net/core/rtnetlink.c

   1 /*
   2  * INET         An implementation of the TCP/IP protocol suite for the LINUX
   3  *              operating system.  INET is implemented using the  BSD Socket
   4  *              interface as the means of communication with the user level.
   5  *
   6  *              Routing netlink socket interface: protocol independent part.
   7  *
   8  * Authors:     Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
   9  *
  10  *              This program is free software; you can redistribute it and/or
  11  *              modify it under the terms of the GNU General Public License
  12  *              as published by the Free Software Foundation; either version
  13  *              2 of the License, or (at your option) any later version.
  14  *
  15  *      Fixes:
  16  *      Vitaly E. Lavrov                RTA_OK arithmetics was wrong.
  17  */
  18
  19 #include <linux/config.h>
  20 #include <linux/errno.h>
  21 #include <linux/module.h>
  22 #include <linux/types.h>
  23 #include <linux/socket.h>
  24 #include <linux/kernel.h>
  25 #include <linux/sched.h>
  26 #include <linux/timer.h>
  27 #include <linux/string.h>
  28 #include <linux/sockios.h>
  29 #include <linux/net.h>
  30 #include <linux/fcntl.h>
  31 #include <linux/mm.h>
  32 #include <linux/slab.h>
  33 #include <linux/interrupt.h>
  34 #include <linux/capability.h>
  35 #include <linux/skbuff.h>
  36 #include <linux/init.h>
  37 #include <linux/security.h>
  38 #include <linux/mutex.h>
  39
  40 #include <asm/uaccess.h>
  41 #include <asm/system.h>
  42 #include <asm/string.h>
  43
  44 #include <linux/inet.h>
  45 #include <linux/netdevice.h>
  46 #include <net/ip.h>
  47 #include <net/protocol.h>
  48 #include <net/arp.h>
  49 #include <net/route.h>
  50 #include <net/udp.h>
  51 #include <net/sock.h>
  52 #include <net/pkt_sched.h>
  53 #include <net/netlink.h>
  54
  55 static DEFINE_MUTEX(rtnl_mutex);
  56
  57 void rtnl_lock(void)
  58 {
  59         mutex_lock(&rtnl_mutex);
  60 }
  61
  62 void __rtnl_unlock(void)
  63 {
  64         mutex_unlock(&rtnl_mutex);
  65 }
  66
  67 void rtnl_unlock(void)
  68 {
  69         mutex_unlock(&rtnl_mutex);
  70         if (rtnl && rtnl->sk_receive_queue.qlen)
  71                 rtnl->sk_data_ready(rtnl, 0);
  72         netdev_run_todo();
  73 }
  74
  75 int rtnl_trylock(void)
  76 {
  77         return mutex_trylock(&rtnl_mutex);
  78 }
  79
  80 int rtattr_parse(struct rtattr *tb[], int maxattr, struct rtattr *rta, int len)
  81 {
  82         memset(tb, 0, sizeof(struct rtattr*)*maxattr);
  83
  84         while (RTA_OK(rta, len)) {
  85                 unsigned flavor = rta->rta_type;
  86                 if (flavor && flavor <= maxattr)
  87                         tb[flavor-1] = rta;
  88                 rta = RTA_NEXT(rta, len);
  89         }
  90         return 0;
  91 }
  92
  93 struct sock *rtnl;
  94
  95 struct rtnetlink_link * rtnetlink_links[NPROTO];
  96
  97 static const int rtm_min[RTM_NR_FAMILIES] =
  98 {
  99         [RTM_FAM(RTM_NEWLINK)]      = NLMSG_LENGTH(sizeof(struct ifinfomsg)),
 100         [RTM_FAM(RTM_NEWADDR)]      = NLMSG_LENGTH(sizeof(struct ifaddrmsg)),
 101         [RTM_FAM(RTM_NEWROUTE)]     = NLMSG_LENGTH(sizeof(struct rtmsg)),
 102         [RTM_FAM(RTM_NEWNEIGH)]     = NLMSG_LENGTH(sizeof(struct ndmsg)),
 103         [RTM_FAM(RTM_NEWRULE)]      = NLMSG_LENGTH(sizeof(struct rtmsg)),
 104         [RTM_FAM(RTM_NEWQDISC)]     = NLMSG_LENGTH(sizeof(struct tcmsg)),
 105         [RTM_FAM(RTM_NEWTCLASS)]    = NLMSG_LENGTH(sizeof(struct tcmsg)),
 106         [RTM_FAM(RTM_NEWTFILTER)]   = NLMSG_LENGTH(sizeof(struct tcmsg)),
 107         [RTM_FAM(RTM_NEWACTION)]    = NLMSG_LENGTH(sizeof(struct tcamsg)),
 108         [RTM_FAM(RTM_NEWPREFIX)]    = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
 109         [RTM_FAM(RTM_GETMULTICAST)] = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
 110         [RTM_FAM(RTM_GETANYCAST)]   = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
 111         [RTM_FAM(RTM_NEWNEIGHTBL)]  = NLMSG_LENGTH(sizeof(struct ndtmsg)),
 112 };
 113
 114 static const int rta_max[RTM_NR_FAMILIES] =
 115 {
 116         [RTM_FAM(RTM_NEWLINK)]      = IFLA_MAX,
 117         [RTM_FAM(RTM_NEWADDR)]      = IFA_MAX,
 118         [RTM_FAM(RTM_NEWROUTE)]     = RTA_MAX,
 119         [RTM_FAM(RTM_NEWNEIGH)]     = NDA_MAX,
 120         [RTM_FAM(RTM_NEWRULE)]      = RTA_MAX,
 121         [RTM_FAM(RTM_NEWQDISC)]     = TCA_MAX,
 122         [RTM_FAM(RTM_NEWTCLASS)]    = TCA_MAX,
 123         [RTM_FAM(RTM_NEWTFILTER)]   = TCA_MAX,
 124         [RTM_FAM(RTM_NEWACTION)]    = TCAA_MAX,
 125         [RTM_FAM(RTM_NEWNEIGHTBL)]  = NDTA_MAX,
 126 };
 127
 128 void __rta_fill(struct sk_buff *skb, int attrtype, int attrlen, const void *data)
 129 {
 130         struct rtattr *rta;
 131         int size = RTA_LENGTH(attrlen);
 132
 133         rta = (struct rtattr*)skb_put(skb, RTA_ALIGN(size));
 134         rta->rta_type = attrtype;
 135         rta->rta_len = size;
 136         memcpy(RTA_DATA(rta), data, attrlen);
 137         memset(RTA_DATA(rta) + attrlen, 0, RTA_ALIGN(size) - size);
 138 }
 139
 140 size_t rtattr_strlcpy(char *dest, const struct rtattr *rta, size_t size)
 141 {
 142         size_t ret = RTA_PAYLOAD(rta);
 143         char *src = RTA_DATA(rta);
 144
 145         if (ret > 0 && src[ret - 1] == '\0')
 146                 ret--;
 147         if (size > 0) {
 148                 size_t len = (ret >= size) ? size - 1 : ret;
 149                 memset(dest, 0, size);
 150                 memcpy(dest, src, len);
 151         }
 152         return ret;
 153 }
 154
 155 int rtnetlink_send(struct sk_buff *skb, u32 pid, unsigned group, int echo)
 156 {
 157         int err = 0;
 158
 159         NETLINK_CB(skb).dst_group = group;
 160         if (echo)
 161                 atomic_inc(&skb->users);
 162         netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
 163         if (echo)
 164                 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
 165         return err;
 166 }
 167
 168 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
 169 {
 170         struct rtattr *mx = (struct rtattr*)skb->tail;
 171         int i;
 172
 173         RTA_PUT(skb, RTA_METRICS, 0, NULL);
 174         for (i=0; i<RTAX_MAX; i++) {
 175                 if (metrics[i])
 176                         RTA_PUT(skb, i+1, sizeof(u32), metrics+i);
 177         }
 178         mx->rta_len = skb->tail - (u8*)mx;
 179         if (mx->rta_len == RTA_LENGTH(0))
 180                 skb_trim(skb, (u8*)mx - skb->data);
 181         return 0;
 182
 183 rtattr_failure:
 184         skb_trim(skb, (u8*)mx - skb->data);
 185         return -1;
 186 }
 187
 188
 189 static void set_operstate(struct net_device *dev, unsigned char transition)
 190 {
 191         unsigned char operstate = dev->operstate;
 192
 193         switch(transition) {
 194         case IF_OPER_UP:
 195                 if ((operstate == IF_OPER_DORMANT ||
 196                      operstate == IF_OPER_UNKNOWN) &&
 197                     !netif_dormant(dev))
 198                         operstate = IF_OPER_UP;
 199                 break;
 200
 201         case IF_OPER_DORMANT:
 202                 if (operstate == IF_OPER_UP ||
 203                     operstate == IF_OPER_UNKNOWN)
 204                         operstate = IF_OPER_DORMANT;
 205                 break;
 206         };
 207
 208         if (dev->operstate != operstate) {
 209                 write_lock_bh(&dev_base_lock);
 210                 dev->operstate = operstate;
 211                 write_unlock_bh(&dev_base_lock);
 212                 netdev_state_change(dev);
 213         }
 214 }
 215
 216 static int rtnetlink_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
 217                                  int type, u32 pid, u32 seq, u32 change,
 218                                  unsigned int flags)
 219 {
 220         struct ifinfomsg *r;
 221         struct nlmsghdr  *nlh;
 222         unsigned char    *b = skb->tail;
 223
 224         nlh = NLMSG_NEW(skb, pid, seq, type, sizeof(*r), flags);
 225         r = NLMSG_DATA(nlh);
 226         r->ifi_family = AF_UNSPEC;
 227         r->__ifi_pad = 0;
 228         r->ifi_type = dev->type;
 229         r->ifi_index = dev->ifindex;
 230         r->ifi_flags = dev_get_flags(dev);
 231         r->ifi_change = change;
 232
 233         RTA_PUT(skb, IFLA_IFNAME, strlen(dev->name)+1, dev->name);
 234
 235         if (1) {
 236                 u32 txqlen = dev->tx_queue_len;
 237                 RTA_PUT(skb, IFLA_TXQLEN, sizeof(txqlen), &txqlen);
 238         }
 239
 240         if (1) {
 241                 u32 weight = dev->weight;
 242                 RTA_PUT(skb, IFLA_WEIGHT, sizeof(weight), &weight);
 243         }
 244
 245         if (1) {
 246                 u8 operstate = netif_running(dev)?dev->operstate:IF_OPER_DOWN;
 247                 u8 link_mode = dev->link_mode;
 248                 RTA_PUT(skb, IFLA_OPERSTATE, sizeof(operstate), &operstate);
 249                 RTA_PUT(skb, IFLA_LINKMODE, sizeof(link_mode), &link_mode);
 250         }
 251
 252         if (1) {
 253                 struct rtnl_link_ifmap map = {
 254                         .mem_start   = dev->mem_start,
 255                         .mem_end     = dev->mem_end,
 256                         .base_addr   = dev->base_addr,
 257                         .irq         = dev->irq,
 258                         .dma         = dev->dma,
 259                         .port        = dev->if_port,
 260                 };
 261                 RTA_PUT(skb, IFLA_MAP, sizeof(map), &map);
 262         }
 263
 264         if (dev->addr_len) {
 265                 RTA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
 266                 RTA_PUT(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast);
 267         }
 268
 269         if (1) {
 270                 u32 mtu = dev->mtu;
 271                 RTA_PUT(skb, IFLA_MTU, sizeof(mtu), &mtu);
 272         }
 273
 274         if (dev->ifindex != dev->iflink) {
 275                 u32 iflink = dev->iflink;
 276                 RTA_PUT(skb, IFLA_LINK, sizeof(iflink), &iflink);
 277         }
 278
 279         if (dev->qdisc_sleeping)
 280                 RTA_PUT(skb, IFLA_QDISC,
 281                         strlen(dev->qdisc_sleeping->ops->id) + 1,
 282                         dev->qdisc_sleeping->ops->id);
 283
 284         if (dev->master) {
 285                 u32 master = dev->master->ifindex;
 286                 RTA_PUT(skb, IFLA_MASTER, sizeof(master), &master);
 287         }
 288
 289         if (dev->get_stats) {
 290                 unsigned long *stats = (unsigned long*)dev->get_stats(dev);
 291                 if (stats) {
 292                         struct rtattr  *a;
 293                         __u32          *s;
 294                         int             i;
 295                         int             n = sizeof(struct rtnl_link_stats)/4;
 296
 297                         a = __RTA_PUT(skb, IFLA_STATS, n*4);
 298                         s = RTA_DATA(a);
 299                         for (i=0; i<n; i++)
 300                                 s[i] = stats[i];
 301                 }
 302         }
 303         nlh->nlmsg_len = skb->tail - b;
 304         return skb->len;
 305
 306 nlmsg_failure:
 307 rtattr_failure:
 308         skb_trim(skb, b - skb->data);
 309         return -1;
 310 }
 311
 312 static int rtnetlink_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
 313 {
 314         int idx;
 315         int s_idx = cb->args[0];
 316         struct net_device *dev;
 317
 318         read_lock(&dev_base_lock);
 319         for (dev=dev_base, idx=0; dev; dev = dev->next, idx++) {
 320                 if (idx < s_idx)
 321                         continue;
 322                 if (rtnetlink_fill_ifinfo(skb, dev, RTM_NEWLINK,
 323                                           NETLINK_CB(cb->skb).pid,
 324                                           cb->nlh->nlmsg_seq, 0,
 325                                           NLM_F_MULTI) <= 0)
 326                         break;
 327         }
 328         read_unlock(&dev_base_lock);
 329         cb->args[0] = idx;
 330
 331         return skb->len;
 332 }
 333
 334 static int do_setlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
 335 {
 336         struct ifinfomsg  *ifm = NLMSG_DATA(nlh);
 337         struct rtattr    **ida = arg;
 338         struct net_device *dev;
 339         int err, send_addr_notify = 0;
 340
 341         if (ifm->ifi_index >= 0)
 342                 dev = dev_get_by_index(ifm->ifi_index);
 343         else if (ida[IFLA_IFNAME - 1]) {
 344                 char ifname[IFNAMSIZ];
 345
 346                 if (rtattr_strlcpy(ifname, ida[IFLA_IFNAME - 1],
 347                                    IFNAMSIZ) >= IFNAMSIZ)
 348                         return -EINVAL;
 349                 dev = dev_get_by_name(ifname);
 350         } else
 351                 return -EINVAL;
 352
 353         if (!dev)
 354                 return -ENODEV;
 355
 356         err = -EINVAL;
 357
 358         if (ifm->ifi_flags)
 359                 dev_change_flags(dev, ifm->ifi_flags);
 360
 361         if (ida[IFLA_MAP - 1]) {
 362                 struct rtnl_link_ifmap *u_map;
 363                 struct ifmap k_map;
 364
 365                 if (!dev->set_config) {
 366                         err = -EOPNOTSUPP;
 367                         goto out;
 368                 }
 369
 370                 if (!netif_device_present(dev)) {
 371                         err = -ENODEV;
 372                         goto out;
 373                 }
 374
 375                 if (ida[IFLA_MAP - 1]->rta_len != RTA_LENGTH(sizeof(*u_map)))
 376                         goto out;
 377
 378                 u_map = RTA_DATA(ida[IFLA_MAP - 1]);
 379
 380                 k_map.mem_start = (unsigned long) u_map->mem_start;
 381                 k_map.mem_end = (unsigned long) u_map->mem_end;
 382                 k_map.base_addr = (unsigned short) u_map->base_addr;
 383                 k_map.irq = (unsigned char) u_map->irq;
 384                 k_map.dma = (unsigned char) u_map->dma;
 385                 k_map.port = (unsigned char) u_map->port;
 386
 387                 err = dev->set_config(dev, &k_map);
 388
 389                 if (err)
 390                         goto out;
 391         }
 392
 393         if (ida[IFLA_ADDRESS - 1]) {
 394                 if (!dev->set_mac_address) {
 395                         err = -EOPNOTSUPP;
 396                         goto out;
 397                 }
 398                 if (!netif_device_present(dev)) {
 399                         err = -ENODEV;
 400                         goto out;
 401                 }
 402                 if (ida[IFLA_ADDRESS - 1]->rta_len != RTA_LENGTH(dev->addr_len))
 403                         goto out;
 404
 405                 err = dev->set_mac_address(dev, RTA_DATA(ida[IFLA_ADDRESS - 1]));
 406                 if (err)
 407                         goto out;
 408                 send_addr_notify = 1;
 409         }
 410
 411         if (ida[IFLA_BROADCAST - 1]) {
 412                 if (ida[IFLA_BROADCAST - 1]->rta_len != RTA_LENGTH(dev->addr_len))
 413                         goto out;
 414                 memcpy(dev->broadcast, RTA_DATA(ida[IFLA_BROADCAST - 1]),
 415                        dev->addr_len);
 416                 send_addr_notify = 1;
 417         }
 418
 419         if (ida[IFLA_MTU - 1]) {
 420                 if (ida[IFLA_MTU - 1]->rta_len != RTA_LENGTH(sizeof(u32)))
 421                         goto out;
 422                 err = dev_set_mtu(dev, *((u32 *) RTA_DATA(ida[IFLA_MTU - 1])));
 423
 424                 if (err)
 425                         goto out;
 426
 427         }
 428
 429         if (ida[IFLA_TXQLEN - 1]) {
 430                 if (ida[IFLA_TXQLEN - 1]->rta_len != RTA_LENGTH(sizeof(u32)))
 431                         goto out;
 432
 433                 dev->tx_queue_len = *((u32 *) RTA_DATA(ida[IFLA_TXQLEN - 1]));
 434         }
 435
 436         if (ida[IFLA_WEIGHT - 1]) {
 437                 if (ida[IFLA_WEIGHT - 1]->rta_len != RTA_LENGTH(sizeof(u32)))
 438                         goto out;
 439
 440                 dev->weight = *((u32 *) RTA_DATA(ida[IFLA_WEIGHT - 1]));
 441         }
 442
 443         if (ida[IFLA_OPERSTATE - 1]) {
 444                 if (ida[IFLA_OPERSTATE - 1]->rta_len != RTA_LENGTH(sizeof(u8)))
 445                         goto out;
 446
 447                 set_operstate(dev, *((u8 *) RTA_DATA(ida[IFLA_OPERSTATE - 1])));
 448         }
 449
 450         if (ida[IFLA_LINKMODE - 1]) {
 451                 if (ida[IFLA_LINKMODE - 1]->rta_len != RTA_LENGTH(sizeof(u8)))
 452                         goto out;
 453
 454                 write_lock_bh(&dev_base_lock);
 455                 dev->link_mode = *((u8 *) RTA_DATA(ida[IFLA_LINKMODE - 1]));
 456                 write_unlock_bh(&dev_base_lock);
 457         }
 458
 459         if (ifm->ifi_index >= 0 && ida[IFLA_IFNAME - 1]) {
 460                 char ifname[IFNAMSIZ];
 461
 462                 if (rtattr_strlcpy(ifname, ida[IFLA_IFNAME - 1],
 463                                    IFNAMSIZ) >= IFNAMSIZ)
 464                         goto out;
 465                 err = dev_change_name(dev, ifname);
 466                 if (err)
 467                         goto out;
 468         }
 469
 470         err = 0;
 471
 472 out:
 473         if (send_addr_notify)
 474                 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
 475
 476         dev_put(dev);
 477         return err;
 478 }
 479
 480 static int rtnetlink_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
 481 {
 482         int idx;
 483         int s_idx = cb->family;
 484
 485         if (s_idx == 0)
 486                 s_idx = 1;
 487         for (idx=1; idx<NPROTO; idx++) {
 488                 int type = cb->nlh->nlmsg_type-RTM_BASE;
 489                 if (idx < s_idx || idx == PF_PACKET)
 490                         continue;
 491                 if (rtnetlink_links[idx] == NULL ||
 492                     rtnetlink_links[idx][type].dumpit == NULL)
 493                         continue;
 494                 if (idx > s_idx)
 495                         memset(&cb->args[0], 0, sizeof(cb->args));
 496                 if (rtnetlink_links[idx][type].dumpit(skb, cb))
 497                         break;
 498         }
 499         cb->family = idx;
 500
 501         return skb->len;
 502 }
 503
 504 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned change)
 505 {
 506         struct sk_buff *skb;
 507         int size = NLMSG_SPACE(sizeof(struct ifinfomsg) +
 508                                sizeof(struct rtnl_link_ifmap) +
 509                                sizeof(struct rtnl_link_stats) + 128);
 510
 511         skb = alloc_skb(size, GFP_KERNEL);
 512         if (!skb)
 513                 return;
 514
 515         if (rtnetlink_fill_ifinfo(skb, dev, type, 0, 0, change, 0) < 0) {
 516                 kfree_skb(skb);
 517                 return;
 518         }
 519         NETLINK_CB(skb).dst_group = RTNLGRP_LINK;
 520         netlink_broadcast(rtnl, skb, 0, RTNLGRP_LINK, GFP_KERNEL);
 521 }
 522
 523 /* Protected by RTNL sempahore.  */
 524 static struct rtattr **rta_buf;
 525 static int rtattr_max;
 526
 527 /* Process one rtnetlink message. */
 528
 529 static __inline__ int
 530 rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh, int *errp)
 531 {
 532         struct rtnetlink_link *link;
 533         struct rtnetlink_link *link_tab;
 534         int sz_idx, kind;
 535         int min_len;
 536         int family;
 537         int type;
 538         int err;
 539
 540         /* Only requests are handled by kernel now */
 541         if (!(nlh->nlmsg_flags&NLM_F_REQUEST))
 542                 return 0;
 543
 544         type = nlh->nlmsg_type;
 545
 546         /* A control message: ignore them */
 547         if (type < RTM_BASE)
 548                 return 0;
 549
 550         /* Unknown message: reply with EINVAL */
 551         if (type > RTM_MAX)
 552                 goto err_inval;
 553
 554         type -= RTM_BASE;
 555
 556         /* All the messages must have at least 1 byte length */
 557         if (nlh->nlmsg_len < NLMSG_LENGTH(sizeof(struct rtgenmsg)))
 558                 return 0;
 559
 560         family = ((struct rtgenmsg*)NLMSG_DATA(nlh))->rtgen_family;
 561         if (family >= NPROTO) {
 562                 *errp = -EAFNOSUPPORT;
 563                 return -1;
 564         }
 565
 566         link_tab = rtnetlink_links[family];
 567         if (link_tab == NULL)
 568                 link_tab = rtnetlink_links[PF_UNSPEC];
 569         link = &link_tab[type];
 570
 571         sz_idx = type>>2;
 572         kind = type&3;
 573
 574         if (kind != 2 && security_netlink_recv(skb)) {
 575                 *errp = -EPERM;
 576                 return -1;
 577         }
 578
 579         if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
 580                 if (link->dumpit == NULL)
 581                         link = &(rtnetlink_links[PF_UNSPEC][type]);
 582
 583                 if (link->dumpit == NULL)
 584                         goto err_inval;
 585
 586                 if ((*errp = netlink_dump_start(rtnl, skb, nlh,
 587                                                 link->dumpit, NULL)) != 0) {
 588                         return -1;
 589                 }
 590
 591                 netlink_queue_skip(nlh, skb);
 592                 return -1;
 593         }
 594
 595         memset(rta_buf, 0, (rtattr_max * sizeof(struct rtattr *)));
 596
 597         min_len = rtm_min[sz_idx];
 598         if (nlh->nlmsg_len < min_len)
 599                 goto err_inval;
 600
 601         if (nlh->nlmsg_len > min_len) {
 602                 int attrlen = nlh->nlmsg_len - NLMSG_ALIGN(min_len);
 603                 struct rtattr *attr = (void*)nlh + NLMSG_ALIGN(min_len);
 604
 605                 while (RTA_OK(attr, attrlen)) {
 606                         unsigned flavor = attr->rta_type;
 607                         if (flavor) {
 608                                 if (flavor > rta_max[sz_idx])
 609                                         goto err_inval;
 610                                 rta_buf[flavor-1] = attr;
 611                         }
 612                         attr = RTA_NEXT(attr, attrlen);
 613                 }
 614         }
 615
 616         if (link->doit == NULL)
 617                 link = &(rtnetlink_links[PF_UNSPEC][type]);
 618         if (link->doit == NULL)
 619                 goto err_inval;
 620         err = link->doit(skb, nlh, (void *)&rta_buf[0]);
 621
 622         *errp = err;
 623         return err;
 624
 625 err_inval:
 626         *errp = -EINVAL;
 627         return -1;
 628 }
 629
 630 static void rtnetlink_rcv(struct sock *sk, int len)
 631 {
 632         unsigned int qlen = 0;
 633
 634         do {
 635                 mutex_lock(&rtnl_mutex);
 636                 netlink_run_queue(sk, &qlen, &rtnetlink_rcv_msg);
 637                 mutex_unlock(&rtnl_mutex);
 638
 639                 netdev_run_todo();
 640         } while (qlen);
 641 }
 642
 643 static struct rtnetlink_link link_rtnetlink_table[RTM_NR_MSGTYPES] =
 644 {
 645         [RTM_GETLINK     - RTM_BASE] = { .dumpit = rtnetlink_dump_ifinfo },
 646         [RTM_SETLINK     - RTM_BASE] = { .doit   = do_setlink            },
 647         [RTM_GETADDR     - RTM_BASE] = { .dumpit = rtnetlink_dump_all    },
 648         [RTM_GETROUTE    - RTM_BASE] = { .dumpit = rtnetlink_dump_all    },
 649         [RTM_NEWNEIGH    - RTM_BASE] = { .doit   = neigh_add             },
 650         [RTM_DELNEIGH    - RTM_BASE] = { .doit   = neigh_delete          },
 651         [RTM_GETNEIGH    - RTM_BASE] = { .dumpit = neigh_dump_info       },
 652         [RTM_GETRULE     - RTM_BASE] = { .dumpit = rtnetlink_dump_all    },
 653         [RTM_GETNEIGHTBL - RTM_BASE] = { .dumpit = neightbl_dump_info    },
 654         [RTM_SETNEIGHTBL - RTM_BASE] = { .doit   = neightbl_set          },
 655 };
 656
 657 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
 658 {
 659         struct net_device *dev = ptr;
 660         switch (event) {
 661         case NETDEV_UNREGISTER:
 662                 rtmsg_ifinfo(RTM_DELLINK, dev, ~0U);
 663                 break;
 664         case NETDEV_REGISTER:
 665                 rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U);
 666                 break;
 667         case NETDEV_UP:
 668         case NETDEV_DOWN:
 669                 rtmsg_ifinfo(RTM_NEWLINK, dev, IFF_UP|IFF_RUNNING);
 670                 break;
 671         case NETDEV_CHANGE:
 672         case NETDEV_GOING_DOWN:
 673                 break;
 674         default:
 675                 rtmsg_ifinfo(RTM_NEWLINK, dev, 0);
 676                 break;
 677         }
 678         return NOTIFY_DONE;
 679 }
 680
 681 static struct notifier_block rtnetlink_dev_notifier = {
 682         .notifier_call  = rtnetlink_event,
 683 };
 684
 685 void __init rtnetlink_init(void)
 686 {
 687         int i;
 688
 689         rtattr_max = 0;
 690         for (i = 0; i < ARRAY_SIZE(rta_max); i++)
 691                 if (rta_max[i] > rtattr_max)
 692                         rtattr_max = rta_max[i];
 693         rta_buf = kmalloc(rtattr_max * sizeof(struct rtattr *), GFP_KERNEL);
 694         if (!rta_buf)
 695                 panic("rtnetlink_init: cannot allocate rta_buf\n");
 696
 697         rtnl = netlink_kernel_create(NETLINK_ROUTE, RTNLGRP_MAX, rtnetlink_rcv,
 698                                      THIS_MODULE);
 699         if (rtnl == NULL)
 700                 panic("rtnetlink_init: cannot initialize rtnetlink\n");
 701         netlink_set_nonroot(NETLINK_ROUTE, NL_NONROOT_RECV);
 702         register_netdevice_notifier(&rtnetlink_dev_notifier);
 703         rtnetlink_links[PF_UNSPEC] = link_rtnetlink_table;
 704         rtnetlink_links[PF_PACKET] = link_rtnetlink_table;
 705 }
 706
 707 EXPORT_SYMBOL(__rta_fill);
 708 EXPORT_SYMBOL(rtattr_strlcpy);
 709 EXPORT_SYMBOL(rtattr_parse);
 710 EXPORT_SYMBOL(rtnetlink_links);
 711 EXPORT_SYMBOL(rtnetlink_put_metrics);
 712 EXPORT_SYMBOL(rtnl);
 713 EXPORT_SYMBOL(rtnl_lock);
 714 EXPORT_SYMBOL(rtnl_trylock);
 715 EXPORT_SYMBOL(rtnl_unlock);