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/module.h>
  36 #include <linux/types.h>
  37 #include <linux/kernel.h>
  38 #include <linux/sched.h>
  39 #include <linux/string.h>
  40 #include <linux/mm.h>
  41 #include <linux/socket.h>
  42 #include <linux/sockios.h>
  43 #include <linux/in.h>
  44 #include <linux/errno.h>
  45 #include <linux/interrupt.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/rtnetlink.h>
  52 #include <linux/init.h>
  53 #include <linux/notifier.h>
  54 #include <linux/inetdevice.h>
  55 #include <linux/igmp.h>
  56 #ifdef CONFIG_SYSCTL
  57 #include <linux/sysctl.h>
  58 #endif
  59 #include <linux/kmod.h>
  60
  61 #include <net/ip.h>
  62 #include <net/route.h>
  63 #include <net/ip_fib.h>
  64
  65 struct ipv4_devconf ipv4_devconf = {
  66         .accept_redirects = 1,
  67         .send_redirects =  1,
  68         .secure_redirects = 1,
  69         .shared_media =   1,
  70 };
  71
  72 static struct ipv4_devconf ipv4_devconf_dflt = {
  73         .accept_redirects =  1,
  74         .send_redirects =    1,
  75         .secure_redirects =  1,
  76         .shared_media =      1,
  77         .accept_source_route = 1,
  78 };
  79
  80 static void rtmsg_ifa(int event, struct in_ifaddr *);
  81
  82 static struct notifier_block *inetaddr_chain;
  83 static void inet_del_ifa(struct in_device *in_dev, struct in_ifaddr **ifap,
  84                          int destroy);
  85 #ifdef CONFIG_SYSCTL
  86 static void devinet_sysctl_register(struct in_device *in_dev,
  87                                     struct ipv4_devconf *p);
  88 static void devinet_sysctl_unregister(struct ipv4_devconf *p);
  89 #endif
  90
  91 /* Locks all the inet devices. */
  92
  93 static struct in_ifaddr *inet_alloc_ifa(void)
  94 {
  95         struct in_ifaddr *ifa = kmalloc(sizeof(*ifa), GFP_KERNEL);
  96
  97         if (ifa) {
  98                 memset(ifa, 0, sizeof(*ifa));
  99                 INIT_RCU_HEAD(&ifa->rcu_head);
 100         }
 101
 102         return ifa;
 103 }
 104
 105 static void inet_rcu_free_ifa(struct rcu_head *head)
 106 {
 107         struct in_ifaddr *ifa = container_of(head, struct in_ifaddr, rcu_head);
 108         if (ifa->ifa_dev)
 109                 in_dev_put(ifa->ifa_dev);
 110         kfree(ifa);
 111 }
 112
 113 static inline void inet_free_ifa(struct in_ifaddr *ifa)
 114 {
 115         call_rcu(&ifa->rcu_head, inet_rcu_free_ifa);
 116 }
 117
 118 void in_dev_finish_destroy(struct in_device *idev)
 119 {
 120         struct net_device *dev = idev->dev;
 121
 122         BUG_TRAP(!idev->ifa_list);
 123         BUG_TRAP(!idev->mc_list);
 124 #ifdef NET_REFCNT_DEBUG
 125         printk(KERN_DEBUG "in_dev_finish_destroy: %p=%s\n",
 126                idev, dev ? dev->name : "NIL");
 127 #endif
 128         dev_put(dev);
 129         if (!idev->dead)
 130                 printk("Freeing alive in_device %p\n", idev);
 131         else {
 132                 kfree(idev);
 133         }
 134 }
 135
 136 struct in_device *inetdev_init(struct net_device *dev)
 137 {
 138         struct in_device *in_dev;
 139
 140         ASSERT_RTNL();
 141
 142         in_dev = kmalloc(sizeof(*in_dev), GFP_KERNEL);
 143         if (!in_dev)
 144                 goto out;
 145         memset(in_dev, 0, sizeof(*in_dev));
 146         INIT_RCU_HEAD(&in_dev->rcu_head);
 147         memcpy(&in_dev->cnf, &ipv4_devconf_dflt, sizeof(in_dev->cnf));
 148         in_dev->cnf.sysctl = NULL;
 149         in_dev->dev = dev;
 150         if ((in_dev->arp_parms = neigh_parms_alloc(dev, &arp_tbl)) == NULL)
 151                 goto out_kfree;
 152         /* Reference in_dev->dev */
 153         dev_hold(dev);
 154 #ifdef CONFIG_SYSCTL
 155         neigh_sysctl_register(dev, in_dev->arp_parms, NET_IPV4,
 156                               NET_IPV4_NEIGH, "ipv4", NULL, NULL);
 157 #endif
 158
 159         /* Account for reference dev->ip_ptr */
 160         in_dev_hold(in_dev);
 161         rcu_assign_pointer(dev->ip_ptr, in_dev);
 162
 163 #ifdef CONFIG_SYSCTL
 164         devinet_sysctl_register(in_dev, &in_dev->cnf);
 165 #endif
 166         ip_mc_init_dev(in_dev);
 167         if (dev->flags & IFF_UP)
 168                 ip_mc_up(in_dev);
 169 out:
 170         return in_dev;
 171 out_kfree:
 172         kfree(in_dev);
 173         in_dev = NULL;
 174         goto out;
 175 }
 176
 177 static void in_dev_rcu_put(struct rcu_head *head)
 178 {
 179         struct in_device *idev = container_of(head, struct in_device, rcu_head);
 180         in_dev_put(idev);
 181 }
 182
 183 static void inetdev_destroy(struct in_device *in_dev)
 184 {
 185         struct in_ifaddr *ifa;
 186         struct net_device *dev;
 187
 188         ASSERT_RTNL();
 189
 190         dev = in_dev->dev;
 191         if (dev == &loopback_dev)
 192                 return;
 193
 194         in_dev->dead = 1;
 195
 196         ip_mc_destroy_dev(in_dev);
 197
 198         while ((ifa = in_dev->ifa_list) != NULL) {
 199                 inet_del_ifa(in_dev, &in_dev->ifa_list, 0);
 200                 inet_free_ifa(ifa);
 201         }
 202
 203 #ifdef CONFIG_SYSCTL
 204         devinet_sysctl_unregister(&in_dev->cnf);
 205 #endif
 206
 207         dev->ip_ptr = NULL;
 208
 209 #ifdef CONFIG_SYSCTL
 210         neigh_sysctl_unregister(in_dev->arp_parms);
 211 #endif
 212         neigh_parms_release(&arp_tbl, in_dev->arp_parms);
 213         arp_ifdown(dev);
 214
 215         call_rcu(&in_dev->rcu_head, in_dev_rcu_put);
 216 }
 217
 218 int inet_addr_onlink(struct in_device *in_dev, u32 a, u32 b)
 219 {
 220         rcu_read_lock();
 221         for_primary_ifa(in_dev) {
 222                 if (inet_ifa_match(a, ifa)) {
 223                         if (!b || inet_ifa_match(b, ifa)) {
 224                                 rcu_read_unlock();
 225                                 return 1;
 226                         }
 227                 }
 228         } endfor_ifa(in_dev);
 229         rcu_read_unlock();
 230         return 0;
 231 }
 232
 233 static void inet_del_ifa(struct in_device *in_dev, struct in_ifaddr **ifap,
 234                          int destroy)
 235 {
 236         struct in_ifaddr *promote = NULL;
 237         struct in_ifaddr *ifa1 = *ifap;
 238
 239         ASSERT_RTNL();
 240
 241         /* 1. Deleting primary ifaddr forces deletion all secondaries
 242          * unless alias promotion is set
 243          **/
 244
 245         if (!(ifa1->ifa_flags & IFA_F_SECONDARY)) {
 246                 struct in_ifaddr *ifa;
 247                 struct in_ifaddr **ifap1 = &ifa1->ifa_next;
 248
 249                 while ((ifa = *ifap1) != NULL) {
 250                         if (!(ifa->ifa_flags & IFA_F_SECONDARY) ||
 251                             ifa1->ifa_mask != ifa->ifa_mask ||
 252                             !inet_ifa_match(ifa1->ifa_address, ifa)) {
 253                                 ifap1 = &ifa->ifa_next;
 254                                 continue;
 255                         }
 256
 257                         if (!IN_DEV_PROMOTE_SECONDARIES(in_dev)) {
 258                                 *ifap1 = ifa->ifa_next;
 259
 260                                 rtmsg_ifa(RTM_DELADDR, ifa);
 261                                 notifier_call_chain(&inetaddr_chain, NETDEV_DOWN, ifa);
 262                                 inet_free_ifa(ifa);
 263                         } else {
 264                                 promote = ifa;
 265                                 break;
 266                         }
 267                 }
 268         }
 269
 270         /* 2. Unlink it */
 271
 272         *ifap = ifa1->ifa_next;
 273
 274         /* 3. Announce address deletion */
 275
 276         /* Send message first, then call notifier.
 277            At first sight, FIB update triggered by notifier
 278            will refer to already deleted ifaddr, that could confuse
 279            netlink listeners. It is not true: look, gated sees
 280            that route deleted and if it still thinks that ifaddr
 281            is valid, it will try to restore deleted routes... Grr.
 282            So that, this order is correct.
 283          */
 284         rtmsg_ifa(RTM_DELADDR, ifa1);
 285         notifier_call_chain(&inetaddr_chain, NETDEV_DOWN, ifa1);
 286         if (destroy) {
 287                 inet_free_ifa(ifa1);
 288
 289                 if (!in_dev->ifa_list)
 290                         inetdev_destroy(in_dev);
 291         }
 292
 293         if (promote && IN_DEV_PROMOTE_SECONDARIES(in_dev)) {
 294                 /* not sure if we should send a delete notify first? */
 295                 promote->ifa_flags &= ~IFA_F_SECONDARY;
 296                 rtmsg_ifa(RTM_NEWADDR, promote);
 297                 notifier_call_chain(&inetaddr_chain, NETDEV_UP, promote);
 298         }
 299 }
 300
 301 static int inet_insert_ifa(struct in_ifaddr *ifa)
 302 {
 303         struct in_device *in_dev = ifa->ifa_dev;
 304         struct in_ifaddr *ifa1, **ifap, **last_primary;
 305
 306         ASSERT_RTNL();
 307
 308         if (!ifa->ifa_local) {
 309                 inet_free_ifa(ifa);
 310                 return 0;
 311         }
 312
 313         ifa->ifa_flags &= ~IFA_F_SECONDARY;
 314         last_primary = &in_dev->ifa_list;
 315
 316         for (ifap = &in_dev->ifa_list; (ifa1 = *ifap) != NULL;
 317              ifap = &ifa1->ifa_next) {
 318                 if (!(ifa1->ifa_flags & IFA_F_SECONDARY) &&
 319                     ifa->ifa_scope <= ifa1->ifa_scope)
 320                         last_primary = &ifa1->ifa_next;
 321                 if (ifa1->ifa_mask == ifa->ifa_mask &&
 322                     inet_ifa_match(ifa1->ifa_address, ifa)) {
 323                         if (ifa1->ifa_local == ifa->ifa_local) {
 324                                 inet_free_ifa(ifa);
 325                                 return -EEXIST;
 326                         }
 327                         if (ifa1->ifa_scope != ifa->ifa_scope) {
 328                                 inet_free_ifa(ifa);
 329                                 return -EINVAL;
 330                         }
 331                         ifa->ifa_flags |= IFA_F_SECONDARY;
 332                 }
 333         }
 334
 335         if (!(ifa->ifa_flags & IFA_F_SECONDARY)) {
 336                 net_srandom(ifa->ifa_local);
 337                 ifap = last_primary;
 338         }
 339
 340         ifa->ifa_next = *ifap;
 341         *ifap = ifa;
 342
 343         /* Send message first, then call notifier.
 344            Notifier will trigger FIB update, so that
 345            listeners of netlink will know about new ifaddr */
 346         rtmsg_ifa(RTM_NEWADDR, ifa);
 347         notifier_call_chain(&inetaddr_chain, NETDEV_UP, ifa);
 348
 349         return 0;
 350 }
 351
 352 static int inet_set_ifa(struct net_device *dev, struct in_ifaddr *ifa)
 353 {
 354         struct in_device *in_dev = __in_dev_get_rtnl(dev);
 355
 356         ASSERT_RTNL();
 357
 358         if (!in_dev) {
 359                 in_dev = inetdev_init(dev);
 360                 if (!in_dev) {
 361                         inet_free_ifa(ifa);
 362                         return -ENOBUFS;
 363                 }
 364         }
 365         if (ifa->ifa_dev != in_dev) {
 366                 BUG_TRAP(!ifa->ifa_dev);
 367                 in_dev_hold(in_dev);
 368                 ifa->ifa_dev = in_dev;
 369         }
 370         if (LOOPBACK(ifa->ifa_local))
 371                 ifa->ifa_scope = RT_SCOPE_HOST;
 372         return inet_insert_ifa(ifa);
 373 }
 374
 375 struct in_device *inetdev_by_index(int ifindex)
 376 {
 377         struct net_device *dev;
 378         struct in_device *in_dev = NULL;
 379         read_lock(&dev_base_lock);
 380         dev = __dev_get_by_index(ifindex);
 381         if (dev)
 382                 in_dev = in_dev_get(dev);
 383         read_unlock(&dev_base_lock);
 384         return in_dev;
 385 }
 386
 387 /* Called only from RTNL semaphored context. No locks. */
 388
 389 struct in_ifaddr *inet_ifa_byprefix(struct in_device *in_dev, u32 prefix,
 390                                     u32 mask)
 391 {
 392         ASSERT_RTNL();
 393
 394         for_primary_ifa(in_dev) {
 395                 if (ifa->ifa_mask == mask && inet_ifa_match(prefix, ifa))
 396                         return ifa;
 397         } endfor_ifa(in_dev);
 398         return NULL;
 399 }
 400
 401 static int inet_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
 402 {
 403         struct rtattr **rta = arg;
 404         struct in_device *in_dev;
 405         struct ifaddrmsg *ifm = NLMSG_DATA(nlh);
 406         struct in_ifaddr *ifa, **ifap;
 407
 408         ASSERT_RTNL();
 409
 410         if ((in_dev = inetdev_by_index(ifm->ifa_index)) == NULL)
 411                 goto out;
 412         __in_dev_put(in_dev);
 413
 414         for (ifap = &in_dev->ifa_list; (ifa = *ifap) != NULL;
 415              ifap = &ifa->ifa_next) {
 416                 if ((rta[IFA_LOCAL - 1] &&
 417                      memcmp(RTA_DATA(rta[IFA_LOCAL - 1]),
 418                             &ifa->ifa_local, 4)) ||
 419                     (rta[IFA_LABEL - 1] &&
 420                      rtattr_strcmp(rta[IFA_LABEL - 1], ifa->ifa_label)) ||
 421                     (rta[IFA_ADDRESS - 1] &&
 422                      (ifm->ifa_prefixlen != ifa->ifa_prefixlen ||
 423                       !inet_ifa_match(*(u32*)RTA_DATA(rta[IFA_ADDRESS - 1]),
 424                                       ifa))))
 425                         continue;
 426                 inet_del_ifa(in_dev, ifap, 1);
 427                 return 0;
 428         }
 429 out:
 430         return -EADDRNOTAVAIL;
 431 }
 432
 433 static int inet_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
 434 {
 435         struct rtattr **rta = arg;
 436         struct net_device *dev;
 437         struct in_device *in_dev;
 438         struct ifaddrmsg *ifm = NLMSG_DATA(nlh);
 439         struct in_ifaddr *ifa;
 440         int rc = -EINVAL;
 441
 442         ASSERT_RTNL();
 443
 444         if (ifm->ifa_prefixlen > 32 || !rta[IFA_LOCAL - 1])
 445                 goto out;
 446
 447         rc = -ENODEV;
 448         if ((dev = __dev_get_by_index(ifm->ifa_index)) == NULL)
 449                 goto out;
 450
 451         rc = -ENOBUFS;
 452         if ((in_dev = __in_dev_get_rtnl(dev)) == NULL) {
 453                 in_dev = inetdev_init(dev);
 454                 if (!in_dev)
 455                         goto out;
 456         }
 457
 458         if ((ifa = inet_alloc_ifa()) == NULL)
 459                 goto out;
 460
 461         if (!rta[IFA_ADDRESS - 1])
 462                 rta[IFA_ADDRESS - 1] = rta[IFA_LOCAL - 1];
 463         memcpy(&ifa->ifa_local, RTA_DATA(rta[IFA_LOCAL - 1]), 4);
 464         memcpy(&ifa->ifa_address, RTA_DATA(rta[IFA_ADDRESS - 1]), 4);
 465         ifa->ifa_prefixlen = ifm->ifa_prefixlen;
 466         ifa->ifa_mask = inet_make_mask(ifm->ifa_prefixlen);
 467         if (rta[IFA_BROADCAST - 1])
 468                 memcpy(&ifa->ifa_broadcast,
 469                        RTA_DATA(rta[IFA_BROADCAST - 1]), 4);
 470         if (rta[IFA_ANYCAST - 1])
 471                 memcpy(&ifa->ifa_anycast, RTA_DATA(rta[IFA_ANYCAST - 1]), 4);
 472         ifa->ifa_flags = ifm->ifa_flags;
 473         ifa->ifa_scope = ifm->ifa_scope;
 474         in_dev_hold(in_dev);
 475         ifa->ifa_dev   = in_dev;
 476         if (rta[IFA_LABEL - 1])
 477                 rtattr_strlcpy(ifa->ifa_label, rta[IFA_LABEL - 1], IFNAMSIZ);
 478         else
 479                 memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
 480
 481         rc = inet_insert_ifa(ifa);
 482 out:
 483         return rc;
 484 }
 485
 486 /*
 487  *      Determine a default network mask, based on the IP address.
 488  */
 489
 490 static __inline__ int inet_abc_len(u32 addr)
 491 {
 492         int rc = -1;    /* Something else, probably a multicast. */
 493
 494         if (ZERONET(addr))
 495                 rc = 0;
 496         else {
 497                 addr = ntohl(addr);
 498
 499                 if (IN_CLASSA(addr))
 500                         rc = 8;
 501                 else if (IN_CLASSB(addr))
 502                         rc = 16;
 503                 else if (IN_CLASSC(addr))
 504                         rc = 24;
 505         }
 506
 507         return rc;
 508 }
 509
 510
 511 int devinet_ioctl(unsigned int cmd, void __user *arg)
 512 {
 513         struct ifreq ifr;
 514         struct sockaddr_in sin_orig;
 515         struct sockaddr_in *sin = (struct sockaddr_in *)&ifr.ifr_addr;
 516         struct in_device *in_dev;
 517         struct in_ifaddr **ifap = NULL;
 518         struct in_ifaddr *ifa = NULL;
 519         struct net_device *dev;
 520         char *colon;
 521         int ret = -EFAULT;
 522         int tryaddrmatch = 0;
 523
 524         /*
 525          *      Fetch the caller's info block into kernel space
 526          */
 527
 528         if (copy_from_user(&ifr, arg, sizeof(struct ifreq)))
 529                 goto out;
 530         ifr.ifr_name[IFNAMSIZ - 1] = 0;
 531
 532         /* save original address for comparison */
 533         memcpy(&sin_orig, sin, sizeof(*sin));
 534
 535         colon = strchr(ifr.ifr_name, ':');
 536         if (colon)
 537                 *colon = 0;
 538
 539 #ifdef CONFIG_KMOD
 540         dev_load(ifr.ifr_name);
 541 #endif
 542
 543         switch(cmd) {
 544         case SIOCGIFADDR:       /* Get interface address */
 545         case SIOCGIFBRDADDR:    /* Get the broadcast address */
 546         case SIOCGIFDSTADDR:    /* Get the destination address */
 547         case SIOCGIFNETMASK:    /* Get the netmask for the interface */
 548                 /* Note that these ioctls will not sleep,
 549                    so that we do not impose a lock.
 550                    One day we will be forced to put shlock here (I mean SMP)
 551                  */
 552                 tryaddrmatch = (sin_orig.sin_family == AF_INET);
 553                 memset(sin, 0, sizeof(*sin));
 554                 sin->sin_family = AF_INET;
 555                 break;
 556
 557         case SIOCSIFFLAGS:
 558                 ret = -EACCES;
 559                 if (!capable(CAP_NET_ADMIN))
 560                         goto out;
 561                 break;
 562         case SIOCSIFADDR:       /* Set interface address (and family) */
 563         case SIOCSIFBRDADDR:    /* Set the broadcast address */
 564         case SIOCSIFDSTADDR:    /* Set the destination address */
 565         case SIOCSIFNETMASK:    /* Set the netmask for the interface */
 566                 ret = -EACCES;
 567                 if (!capable(CAP_NET_ADMIN))
 568                         goto out;
 569                 ret = -EINVAL;
 570                 if (sin->sin_family != AF_INET)
 571                         goto out;
 572                 break;
 573         default:
 574                 ret = -EINVAL;
 575                 goto out;
 576         }
 577
 578         rtnl_lock();
 579
 580         ret = -ENODEV;
 581         if ((dev = __dev_get_by_name(ifr.ifr_name)) == NULL)
 582                 goto done;
 583
 584         if (colon)
 585                 *colon = ':';
 586
 587         if ((in_dev = __in_dev_get_rtnl(dev)) != NULL) {
 588                 if (tryaddrmatch) {
 589                         /* Matthias Andree */
 590                         /* compare label and address (4.4BSD style) */
 591                         /* note: we only do this for a limited set of ioctls
 592                            and only if the original address family was AF_INET.
 593                            This is checked above. */
 594                         for (ifap = &in_dev->ifa_list; (ifa = *ifap) != NULL;
 595                              ifap = &ifa->ifa_next) {
 596                                 if (!strcmp(ifr.ifr_name, ifa->ifa_label) &&
 597                                     sin_orig.sin_addr.s_addr ==
 598                                                         ifa->ifa_address) {
 599                                         break; /* found */
 600                                 }
 601                         }
 602                 }
 603                 /* we didn't get a match, maybe the application is
 604                    4.3BSD-style and passed in junk so we fall back to
 605                    comparing just the label */
 606                 if (!ifa) {
 607                         for (ifap = &in_dev->ifa_list; (ifa = *ifap) != NULL;
 608                              ifap = &ifa->ifa_next)
 609                                 if (!strcmp(ifr.ifr_name, ifa->ifa_label))
 610                                         break;
 611                 }
 612         }
 613
 614         ret = -EADDRNOTAVAIL;
 615         if (!ifa && cmd != SIOCSIFADDR && cmd != SIOCSIFFLAGS)
 616                 goto done;
 617
 618         switch(cmd) {
 619         case SIOCGIFADDR:       /* Get interface address */
 620                 sin->sin_addr.s_addr = ifa->ifa_local;
 621                 goto rarok;
 622
 623         case SIOCGIFBRDADDR:    /* Get the broadcast address */
 624                 sin->sin_addr.s_addr = ifa->ifa_broadcast;
 625                 goto rarok;
 626
 627         case SIOCGIFDSTADDR:    /* Get the destination address */
 628                 sin->sin_addr.s_addr = ifa->ifa_address;
 629                 goto rarok;
 630
 631         case SIOCGIFNETMASK:    /* Get the netmask for the interface */
 632                 sin->sin_addr.s_addr = ifa->ifa_mask;
 633                 goto rarok;
 634
 635         case SIOCSIFFLAGS:
 636                 if (colon) {
 637                         ret = -EADDRNOTAVAIL;
 638                         if (!ifa)
 639                                 break;
 640                         ret = 0;
 641                         if (!(ifr.ifr_flags & IFF_UP))
 642                                 inet_del_ifa(in_dev, ifap, 1);
 643                         break;
 644                 }
 645                 ret = dev_change_flags(dev, ifr.ifr_flags);
 646                 break;
 647
 648         case SIOCSIFADDR:       /* Set interface address (and family) */
 649                 ret = -EINVAL;
 650                 if (inet_abc_len(sin->sin_addr.s_addr) < 0)
 651                         break;
 652
 653                 if (!ifa) {
 654                         ret = -ENOBUFS;
 655                         if ((ifa = inet_alloc_ifa()) == NULL)
 656                                 break;
 657                         if (colon)
 658                                 memcpy(ifa->ifa_label, ifr.ifr_name, IFNAMSIZ);
 659                         else
 660                                 memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
 661                 } else {
 662                         ret = 0;
 663                         if (ifa->ifa_local == sin->sin_addr.s_addr)
 664                                 break;
 665                         inet_del_ifa(in_dev, ifap, 0);
 666                         ifa->ifa_broadcast = 0;
 667                         ifa->ifa_anycast = 0;
 668                 }
 669
 670                 ifa->ifa_address = ifa->ifa_local = sin->sin_addr.s_addr;
 671
 672                 if (!(dev->flags & IFF_POINTOPOINT)) {
 673                         ifa->ifa_prefixlen = inet_abc_len(ifa->ifa_address);
 674                         ifa->ifa_mask = inet_make_mask(ifa->ifa_prefixlen);
 675                         if ((dev->flags & IFF_BROADCAST) &&
 676                             ifa->ifa_prefixlen < 31)
 677                                 ifa->ifa_broadcast = ifa->ifa_address |
 678                                                      ~ifa->ifa_mask;
 679                 } else {
 680                         ifa->ifa_prefixlen = 32;
 681                         ifa->ifa_mask = inet_make_mask(32);
 682                 }
 683                 ret = inet_set_ifa(dev, ifa);
 684                 break;
 685
 686         case SIOCSIFBRDADDR:    /* Set the broadcast address */
 687                 ret = 0;
 688                 if (ifa->ifa_broadcast != sin->sin_addr.s_addr) {
 689                         inet_del_ifa(in_dev, ifap, 0);
 690                         ifa->ifa_broadcast = sin->sin_addr.s_addr;
 691                         inet_insert_ifa(ifa);
 692                 }
 693                 break;
 694
 695         case SIOCSIFDSTADDR:    /* Set the destination address */
 696                 ret = 0;
 697                 if (ifa->ifa_address == sin->sin_addr.s_addr)
 698                         break;
 699                 ret = -EINVAL;
 700                 if (inet_abc_len(sin->sin_addr.s_addr) < 0)
 701                         break;
 702                 ret = 0;
 703                 inet_del_ifa(in_dev, ifap, 0);
 704                 ifa->ifa_address = sin->sin_addr.s_addr;
 705                 inet_insert_ifa(ifa);
 706                 break;
 707
 708         case SIOCSIFNETMASK:    /* Set the netmask for the interface */
 709
 710                 /*
 711                  *      The mask we set must be legal.
 712                  */
 713                 ret = -EINVAL;
 714                 if (bad_mask(sin->sin_addr.s_addr, 0))
 715                         break;
 716                 ret = 0;
 717                 if (ifa->ifa_mask != sin->sin_addr.s_addr) {
 718                         u32 old_mask = ifa->ifa_mask;
 719                         inet_del_ifa(in_dev, ifap, 0);
 720                         ifa->ifa_mask = sin->sin_addr.s_addr;
 721                         ifa->ifa_prefixlen = inet_mask_len(ifa->ifa_mask);
 722
 723                         /* See if current broadcast address matches
 724                          * with current netmask, then recalculate
 725                          * the broadcast address. Otherwise it's a
 726                          * funny address, so don't touch it since
 727                          * the user seems to know what (s)he's doing...
 728                          */
 729                         if ((dev->flags & IFF_BROADCAST) &&
 730                             (ifa->ifa_prefixlen < 31) &&
 731                             (ifa->ifa_broadcast ==
 732                              (ifa->ifa_local|~old_mask))) {
 733                                 ifa->ifa_broadcast = (ifa->ifa_local |
 734                                                       ~sin->sin_addr.s_addr);
 735                         }
 736                         inet_insert_ifa(ifa);
 737                 }
 738                 break;
 739         }
 740 done:
 741         rtnl_unlock();
 742 out:
 743         return ret;
 744 rarok:
 745         rtnl_unlock();
 746         ret = copy_to_user(arg, &ifr, sizeof(struct ifreq)) ? -EFAULT : 0;
 747         goto out;
 748 }
 749
 750 static int inet_gifconf(struct net_device *dev, char __user *buf, int len)
 751 {
 752         struct in_device *in_dev = __in_dev_get_rtnl(dev);
 753         struct in_ifaddr *ifa;
 754         struct ifreq ifr;
 755         int done = 0;
 756
 757         if (!in_dev || (ifa = in_dev->ifa_list) == NULL)
 758                 goto out;
 759
 760         for (; ifa; ifa = ifa->ifa_next) {
 761                 if (!buf) {
 762                         done += sizeof(ifr);
 763                         continue;
 764                 }
 765                 if (len < (int) sizeof(ifr))
 766                         break;
 767                 memset(&ifr, 0, sizeof(struct ifreq));
 768                 if (ifa->ifa_label)
 769                         strcpy(ifr.ifr_name, ifa->ifa_label);
 770                 else
 771                         strcpy(ifr.ifr_name, dev->name);
 772
 773                 (*(struct sockaddr_in *)&ifr.ifr_addr).sin_family = AF_INET;
 774                 (*(struct sockaddr_in *)&ifr.ifr_addr).sin_addr.s_addr =
 775                                                                 ifa->ifa_local;
 776
 777                 if (copy_to_user(buf, &ifr, sizeof(struct ifreq))) {
 778                         done = -EFAULT;
 779                         break;
 780                 }
 781                 buf  += sizeof(struct ifreq);
 782                 len  -= sizeof(struct ifreq);
 783                 done += sizeof(struct ifreq);
 784         }
 785 out:
 786         return done;
 787 }
 788
 789 u32 inet_select_addr(const struct net_device *dev, u32 dst, int scope)
 790 {
 791         u32 addr = 0;
 792         struct in_device *in_dev;
 793
 794         rcu_read_lock();
 795         in_dev = __in_dev_get_rcu(dev);
 796         if (!in_dev)
 797                 goto no_in_dev;
 798
 799         for_primary_ifa(in_dev) {
 800                 if (ifa->ifa_scope > scope)
 801                         continue;
 802                 if (!dst || inet_ifa_match(dst, ifa)) {
 803                         addr = ifa->ifa_local;
 804                         break;
 805                 }
 806                 if (!addr)
 807                         addr = ifa->ifa_local;
 808         } endfor_ifa(in_dev);
 809 no_in_dev:
 810         rcu_read_unlock();
 811
 812         if (addr)
 813                 goto out;
 814
 815         /* Not loopback addresses on loopback should be preferred
 816            in this case. It is importnat that lo is the first interface
 817            in dev_base list.
 818          */
 819         read_lock(&dev_base_lock);
 820         rcu_read_lock();
 821         for (dev = dev_base; dev; dev = dev->next) {
 822                 if ((in_dev = __in_dev_get_rcu(dev)) == NULL)
 823                         continue;
 824
 825                 for_primary_ifa(in_dev) {
 826                         if (ifa->ifa_scope != RT_SCOPE_LINK &&
 827                             ifa->ifa_scope <= scope) {
 828                                 addr = ifa->ifa_local;
 829                                 goto out_unlock_both;
 830                         }
 831                 } endfor_ifa(in_dev);
 832         }
 833 out_unlock_both:
 834         read_unlock(&dev_base_lock);
 835         rcu_read_unlock();
 836 out:
 837         return addr;
 838 }
 839
 840 static u32 confirm_addr_indev(struct in_device *in_dev, u32 dst,
 841                               u32 local, int scope)
 842 {
 843         int same = 0;
 844         u32 addr = 0;
 845
 846         for_ifa(in_dev) {
 847                 if (!addr &&
 848                     (local == ifa->ifa_local || !local) &&
 849                     ifa->ifa_scope <= scope) {
 850                         addr = ifa->ifa_local;
 851                         if (same)
 852                                 break;
 853                 }
 854                 if (!same) {
 855                         same = (!local || inet_ifa_match(local, ifa)) &&
 856                                 (!dst || inet_ifa_match(dst, ifa));
 857                         if (same && addr) {
 858                                 if (local || !dst)
 859                                         break;
 860                                 /* Is the selected addr into dst subnet? */
 861                                 if (inet_ifa_match(addr, ifa))
 862                                         break;
 863                                 /* No, then can we use new local src? */
 864                                 if (ifa->ifa_scope <= scope) {
 865                                         addr = ifa->ifa_local;
 866                                         break;
 867                                 }
 868                                 /* search for large dst subnet for addr */
 869                                 same = 0;
 870                         }
 871                 }
 872         } endfor_ifa(in_dev);
 873
 874         return same? addr : 0;
 875 }
 876
 877 /*
 878  * Confirm that local IP address exists using wildcards:
 879  * - dev: only on this interface, 0=any interface
 880  * - dst: only in the same subnet as dst, 0=any dst
 881  * - local: address, 0=autoselect the local address
 882  * - scope: maximum allowed scope value for the local address
 883  */
 884 u32 inet_confirm_addr(const struct net_device *dev, u32 dst, u32 local, int scope)
 885 {
 886         u32 addr = 0;
 887         struct in_device *in_dev;
 888
 889         if (dev) {
 890                 rcu_read_lock();
 891                 if ((in_dev = __in_dev_get_rcu(dev)))
 892                         addr = confirm_addr_indev(in_dev, dst, local, scope);
 893                 rcu_read_unlock();
 894
 895                 return addr;
 896         }
 897
 898         read_lock(&dev_base_lock);
 899         rcu_read_lock();
 900         for (dev = dev_base; dev; dev = dev->next) {
 901                 if ((in_dev = __in_dev_get_rcu(dev))) {
 902                         addr = confirm_addr_indev(in_dev, dst, local, scope);
 903                         if (addr)
 904                                 break;
 905                 }
 906         }
 907         rcu_read_unlock();
 908         read_unlock(&dev_base_lock);
 909
 910         return addr;
 911 }
 912
 913 /*
 914  *      Device notifier
 915  */
 916
 917 int register_inetaddr_notifier(struct notifier_block *nb)
 918 {
 919         return notifier_chain_register(&inetaddr_chain, nb);
 920 }
 921
 922 int unregister_inetaddr_notifier(struct notifier_block *nb)
 923 {
 924         return notifier_chain_unregister(&inetaddr_chain, nb);
 925 }
 926
 927 /* Rename ifa_labels for a device name change. Make some effort to preserve existing
 928  * alias numbering and to create unique labels if possible.
 929 */
 930 static void inetdev_changename(struct net_device *dev, struct in_device *in_dev)
 931 {
 932         struct in_ifaddr *ifa;
 933         int named = 0;
 934
 935         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
 936                 char old[IFNAMSIZ], *dot;
 937
 938                 memcpy(old, ifa->ifa_label, IFNAMSIZ);
 939                 memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
 940                 if (named++ == 0)
 941                         continue;
 942                 dot = strchr(ifa->ifa_label, ':');
 943                 if (dot == NULL) {
 944                         sprintf(old, ":%d", named);
 945                         dot = old;
 946                 }
 947                 if (strlen(dot) + strlen(dev->name) < IFNAMSIZ) {
 948                         strcat(ifa->ifa_label, dot);
 949                 } else {
 950                         strcpy(ifa->ifa_label + (IFNAMSIZ - strlen(dot) - 1), dot);
 951                 }
 952         }
 953 }
 954
 955 /* Called only under RTNL semaphore */
 956
 957 static int inetdev_event(struct notifier_block *this, unsigned long event,
 958                          void *ptr)
 959 {
 960         struct net_device *dev = ptr;
 961         struct in_device *in_dev = __in_dev_get_rtnl(dev);
 962
 963         ASSERT_RTNL();
 964
 965         if (!in_dev) {
 966                 if (event == NETDEV_REGISTER && dev == &loopback_dev) {
 967                         in_dev = inetdev_init(dev);
 968                         if (!in_dev)
 969                                 panic("devinet: Failed to create loopback\n");
 970                         in_dev->cnf.no_xfrm = 1;
 971                         in_dev->cnf.no_policy = 1;
 972                 }
 973                 goto out;
 974         }
 975
 976         switch (event) {
 977         case NETDEV_REGISTER:
 978                 printk(KERN_DEBUG "inetdev_event: bug\n");
 979                 dev->ip_ptr = NULL;
 980                 break;
 981         case NETDEV_UP:
 982                 if (dev->mtu < 68)
 983                         break;
 984                 if (dev == &loopback_dev) {
 985                         struct in_ifaddr *ifa;
 986                         if ((ifa = inet_alloc_ifa()) != NULL) {
 987                                 ifa->ifa_local =
 988                                   ifa->ifa_address = htonl(INADDR_LOOPBACK);
 989                                 ifa->ifa_prefixlen = 8;
 990                                 ifa->ifa_mask = inet_make_mask(8);
 991                                 in_dev_hold(in_dev);
 992                                 ifa->ifa_dev = in_dev;
 993                                 ifa->ifa_scope = RT_SCOPE_HOST;
 994                                 memcpy(ifa->ifa_label, dev->name, IFNAMSIZ);
 995                                 inet_insert_ifa(ifa);
 996                         }
 997                 }
 998                 ip_mc_up(in_dev);
 999                 break;
1000         case NETDEV_DOWN:
1001                 ip_mc_down(in_dev);
1002                 break;
1003         case NETDEV_CHANGEMTU:
1004                 if (dev->mtu >= 68)
1005                         break;
1006                 /* MTU falled under 68, disable IP */
1007         case NETDEV_UNREGISTER:
1008                 inetdev_destroy(in_dev);
1009                 break;
1010         case NETDEV_CHANGENAME:
1011                 /* Do not notify about label change, this event is
1012                  * not interesting to applications using netlink.
1013                  */
1014                 inetdev_changename(dev, in_dev);
1015
1016 #ifdef CONFIG_SYSCTL
1017                 devinet_sysctl_unregister(&in_dev->cnf);
1018                 neigh_sysctl_unregister(in_dev->arp_parms);
1019                 neigh_sysctl_register(dev, in_dev->arp_parms, NET_IPV4,
1020                                       NET_IPV4_NEIGH, "ipv4", NULL, NULL);
1021                 devinet_sysctl_register(in_dev, &in_dev->cnf);
1022 #endif
1023                 break;
1024         }
1025 out:
1026         return NOTIFY_DONE;
1027 }
1028
1029 static struct notifier_block ip_netdev_notifier = {
1030         .notifier_call =inetdev_event,
1031 };
1032
1033 static int inet_fill_ifaddr(struct sk_buff *skb, struct in_ifaddr *ifa,
1034                             u32 pid, u32 seq, int event, unsigned int flags)
1035 {
1036         struct ifaddrmsg *ifm;
1037         struct nlmsghdr  *nlh;
1038         unsigned char    *b = skb->tail;
1039
1040         nlh = NLMSG_NEW(skb, pid, seq, event, sizeof(*ifm), flags);
1041         ifm = NLMSG_DATA(nlh);
1042         ifm->ifa_family = AF_INET;
1043         ifm->ifa_prefixlen = ifa->ifa_prefixlen;
1044         ifm->ifa_flags = ifa->ifa_flags|IFA_F_PERMANENT;
1045         ifm->ifa_scope = ifa->ifa_scope;
1046         ifm->ifa_index = ifa->ifa_dev->dev->ifindex;
1047         if (ifa->ifa_address)
1048                 RTA_PUT(skb, IFA_ADDRESS, 4, &ifa->ifa_address);
1049         if (ifa->ifa_local)
1050                 RTA_PUT(skb, IFA_LOCAL, 4, &ifa->ifa_local);
1051         if (ifa->ifa_broadcast)
1052                 RTA_PUT(skb, IFA_BROADCAST, 4, &ifa->ifa_broadcast);
1053         if (ifa->ifa_anycast)
1054                 RTA_PUT(skb, IFA_ANYCAST, 4, &ifa->ifa_anycast);
1055         if (ifa->ifa_label[0])
1056                 RTA_PUT(skb, IFA_LABEL, IFNAMSIZ, &ifa->ifa_label);
1057         nlh->nlmsg_len = skb->tail - b;
1058         return skb->len;
1059
1060 nlmsg_failure:
1061 rtattr_failure:
1062         skb_trim(skb, b - skb->data);
1063         return -1;
1064 }
1065
1066 static int inet_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
1067 {
1068         int idx, ip_idx;
1069         struct net_device *dev;
1070         struct in_device *in_dev;
1071         struct in_ifaddr *ifa;
1072         int s_ip_idx, s_idx = cb->args[0];
1073
1074         s_ip_idx = ip_idx = cb->args[1];
1075         read_lock(&dev_base_lock);
1076         for (dev = dev_base, idx = 0; dev; dev = dev->next, idx++) {
1077                 if (idx < s_idx)
1078                         continue;
1079                 if (idx > s_idx)
1080                         s_ip_idx = 0;
1081                 rcu_read_lock();
1082                 if ((in_dev = __in_dev_get_rcu(dev)) == NULL) {
1083                         rcu_read_unlock();
1084                         continue;
1085                 }
1086
1087                 for (ifa = in_dev->ifa_list, ip_idx = 0; ifa;
1088                      ifa = ifa->ifa_next, ip_idx++) {
1089                         if (ip_idx < s_ip_idx)
1090                                 continue;
1091                         if (inet_fill_ifaddr(skb, ifa, NETLINK_CB(cb->skb).pid,
1092                                              cb->nlh->nlmsg_seq,
1093                                              RTM_NEWADDR, NLM_F_MULTI) <= 0) {
1094                                 rcu_read_unlock();
1095                                 goto done;
1096                         }
1097                 }
1098                 rcu_read_unlock();
1099         }
1100
1101 done:
1102         read_unlock(&dev_base_lock);
1103         cb->args[0] = idx;
1104         cb->args[1] = ip_idx;
1105
1106         return skb->len;
1107 }
1108
1109 static void rtmsg_ifa(int event, struct in_ifaddr* ifa)
1110 {
1111         int size = NLMSG_SPACE(sizeof(struct ifaddrmsg) + 128);
1112         struct sk_buff *skb = alloc_skb(size, GFP_KERNEL);
1113
1114         if (!skb)
1115                 netlink_set_err(rtnl, 0, RTNLGRP_IPV4_IFADDR, ENOBUFS);
1116         else if (inet_fill_ifaddr(skb, ifa, current->pid, 0, event, 0) < 0) {
1117                 kfree_skb(skb);
1118                 netlink_set_err(rtnl, 0, RTNLGRP_IPV4_IFADDR, EINVAL);
1119         } else {
1120                 netlink_broadcast(rtnl, skb, 0, RTNLGRP_IPV4_IFADDR, GFP_KERNEL);
1121         }
1122 }
1123
1124 static struct rtnetlink_link inet_rtnetlink_table[RTM_NR_MSGTYPES] = {
1125         [RTM_NEWADDR  - RTM_BASE] = { .doit     = inet_rtm_newaddr,     },
1126         [RTM_DELADDR  - RTM_BASE] = { .doit     = inet_rtm_deladdr,     },
1127         [RTM_GETADDR  - RTM_BASE] = { .dumpit   = inet_dump_ifaddr,     },
1128         [RTM_NEWROUTE - RTM_BASE] = { .doit     = inet_rtm_newroute,    },
1129         [RTM_DELROUTE - RTM_BASE] = { .doit     = inet_rtm_delroute,    },
1130         [RTM_GETROUTE - RTM_BASE] = { .doit     = inet_rtm_getroute,
1131                                       .dumpit   = inet_dump_fib,        },
1132 #ifdef CONFIG_IP_MULTIPLE_TABLES
1133         [RTM_NEWRULE  - RTM_BASE] = { .doit     = inet_rtm_newrule,     },
1134         [RTM_DELRULE  - RTM_BASE] = { .doit     = inet_rtm_delrule,     },
1135         [RTM_GETRULE  - RTM_BASE] = { .dumpit   = inet_dump_rules,      },
1136 #endif
1137 };
1138
1139 #ifdef CONFIG_SYSCTL
1140
1141 void inet_forward_change(void)
1142 {
1143         struct net_device *dev;
1144         int on = ipv4_devconf.forwarding;
1145
1146         ipv4_devconf.accept_redirects = !on;
1147         ipv4_devconf_dflt.forwarding = on;
1148
1149         read_lock(&dev_base_lock);
1150         for (dev = dev_base; dev; dev = dev->next) {
1151                 struct in_device *in_dev;
1152                 rcu_read_lock();
1153                 in_dev = __in_dev_get_rcu(dev);
1154                 if (in_dev)
1155                         in_dev->cnf.forwarding = on;
1156                 rcu_read_unlock();
1157         }
1158         read_unlock(&dev_base_lock);
1159
1160         rt_cache_flush(0);
1161 }
1162
1163 static int devinet_sysctl_forward(ctl_table *ctl, int write,
1164                                   struct file* filp, void __user *buffer,
1165                                   size_t *lenp, loff_t *ppos)
1166 {
1167         int *valp = ctl->data;
1168         int val = *valp;
1169         int ret = proc_dointvec(ctl, write, filp, buffer, lenp, ppos);
1170
1171         if (write && *valp != val) {
1172                 if (valp == &ipv4_devconf.forwarding)
1173                         inet_forward_change();
1174                 else if (valp != &ipv4_devconf_dflt.forwarding)
1175                         rt_cache_flush(0);
1176         }
1177
1178         return ret;
1179 }
1180
1181 int ipv4_doint_and_flush(ctl_table *ctl, int write,
1182                          struct file* filp, void __user *buffer,
1183                          size_t *lenp, loff_t *ppos)
1184 {
1185         int *valp = ctl->data;
1186         int val = *valp;
1187         int ret = proc_dointvec(ctl, write, filp, buffer, lenp, ppos);
1188
1189         if (write && *valp != val)
1190                 rt_cache_flush(0);
1191
1192         return ret;
1193 }
1194
1195 int ipv4_doint_and_flush_strategy(ctl_table *table, int __user *name, int nlen,
1196                                   void __user *oldval, size_t __user *oldlenp,
1197                                   void __user *newval, size_t newlen,
1198                                   void **context)
1199 {
1200         int *valp = table->data;
1201         int new;
1202
1203         if (!newval || !newlen)
1204                 return 0;
1205
1206         if (newlen != sizeof(int))
1207                 return -EINVAL;
1208
1209         if (get_user(new, (int __user *)newval))
1210                 return -EFAULT;
1211
1212         if (new == *valp)
1213                 return 0;
1214
1215         if (oldval && oldlenp) {
1216                 size_t len;
1217
1218                 if (get_user(len, oldlenp))
1219                         return -EFAULT;
1220
1221                 if (len) {
1222                         if (len > table->maxlen)
1223                                 len = table->maxlen;
1224                         if (copy_to_user(oldval, valp, len))
1225                                 return -EFAULT;
1226                         if (put_user(len, oldlenp))
1227                                 return -EFAULT;
1228                 }
1229         }
1230
1231         *valp = new;
1232         rt_cache_flush(0);
1233         return 1;
1234 }
1235
1236
1237 static struct devinet_sysctl_table {
1238         struct ctl_table_header *sysctl_header;
1239         ctl_table               devinet_vars[__NET_IPV4_CONF_MAX];
1240         ctl_table               devinet_dev[2];
1241         ctl_table               devinet_conf_dir[2];
1242         ctl_table               devinet_proto_dir[2];
1243         ctl_table               devinet_root_dir[2];
1244 } devinet_sysctl = {
1245         .devinet_vars = {
1246                 {
1247                         .ctl_name       = NET_IPV4_CONF_FORWARDING,
1248                         .procname       = "forwarding",
1249                         .data           = &ipv4_devconf.forwarding,
1250                         .maxlen         = sizeof(int),
1251                         .mode           = 0644,
1252                         .proc_handler   = &devinet_sysctl_forward,
1253                 },
1254                 {
1255                         .ctl_name       = NET_IPV4_CONF_MC_FORWARDING,
1256                         .procname       = "mc_forwarding",
1257                         .data           = &ipv4_devconf.mc_forwarding,
1258                         .maxlen         = sizeof(int),
1259                         .mode           = 0444,
1260                         .proc_handler   = &proc_dointvec,
1261                 },
1262                 {
1263                         .ctl_name       = NET_IPV4_CONF_ACCEPT_REDIRECTS,
1264                         .procname       = "accept_redirects",
1265                         .data           = &ipv4_devconf.accept_redirects,
1266                         .maxlen         = sizeof(int),
1267                         .mode           = 0644,
1268                         .proc_handler   = &proc_dointvec,
1269                 },
1270                 {
1271                         .ctl_name       = NET_IPV4_CONF_SECURE_REDIRECTS,
1272                         .procname       = "secure_redirects",
1273                         .data           = &ipv4_devconf.secure_redirects,
1274                         .maxlen         = sizeof(int),
1275                         .mode           = 0644,
1276                         .proc_handler   = &proc_dointvec,
1277                 },
1278                 {
1279                         .ctl_name       = NET_IPV4_CONF_SHARED_MEDIA,
1280                         .procname       = "shared_media",
1281                         .data           = &ipv4_devconf.shared_media,
1282                         .maxlen         = sizeof(int),
1283                         .mode           = 0644,
1284                         .proc_handler   = &proc_dointvec,
1285                 },
1286                 {
1287                         .ctl_name       = NET_IPV4_CONF_RP_FILTER,
1288                         .procname       = "rp_filter",
1289                         .data           = &ipv4_devconf.rp_filter,
1290                         .maxlen         = sizeof(int),
1291                         .mode           = 0644,
1292                         .proc_handler   = &proc_dointvec,
1293                 },
1294                 {
1295                         .ctl_name       = NET_IPV4_CONF_SEND_REDIRECTS,
1296                         .procname       = "send_redirects",
1297                         .data           = &ipv4_devconf.send_redirects,
1298                         .maxlen         = sizeof(int),
1299                         .mode           = 0644,
1300                         .proc_handler   = &proc_dointvec,
1301                 },
1302                 {
1303                         .ctl_name       = NET_IPV4_CONF_ACCEPT_SOURCE_ROUTE,
1304                         .procname       = "accept_source_route",
1305                         .data           = &ipv4_devconf.accept_source_route,
1306                         .maxlen         = sizeof(int),
1307                         .mode           = 0644,
1308                         .proc_handler   = &proc_dointvec,
1309                 },
1310                 {
1311                         .ctl_name       = NET_IPV4_CONF_PROXY_ARP,
1312                         .procname       = "proxy_arp",
1313                         .data           = &ipv4_devconf.proxy_arp,
1314                         .maxlen         = sizeof(int),
1315                         .mode           = 0644,
1316                         .proc_handler   = &proc_dointvec,
1317                 },
1318                 {
1319                         .ctl_name       = NET_IPV4_CONF_MEDIUM_ID,
1320                         .procname       = "medium_id",
1321                         .data           = &ipv4_devconf.medium_id,
1322                         .maxlen         = sizeof(int),
1323                         .mode           = 0644,
1324                         .proc_handler   = &proc_dointvec,
1325                 },
1326                 {
1327                         .ctl_name       = NET_IPV4_CONF_BOOTP_RELAY,
1328                         .procname       = "bootp_relay",
1329                         .data           = &ipv4_devconf.bootp_relay,
1330                         .maxlen         = sizeof(int),
1331                         .mode           = 0644,
1332                         .proc_handler   = &proc_dointvec,
1333                 },
1334                 {
1335                         .ctl_name       = NET_IPV4_CONF_LOG_MARTIANS,
1336                         .procname       = "log_martians",
1337                         .data           = &ipv4_devconf.log_martians,
1338                         .maxlen         = sizeof(int),
1339                         .mode           = 0644,
1340                         .proc_handler   = &proc_dointvec,
1341                 },
1342                 {
1343                         .ctl_name       = NET_IPV4_CONF_TAG,
1344                         .procname       = "tag",
1345                         .data           = &ipv4_devconf.tag,
1346                         .maxlen         = sizeof(int),
1347                         .mode           = 0644,
1348                         .proc_handler   = &proc_dointvec,
1349                 },
1350                 {
1351                         .ctl_name       = NET_IPV4_CONF_ARPFILTER,
1352                         .procname       = "arp_filter",
1353                         .data           = &ipv4_devconf.arp_filter,
1354                         .maxlen         = sizeof(int),
1355                         .mode           = 0644,
1356                         .proc_handler   = &proc_dointvec,
1357                 },
1358                 {
1359                         .ctl_name       = NET_IPV4_CONF_ARP_ANNOUNCE,
1360                         .procname       = "arp_announce",
1361                         .data           = &ipv4_devconf.arp_announce,
1362                         .maxlen         = sizeof(int),
1363                         .mode           = 0644,
1364                         .proc_handler   = &proc_dointvec,
1365                 },
1366                 {
1367                         .ctl_name       = NET_IPV4_CONF_ARP_IGNORE,
1368                         .procname       = "arp_ignore",
1369                         .data           = &ipv4_devconf.arp_ignore,
1370                         .maxlen         = sizeof(int),
1371                         .mode           = 0644,
1372                         .proc_handler   = &proc_dointvec,
1373                 },
1374                 {
1375                         .ctl_name       = NET_IPV4_CONF_NOXFRM,
1376                         .procname       = "disable_xfrm",
1377                         .data           = &ipv4_devconf.no_xfrm,
1378                         .maxlen         = sizeof(int),
1379                         .mode           = 0644,
1380                         .proc_handler   = &ipv4_doint_and_flush,
1381                         .strategy       = &ipv4_doint_and_flush_strategy,
1382                 },
1383                 {
1384                         .ctl_name       = NET_IPV4_CONF_NOPOLICY,
1385                         .procname       = "disable_policy",
1386                         .data           = &ipv4_devconf.no_policy,
1387                         .maxlen         = sizeof(int),
1388                         .mode           = 0644,
1389                         .proc_handler   = &ipv4_doint_and_flush,
1390                         .strategy       = &ipv4_doint_and_flush_strategy,
1391                 },
1392                 {
1393                         .ctl_name       = NET_IPV4_CONF_FORCE_IGMP_VERSION,
1394                         .procname       = "force_igmp_version",
1395                         .data           = &ipv4_devconf.force_igmp_version,
1396                         .maxlen         = sizeof(int),
1397                         .mode           = 0644,
1398                         .proc_handler   = &ipv4_doint_and_flush,
1399                         .strategy       = &ipv4_doint_and_flush_strategy,
1400                 },
1401                 {
1402                         .ctl_name       = NET_IPV4_CONF_PROMOTE_SECONDARIES,
1403                         .procname       = "promote_secondaries",
1404                         .data           = &ipv4_devconf.promote_secondaries,
1405                         .maxlen         = sizeof(int),
1406                         .mode           = 0644,
1407                         .proc_handler   = &ipv4_doint_and_flush,
1408                         .strategy       = &ipv4_doint_and_flush_strategy,
1409                 },
1410         },
1411         .devinet_dev = {
1412                 {
1413                         .ctl_name       = NET_PROTO_CONF_ALL,
1414                         .procname       = "all",
1415                         .mode           = 0555,
1416                         .child          = devinet_sysctl.devinet_vars,
1417                 },
1418         },
1419         .devinet_conf_dir = {
1420                 {
1421                         .ctl_name       = NET_IPV4_CONF,
1422                         .procname       = "conf",
1423                         .mode           = 0555,
1424                         .child          = devinet_sysctl.devinet_dev,
1425                 },
1426         },
1427         .devinet_proto_dir = {
1428                 {
1429                         .ctl_name       = NET_IPV4,
1430                         .procname       = "ipv4",
1431                         .mode           = 0555,
1432                         .child          = devinet_sysctl.devinet_conf_dir,
1433                 },
1434         },
1435         .devinet_root_dir = {
1436                 {
1437                         .ctl_name       = CTL_NET,
1438                         .procname       = "net",
1439                         .mode           = 0555,
1440                         .child          = devinet_sysctl.devinet_proto_dir,
1441                 },
1442         },
1443 };
1444
1445 static void devinet_sysctl_register(struct in_device *in_dev,
1446                                     struct ipv4_devconf *p)
1447 {
1448         int i;
1449         struct net_device *dev = in_dev ? in_dev->dev : NULL;
1450         struct devinet_sysctl_table *t = kmalloc(sizeof(*t), GFP_KERNEL);
1451         char *dev_name = NULL;
1452
1453         if (!t)
1454                 return;
1455         memcpy(t, &devinet_sysctl, sizeof(*t));
1456         for (i = 0; i < ARRAY_SIZE(t->devinet_vars) - 1; i++) {
1457                 t->devinet_vars[i].data += (char *)p - (char *)&ipv4_devconf;
1458                 t->devinet_vars[i].de = NULL;
1459         }
1460
1461         if (dev) {
1462                 dev_name = dev->name;
1463                 t->devinet_dev[0].ctl_name = dev->ifindex;
1464         } else {
1465                 dev_name = "default";
1466                 t->devinet_dev[0].ctl_name = NET_PROTO_CONF_DEFAULT;
1467         }
1468
1469         /*
1470          * Make a copy of dev_name, because '.procname' is regarded as const
1471          * by sysctl and we wouldn't want anyone to change it under our feet
1472          * (see SIOCSIFNAME).
1473          */
1474         dev_name = kstrdup(dev_name, GFP_KERNEL);
1475         if (!dev_name)
1476             goto free;
1477
1478         t->devinet_dev[0].procname    = dev_name;
1479         t->devinet_dev[0].child       = t->devinet_vars;
1480         t->devinet_dev[0].de          = NULL;
1481         t->devinet_conf_dir[0].child  = t->devinet_dev;
1482         t->devinet_conf_dir[0].de     = NULL;
1483         t->devinet_proto_dir[0].child = t->devinet_conf_dir;
1484         t->devinet_proto_dir[0].de    = NULL;
1485         t->devinet_root_dir[0].child  = t->devinet_proto_dir;
1486         t->devinet_root_dir[0].de     = NULL;
1487
1488         t->sysctl_header = register_sysctl_table(t->devinet_root_dir, 0);
1489         if (!t->sysctl_header)
1490             goto free_procname;
1491
1492         p->sysctl = t;
1493         return;
1494
1495         /* error path */
1496  free_procname:
1497         kfree(dev_name);
1498  free:
1499         kfree(t);
1500         return;
1501 }
1502
1503 static void devinet_sysctl_unregister(struct ipv4_devconf *p)
1504 {
1505         if (p->sysctl) {
1506                 struct devinet_sysctl_table *t = p->sysctl;
1507                 p->sysctl = NULL;
1508                 unregister_sysctl_table(t->sysctl_header);
1509                 kfree(t->devinet_dev[0].procname);
1510                 kfree(t);
1511         }
1512 }
1513 #endif
1514
1515 void __init devinet_init(void)
1516 {
1517         register_gifconf(PF_INET, inet_gifconf);
1518         register_netdevice_notifier(&ip_netdev_notifier);
1519         rtnetlink_links[PF_INET] = inet_rtnetlink_table;
1520 #ifdef CONFIG_SYSCTL
1521         devinet_sysctl.sysctl_header =
1522                 register_sysctl_table(devinet_sysctl.devinet_root_dir, 0);
1523         devinet_sysctl_register(NULL, &ipv4_devconf_dflt);
1524 #endif
1525 }
1526
1527 EXPORT_SYMBOL(devinet_ioctl);
1528 EXPORT_SYMBOL(in_dev_finish_destroy);
1529 EXPORT_SYMBOL(inet_select_addr);
1530 EXPORT_SYMBOL(inetdev_by_index);
1531 EXPORT_SYMBOL(register_inetaddr_notifier);
1532 EXPORT_SYMBOL(unregister_inetaddr_notifier);