlen         = sizeof(int),
1387                         .mode           = 0644,
1388                         .proc_handler   = &proc_dointvec,
1389                 },
1390                 {
1391                         .ctl_name       = NET_IPV4_CONF_ARP_IGNORE,
1392                         .procname       = "arp_ignore",
1393                         .data           = &ipv4_devconf.arp_ignore,
1394                         .maxlen         = sizeof(int),
1395                         .mode           = 0644,
1396                         .proc_handler   = &proc_dointvec,
1397                 },
1398                 {
1399                         .ctl_name       = NET_IPV4_CONF_ARP_ACCEPT,
1400                         .procname       = "arp_accept",
1401                         .data           = &ipv4_devconf.arp_accept,
1402                         .maxlen         = sizeof(int),
1403                         .mode           = 0644,
1404                         .proc_handler   = &proc_dointvec,
1405                 },
1406                 {
1407                         .ctl_name       = NET_IPV4_CONF_NOXFRM,
1408                         .procname       = "disable_xfrm",
1409                         .data           = &ipv4_devconf.no_xfrm,
1410                         .maxlen         = sizeof(int),
1411                         .mode           = 0644,
1412                         .proc_handler   = &ipv4_doint_and_flush,
1413                         .strategy       = &ipv4_doint_and_flush_strategy,
1414                 },
1415                 {
1416                         .ctl_name       = NET_IPV4_CONF_NOPOLICY,
1417                         .procname       = "disable_policy",
1418                         .data           = &ipv4_devconf.no_policy,
1419                         .maxlen         = sizeof(int),
1420                         .mode           = 0644,
1421                         .proc_handler   = &ipv4_doint_and_flush,
1422                         .strategy       = &ipv4_doint_and_flush_strategy,
1423                 },
1424                 {
1425                         .ctl_name       = NET_IPV4_CONF_FORCE_IGMP_VERSION,
1426                         .procname       = "force_igmp_version",
1427                         .data           = &ipv4_devconf.force_igmp_version,
1428                         .maxlen         = sizeof(int),
1429                         .mode           = 0644,
1430                         .proc_handler   = &ipv4_doint_and_flush,
1431                         .strategy       = &ipv4_doint_and_flush_strategy,
1432                 },
1433                 {
1434                         .ctl_name       = NET_IPV4_CONF_PROMOTE_SECONDARIES,
1435                         .procname       = "promote_secondaries",
1436                         .data           = &ipv4_devconf.promote_secondaries,
1437                         .maxlen         = sizeof(int),
1438                         .mode           = 0644,
1439                         .proc_handler   = &ipv4_doint_and_flush,
1440                         .strategy       = &ipv4_doint_and_flush_strategy,
1441                 },
1442         },
1443         .devinet_dev = {
1444                 {
1445                         .ctl_name       = NET_PROTO_CONF_ALL,
1446                         .procname       = "all",
1447                         .mode           = 0555,
1448                         .child          = devinet_sysctl.devinet_vars,
1449                 },
1450         },
1451         .devinet_conf_dir = {
1452                 {
1453                         .ctl_name       = NET_IPV4_CONF,
1454                         .procname       = "conf",
1455                         .mode           = 0555,
1456                         .child          = devinet_sysctl.devinet_dev,
1457                 },
1458         },
1459         .devinet_proto_dir = {
1460                 {
1461                         .ctl_name       = NET_IPV4,
1462                         .procname       = "ipv4",
1463                         .mode           = 0555,
1464                         .child          = devinet_sysctl.devinet_conf_dir,
1465                 },
1466         },
1467         .devinet_root_dir = {
1468                 {
1469                         .ctl_name       = CTL_NET,
1470                         .procname       = "net",
1471                         .mode           = 0555,
1472                         .child          = devinet_sysctl.devinet_proto_dir,
1473                 },
1474         },
1475 };
1476
1477 static void devinet_sysctl_register(struct in_device *in_dev,
1478                                     struct ipv4_devconf *p)
1479 {
1480         int i;
1481         struct net_device *dev = in_dev ? in_dev->dev : NULL;
1482         struct devinet_sysctl_table *t = kmalloc(sizeof(*t), GFP_KERNEL);
1483         char *dev_name = NULL;
1484
1485         if (!t)
1486                 return;
1487         memcpy(t, &devinet_sysctl, sizeof(*t));
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].de = NULL;
1491         }
1492
1493         if (dev) {
1494                 dev_name = dev->name;
1495                 t->devinet_dev[0].ctl_name = dev->ifindex;
1496         } else {
1497                 dev_name = "default";
1498                 t->devinet_dev[0].ctl_name = NET_PROTO_CONF_DEFAULT;
1499         }
1500
1501         /*
1502          * Make a copy of dev_name, because '.procname' is regarded as const
1503          * by sysctl and we wouldn't want anyone to change it under our feet
1504          * (see SIOCSIFNAME).
1505          */
1506         dev_name = kstrdup(dev_name, GFP_KERNEL);
1507         if (!dev_name)
1508             goto free;
1509
1510         t->devinet_dev[0].procname    = dev_name;
1511         t->devinet_dev[0].child       = t->devinet_vars;
1512         t->devinet_dev[0].de          = NULL;
1513         t->devinet_conf_dir[0].child  = t->devinet_dev;
1514         t->devinet_conf_dir[0].de     = NULL;
1515         t->devinet_proto_dir[0].child = t->devinet_conf_dir;
1516         t->devinet_proto_dir[0].de    = NULL;
1517         t->devinet_root_dir[0].child  = t->devinet_proto_dir;
1518         t->devinet_root_dir[0].de     = NULL;
1519
1520         t->sysctl_header = register_sysctl_table(t->devinet_root_dir, 0);
1521         if (!t->sysctl_header)
1522             goto free_procname;
1523
1524         p->sysctl = t;
1525         return;
1526
1527         /* error path */
1528  free_procname:
1529         kfree(dev_name);
1530  free:
1531         kfree(t);
1532         return;
1533 }
1534
1535 static void devinet_sysctl_unregister(struct ipv4_devconf *p)
1536 {
1537         if (p->sysctl) {
1538                 struct devinet_sysctl_table *t = p->sysctl;
1539                 p->sysctl = NULL;
1540                 unregister_sysctl_table(t->sysctl_header);
1541                 kfree(t->devinet_dev[0].procname);
1542                 kfree(t);
1543         }
1544 }
1545 #endif
1546
1547 void __init devinet_init(void)
1548 {
1549         register_gifconf(PF_INET, inet_gifconf);
1550         register_netdevice_notifier(&ip_netdev_notifier);
1551         rtnetlink_links[PF_INET] = inet_rtnetlink_table;
1552 #ifdef CONFIG_SYSCTL
1553         devinet_sysctl.sysctl_header =
1554                 register_sysctl_table(devinet_sysctl.devinet_root_dir, 0);
1555         devinet_sysctl_register(NULL, &ipv4_devconf_dflt);
1556 #endif
1557 }
1558
1559 EXPORT_SYMBOL(in_dev_finish_destroy);
1560 EXPORT_SYMBOL(inet_select_addr);
1561 EXPORT_SYMBOL(inetdev_by_index);
1562 EXPORT_SYMBOL(register_inetaddr_notifier);
1563 EXPORT_SYMBOL(unregister_inetaddr_notifier);