linux/net/core/rtnetlink.c
<<
>>
Prefs
   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/errno.h>
  20#include <linux/module.h>
  21#include <linux/types.h>
  22#include <linux/socket.h>
  23#include <linux/kernel.h>
  24#include <linux/timer.h>
  25#include <linux/string.h>
  26#include <linux/sockios.h>
  27#include <linux/net.h>
  28#include <linux/fcntl.h>
  29#include <linux/mm.h>
  30#include <linux/slab.h>
  31#include <linux/interrupt.h>
  32#include <linux/capability.h>
  33#include <linux/skbuff.h>
  34#include <linux/init.h>
  35#include <linux/security.h>
  36#include <linux/mutex.h>
  37#include <linux/if_addr.h>
  38#include <linux/nsproxy.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/fib_rules.h>
  54#include <net/rtnetlink.h>
  55
  56struct rtnl_link
  57{
  58        rtnl_doit_func          doit;
  59        rtnl_dumpit_func        dumpit;
  60};
  61
  62static DEFINE_MUTEX(rtnl_mutex);
  63
  64void rtnl_lock(void)
  65{
  66        mutex_lock(&rtnl_mutex);
  67}
  68
  69void __rtnl_unlock(void)
  70{
  71        mutex_unlock(&rtnl_mutex);
  72}
  73
  74void rtnl_unlock(void)
  75{
  76        /* This fellow will unlock it for us. */
  77        netdev_run_todo();
  78}
  79
  80int rtnl_trylock(void)
  81{
  82        return mutex_trylock(&rtnl_mutex);
  83}
  84
  85int rtnl_is_locked(void)
  86{
  87        return mutex_is_locked(&rtnl_mutex);
  88}
  89
  90static struct rtnl_link *rtnl_msg_handlers[NPROTO];
  91
  92static inline int rtm_msgindex(int msgtype)
  93{
  94        int msgindex = msgtype - RTM_BASE;
  95
  96        /*
  97         * msgindex < 0 implies someone tried to register a netlink
  98         * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
  99         * the message type has not been added to linux/rtnetlink.h
 100         */
 101        BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
 102
 103        return msgindex;
 104}
 105
 106static rtnl_doit_func rtnl_get_doit(int protocol, int msgindex)
 107{
 108        struct rtnl_link *tab;
 109
 110        tab = rtnl_msg_handlers[protocol];
 111        if (tab == NULL || tab[msgindex].doit == NULL)
 112                tab = rtnl_msg_handlers[PF_UNSPEC];
 113
 114        return tab ? tab[msgindex].doit : NULL;
 115}
 116
 117static rtnl_dumpit_func rtnl_get_dumpit(int protocol, int msgindex)
 118{
 119        struct rtnl_link *tab;
 120
 121        tab = rtnl_msg_handlers[protocol];
 122        if (tab == NULL || tab[msgindex].dumpit == NULL)
 123                tab = rtnl_msg_handlers[PF_UNSPEC];
 124
 125        return tab ? tab[msgindex].dumpit : NULL;
 126}
 127
 128/**
 129 * __rtnl_register - Register a rtnetlink message type
 130 * @protocol: Protocol family or PF_UNSPEC
 131 * @msgtype: rtnetlink message type
 132 * @doit: Function pointer called for each request message
 133 * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
 134 *
 135 * Registers the specified function pointers (at least one of them has
 136 * to be non-NULL) to be called whenever a request message for the
 137 * specified protocol family and message type is received.
 138 *
 139 * The special protocol family PF_UNSPEC may be used to define fallback
 140 * function pointers for the case when no entry for the specific protocol
 141 * family exists.
 142 *
 143 * Returns 0 on success or a negative error code.
 144 */
 145int __rtnl_register(int protocol, int msgtype,
 146                    rtnl_doit_func doit, rtnl_dumpit_func dumpit)
 147{
 148        struct rtnl_link *tab;
 149        int msgindex;
 150
 151        BUG_ON(protocol < 0 || protocol >= NPROTO);
 152        msgindex = rtm_msgindex(msgtype);
 153
 154        tab = rtnl_msg_handlers[protocol];
 155        if (tab == NULL) {
 156                tab = kcalloc(RTM_NR_MSGTYPES, sizeof(*tab), GFP_KERNEL);
 157                if (tab == NULL)
 158                        return -ENOBUFS;
 159
 160                rtnl_msg_handlers[protocol] = tab;
 161        }
 162
 163        if (doit)
 164                tab[msgindex].doit = doit;
 165
 166        if (dumpit)
 167                tab[msgindex].dumpit = dumpit;
 168
 169        return 0;
 170}
 171
 172EXPORT_SYMBOL_GPL(__rtnl_register);
 173
 174/**
 175 * rtnl_register - Register a rtnetlink message type
 176 *
 177 * Identical to __rtnl_register() but panics on failure. This is useful
 178 * as failure of this function is very unlikely, it can only happen due
 179 * to lack of memory when allocating the chain to store all message
 180 * handlers for a protocol. Meant for use in init functions where lack
 181 * of memory implies no sense in continueing.
 182 */
 183void rtnl_register(int protocol, int msgtype,
 184                   rtnl_doit_func doit, rtnl_dumpit_func dumpit)
 185{
 186        if (__rtnl_register(protocol, msgtype, doit, dumpit) < 0)
 187                panic("Unable to register rtnetlink message handler, "
 188                      "protocol = %d, message type = %d\n",
 189                      protocol, msgtype);
 190}
 191
 192EXPORT_SYMBOL_GPL(rtnl_register);
 193
 194/**
 195 * rtnl_unregister - Unregister a rtnetlink message type
 196 * @protocol: Protocol family or PF_UNSPEC
 197 * @msgtype: rtnetlink message type
 198 *
 199 * Returns 0 on success or a negative error code.
 200 */
 201int rtnl_unregister(int protocol, int msgtype)
 202{
 203        int msgindex;
 204
 205        BUG_ON(protocol < 0 || protocol >= NPROTO);
 206        msgindex = rtm_msgindex(msgtype);
 207
 208        if (rtnl_msg_handlers[protocol] == NULL)
 209                return -ENOENT;
 210
 211        rtnl_msg_handlers[protocol][msgindex].doit = NULL;
 212        rtnl_msg_handlers[protocol][msgindex].dumpit = NULL;
 213
 214        return 0;
 215}
 216
 217EXPORT_SYMBOL_GPL(rtnl_unregister);
 218
 219/**
 220 * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
 221 * @protocol : Protocol family or PF_UNSPEC
 222 *
 223 * Identical to calling rtnl_unregster() for all registered message types
 224 * of a certain protocol family.
 225 */
 226void rtnl_unregister_all(int protocol)
 227{
 228        BUG_ON(protocol < 0 || protocol >= NPROTO);
 229
 230        kfree(rtnl_msg_handlers[protocol]);
 231        rtnl_msg_handlers[protocol] = NULL;
 232}
 233
 234EXPORT_SYMBOL_GPL(rtnl_unregister_all);
 235
 236static LIST_HEAD(link_ops);
 237
 238/**
 239 * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 240 * @ops: struct rtnl_link_ops * to register
 241 *
 242 * The caller must hold the rtnl_mutex. This function should be used
 243 * by drivers that create devices during module initialization. It
 244 * must be called before registering the devices.
 245 *
 246 * Returns 0 on success or a negative error code.
 247 */
 248int __rtnl_link_register(struct rtnl_link_ops *ops)
 249{
 250        if (!ops->dellink)
 251                ops->dellink = unregister_netdevice;
 252
 253        list_add_tail(&ops->list, &link_ops);
 254        return 0;
 255}
 256
 257EXPORT_SYMBOL_GPL(__rtnl_link_register);
 258
 259/**
 260 * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 261 * @ops: struct rtnl_link_ops * to register
 262 *
 263 * Returns 0 on success or a negative error code.
 264 */
 265int rtnl_link_register(struct rtnl_link_ops *ops)
 266{
 267        int err;
 268
 269        rtnl_lock();
 270        err = __rtnl_link_register(ops);
 271        rtnl_unlock();
 272        return err;
 273}
 274
 275EXPORT_SYMBOL_GPL(rtnl_link_register);
 276
 277static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
 278{
 279        struct net_device *dev;
 280restart:
 281        for_each_netdev(net, dev) {
 282                if (dev->rtnl_link_ops == ops) {
 283                        ops->dellink(dev);
 284                        goto restart;
 285                }
 286        }
 287}
 288
 289void rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
 290{
 291        rtnl_lock();
 292        __rtnl_kill_links(net, ops);
 293        rtnl_unlock();
 294}
 295EXPORT_SYMBOL_GPL(rtnl_kill_links);
 296
 297/**
 298 * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 299 * @ops: struct rtnl_link_ops * to unregister
 300 *
 301 * The caller must hold the rtnl_mutex.
 302 */
 303void __rtnl_link_unregister(struct rtnl_link_ops *ops)
 304{
 305        struct net *net;
 306
 307        for_each_net(net) {
 308                __rtnl_kill_links(net, ops);
 309        }
 310        list_del(&ops->list);
 311}
 312
 313EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
 314
 315/**
 316 * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 317 * @ops: struct rtnl_link_ops * to unregister
 318 */
 319void rtnl_link_unregister(struct rtnl_link_ops *ops)
 320{
 321        rtnl_lock();
 322        __rtnl_link_unregister(ops);
 323        rtnl_unlock();
 324}
 325
 326EXPORT_SYMBOL_GPL(rtnl_link_unregister);
 327
 328static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
 329{
 330        const struct rtnl_link_ops *ops;
 331
 332        list_for_each_entry(ops, &link_ops, list) {
 333                if (!strcmp(ops->kind, kind))
 334                        return ops;
 335        }
 336        return NULL;
 337}
 338
 339static size_t rtnl_link_get_size(const struct net_device *dev)
 340{
 341        const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 342        size_t size;
 343
 344        if (!ops)
 345                return 0;
 346
 347        size = nlmsg_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
 348               nlmsg_total_size(strlen(ops->kind) + 1);  /* IFLA_INFO_KIND */
 349
 350        if (ops->get_size)
 351                /* IFLA_INFO_DATA + nested data */
 352                size += nlmsg_total_size(sizeof(struct nlattr)) +
 353                        ops->get_size(dev);
 354
 355        if (ops->get_xstats_size)
 356                size += ops->get_xstats_size(dev);      /* IFLA_INFO_XSTATS */
 357
 358        return size;
 359}
 360
 361static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
 362{
 363        const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 364        struct nlattr *linkinfo, *data;
 365        int err = -EMSGSIZE;
 366
 367        linkinfo = nla_nest_start(skb, IFLA_LINKINFO);
 368        if (linkinfo == NULL)
 369                goto out;
 370
 371        if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
 372                goto err_cancel_link;
 373        if (ops->fill_xstats) {
 374                err = ops->fill_xstats(skb, dev);
 375                if (err < 0)
 376                        goto err_cancel_link;
 377        }
 378        if (ops->fill_info) {
 379                data = nla_nest_start(skb, IFLA_INFO_DATA);
 380                if (data == NULL)
 381                        goto err_cancel_link;
 382                err = ops->fill_info(skb, dev);
 383                if (err < 0)
 384                        goto err_cancel_data;
 385                nla_nest_end(skb, data);
 386        }
 387
 388        nla_nest_end(skb, linkinfo);
 389        return 0;
 390
 391err_cancel_data:
 392        nla_nest_cancel(skb, data);
 393err_cancel_link:
 394        nla_nest_cancel(skb, linkinfo);
 395out:
 396        return err;
 397}
 398
 399static const int rtm_min[RTM_NR_FAMILIES] =
 400{
 401        [RTM_FAM(RTM_NEWLINK)]      = NLMSG_LENGTH(sizeof(struct ifinfomsg)),
 402        [RTM_FAM(RTM_NEWADDR)]      = NLMSG_LENGTH(sizeof(struct ifaddrmsg)),
 403        [RTM_FAM(RTM_NEWROUTE)]     = NLMSG_LENGTH(sizeof(struct rtmsg)),
 404        [RTM_FAM(RTM_NEWRULE)]      = NLMSG_LENGTH(sizeof(struct fib_rule_hdr)),
 405        [RTM_FAM(RTM_NEWQDISC)]     = NLMSG_LENGTH(sizeof(struct tcmsg)),
 406        [RTM_FAM(RTM_NEWTCLASS)]    = NLMSG_LENGTH(sizeof(struct tcmsg)),
 407        [RTM_FAM(RTM_NEWTFILTER)]   = NLMSG_LENGTH(sizeof(struct tcmsg)),
 408        [RTM_FAM(RTM_NEWACTION)]    = NLMSG_LENGTH(sizeof(struct tcamsg)),
 409        [RTM_FAM(RTM_GETMULTICAST)] = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
 410        [RTM_FAM(RTM_GETANYCAST)]   = NLMSG_LENGTH(sizeof(struct rtgenmsg)),
 411};
 412
 413static const int rta_max[RTM_NR_FAMILIES] =
 414{
 415        [RTM_FAM(RTM_NEWLINK)]      = IFLA_MAX,
 416        [RTM_FAM(RTM_NEWADDR)]      = IFA_MAX,
 417        [RTM_FAM(RTM_NEWROUTE)]     = RTA_MAX,
 418        [RTM_FAM(RTM_NEWRULE)]      = FRA_MAX,
 419        [RTM_FAM(RTM_NEWQDISC)]     = TCA_MAX,
 420        [RTM_FAM(RTM_NEWTCLASS)]    = TCA_MAX,
 421        [RTM_FAM(RTM_NEWTFILTER)]   = TCA_MAX,
 422        [RTM_FAM(RTM_NEWACTION)]    = TCAA_MAX,
 423};
 424
 425void __rta_fill(struct sk_buff *skb, int attrtype, int attrlen, const void *data)
 426{
 427        struct rtattr *rta;
 428        int size = RTA_LENGTH(attrlen);
 429
 430        rta = (struct rtattr*)skb_put(skb, RTA_ALIGN(size));
 431        rta->rta_type = attrtype;
 432        rta->rta_len = size;
 433        memcpy(RTA_DATA(rta), data, attrlen);
 434        memset(RTA_DATA(rta) + attrlen, 0, RTA_ALIGN(size) - size);
 435}
 436
 437int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned group, int echo)
 438{
 439        struct sock *rtnl = net->rtnl;
 440        int err = 0;
 441
 442        NETLINK_CB(skb).dst_group = group;
 443        if (echo)
 444                atomic_inc(&skb->users);
 445        netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
 446        if (echo)
 447                err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
 448        return err;
 449}
 450
 451int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
 452{
 453        struct sock *rtnl = net->rtnl;
 454
 455        return nlmsg_unicast(rtnl, skb, pid);
 456}
 457
 458void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
 459                 struct nlmsghdr *nlh, gfp_t flags)
 460{
 461        struct sock *rtnl = net->rtnl;
 462        int report = 0;
 463
 464        if (nlh)
 465                report = nlmsg_report(nlh);
 466
 467        nlmsg_notify(rtnl, skb, pid, group, report, flags);
 468}
 469
 470void rtnl_set_sk_err(struct net *net, u32 group, int error)
 471{
 472        struct sock *rtnl = net->rtnl;
 473
 474        netlink_set_err(rtnl, 0, group, error);
 475}
 476
 477int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
 478{
 479        struct nlattr *mx;
 480        int i, valid = 0;
 481
 482        mx = nla_nest_start(skb, RTA_METRICS);
 483        if (mx == NULL)
 484                return -ENOBUFS;
 485
 486        for (i = 0; i < RTAX_MAX; i++) {
 487                if (metrics[i]) {
 488                        valid++;
 489                        NLA_PUT_U32(skb, i+1, metrics[i]);
 490                }
 491        }
 492
 493        if (!valid) {
 494                nla_nest_cancel(skb, mx);
 495                return 0;
 496        }
 497
 498        return nla_nest_end(skb, mx);
 499
 500nla_put_failure:
 501        nla_nest_cancel(skb, mx);
 502        return -EMSGSIZE;
 503}
 504
 505int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
 506                       u32 ts, u32 tsage, long expires, u32 error)
 507{
 508        struct rta_cacheinfo ci = {
 509                .rta_lastuse = jiffies_to_clock_t(jiffies - dst->lastuse),
 510                .rta_used = dst->__use,
 511                .rta_clntref = atomic_read(&(dst->__refcnt)),
 512                .rta_error = error,
 513                .rta_id =  id,
 514                .rta_ts = ts,
 515                .rta_tsage = tsage,
 516        };
 517
 518        if (expires)
 519                ci.rta_expires = jiffies_to_clock_t(expires);
 520
 521        return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
 522}
 523
 524EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
 525
 526static void set_operstate(struct net_device *dev, unsigned char transition)
 527{
 528        unsigned char operstate = dev->operstate;
 529
 530        switch(transition) {
 531        case IF_OPER_UP:
 532                if ((operstate == IF_OPER_DORMANT ||
 533                     operstate == IF_OPER_UNKNOWN) &&
 534                    !netif_dormant(dev))
 535                        operstate = IF_OPER_UP;
 536                break;
 537
 538        case IF_OPER_DORMANT:
 539                if (operstate == IF_OPER_UP ||
 540                    operstate == IF_OPER_UNKNOWN)
 541                        operstate = IF_OPER_DORMANT;
 542                break;
 543        }
 544
 545        if (dev->operstate != operstate) {
 546                write_lock_bh(&dev_base_lock);
 547                dev->operstate = operstate;
 548                write_unlock_bh(&dev_base_lock);
 549                netdev_state_change(dev);
 550        }
 551}
 552
 553static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
 554                                 const struct net_device_stats *b)
 555{
 556        a->rx_packets = b->rx_packets;
 557        a->tx_packets = b->tx_packets;
 558        a->rx_bytes = b->rx_bytes;
 559        a->tx_bytes = b->tx_bytes;
 560        a->rx_errors = b->rx_errors;
 561        a->tx_errors = b->tx_errors;
 562        a->rx_dropped = b->rx_dropped;
 563        a->tx_dropped = b->tx_dropped;
 564
 565        a->multicast = b->multicast;
 566        a->collisions = b->collisions;
 567
 568        a->rx_length_errors = b->rx_length_errors;
 569        a->rx_over_errors = b->rx_over_errors;
 570        a->rx_crc_errors = b->rx_crc_errors;
 571        a->rx_frame_errors = b->rx_frame_errors;
 572        a->rx_fifo_errors = b->rx_fifo_errors;
 573        a->rx_missed_errors = b->rx_missed_errors;
 574
 575        a->tx_aborted_errors = b->tx_aborted_errors;
 576        a->tx_carrier_errors = b->tx_carrier_errors;
 577        a->tx_fifo_errors = b->tx_fifo_errors;
 578        a->tx_heartbeat_errors = b->tx_heartbeat_errors;
 579        a->tx_window_errors = b->tx_window_errors;
 580
 581        a->rx_compressed = b->rx_compressed;
 582        a->tx_compressed = b->tx_compressed;
 583};
 584
 585static inline size_t if_nlmsg_size(const struct net_device *dev)
 586{
 587        return NLMSG_ALIGN(sizeof(struct ifinfomsg))
 588               + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
 589               + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
 590               + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
 591               + nla_total_size(sizeof(struct rtnl_link_ifmap))
 592               + nla_total_size(sizeof(struct rtnl_link_stats))
 593               + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
 594               + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
 595               + nla_total_size(4) /* IFLA_TXQLEN */
 596               + nla_total_size(4) /* IFLA_WEIGHT */
 597               + nla_total_size(4) /* IFLA_MTU */
 598               + nla_total_size(4) /* IFLA_LINK */
 599               + nla_total_size(4) /* IFLA_MASTER */
 600               + nla_total_size(1) /* IFLA_OPERSTATE */
 601               + nla_total_size(1) /* IFLA_LINKMODE */
 602               + rtnl_link_get_size(dev); /* IFLA_LINKINFO */
 603}
 604
 605static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
 606                            int type, u32 pid, u32 seq, u32 change,
 607                            unsigned int flags)
 608{
 609        struct netdev_queue *txq;
 610        struct ifinfomsg *ifm;
 611        struct nlmsghdr *nlh;
 612        const struct net_device_stats *stats;
 613        struct nlattr *attr;
 614
 615        nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
 616        if (nlh == NULL)
 617                return -EMSGSIZE;
 618
 619        ifm = nlmsg_data(nlh);
 620        ifm->ifi_family = AF_UNSPEC;
 621        ifm->__ifi_pad = 0;
 622        ifm->ifi_type = dev->type;
 623        ifm->ifi_index = dev->ifindex;
 624        ifm->ifi_flags = dev_get_flags(dev);
 625        ifm->ifi_change = change;
 626
 627        NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
 628        NLA_PUT_U32(skb, IFLA_TXQLEN, dev->tx_queue_len);
 629        NLA_PUT_U8(skb, IFLA_OPERSTATE,
 630                   netif_running(dev) ? dev->operstate : IF_OPER_DOWN);
 631        NLA_PUT_U8(skb, IFLA_LINKMODE, dev->link_mode);
 632        NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
 633
 634        if (dev->ifindex != dev->iflink)
 635                NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
 636
 637        if (dev->master)
 638                NLA_PUT_U32(skb, IFLA_MASTER, dev->master->ifindex);
 639
 640        txq = netdev_get_tx_queue(dev, 0);
 641        if (txq->qdisc_sleeping)
 642                NLA_PUT_STRING(skb, IFLA_QDISC, txq->qdisc_sleeping->ops->id);
 643
 644        if (dev->ifalias)
 645                NLA_PUT_STRING(skb, IFLA_IFALIAS, dev->ifalias);
 646
 647        if (1) {
 648                struct rtnl_link_ifmap map = {
 649                        .mem_start   = dev->mem_start,
 650                        .mem_end     = dev->mem_end,
 651                        .base_addr   = dev->base_addr,
 652                        .irq         = dev->irq,
 653                        .dma         = dev->dma,
 654                        .port        = dev->if_port,
 655                };
 656                NLA_PUT(skb, IFLA_MAP, sizeof(map), &map);
 657        }
 658
 659        if (dev->addr_len) {
 660                NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
 661                NLA_PUT(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast);
 662        }
 663
 664        attr = nla_reserve(skb, IFLA_STATS,
 665                        sizeof(struct rtnl_link_stats));
 666        if (attr == NULL)
 667                goto nla_put_failure;
 668
 669        stats = dev_get_stats(dev);
 670        copy_rtnl_link_stats(nla_data(attr), stats);
 671
 672        if (dev->rtnl_link_ops) {
 673                if (rtnl_link_fill(skb, dev) < 0)
 674                        goto nla_put_failure;
 675        }
 676
 677        return nlmsg_end(skb, nlh);
 678
 679nla_put_failure:
 680        nlmsg_cancel(skb, nlh);
 681        return -EMSGSIZE;
 682}
 683
 684static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
 685{
 686        struct net *net = sock_net(skb->sk);
 687        int idx;
 688        int s_idx = cb->args[0];
 689        struct net_device *dev;
 690
 691        idx = 0;
 692        for_each_netdev(net, dev) {
 693                if (idx < s_idx)
 694                        goto cont;
 695                if (rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
 696                                     NETLINK_CB(cb->skb).pid,
 697                                     cb->nlh->nlmsg_seq, 0, NLM_F_MULTI) <= 0)
 698                        break;
 699cont:
 700                idx++;
 701        }
 702        cb->args[0] = idx;
 703
 704        return skb->len;
 705}
 706
 707const struct nla_policy ifla_policy[IFLA_MAX+1] = {
 708        [IFLA_IFNAME]           = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
 709        [IFLA_ADDRESS]          = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
 710        [IFLA_BROADCAST]        = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
 711        [IFLA_MAP]              = { .len = sizeof(struct rtnl_link_ifmap) },
 712        [IFLA_MTU]              = { .type = NLA_U32 },
 713        [IFLA_LINK]             = { .type = NLA_U32 },
 714        [IFLA_TXQLEN]           = { .type = NLA_U32 },
 715        [IFLA_WEIGHT]           = { .type = NLA_U32 },
 716        [IFLA_OPERSTATE]        = { .type = NLA_U8 },
 717        [IFLA_LINKMODE]         = { .type = NLA_U8 },
 718        [IFLA_LINKINFO]         = { .type = NLA_NESTED },
 719        [IFLA_NET_NS_PID]       = { .type = NLA_U32 },
 720        [IFLA_IFALIAS]          = { .type = NLA_STRING, .len = IFALIASZ-1 },
 721};
 722
 723static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
 724        [IFLA_INFO_KIND]        = { .type = NLA_STRING },
 725        [IFLA_INFO_DATA]        = { .type = NLA_NESTED },
 726};
 727
 728static struct net *get_net_ns_by_pid(pid_t pid)
 729{
 730        struct task_struct *tsk;
 731        struct net *net;
 732
 733        /* Lookup the network namespace */
 734        net = ERR_PTR(-ESRCH);
 735        rcu_read_lock();
 736        tsk = find_task_by_vpid(pid);
 737        if (tsk) {
 738                struct nsproxy *nsproxy;
 739                nsproxy = task_nsproxy(tsk);
 740                if (nsproxy)
 741                        net = get_net(nsproxy->net_ns);
 742        }
 743        rcu_read_unlock();
 744        return net;
 745}
 746
 747static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
 748{
 749        if (dev) {
 750                if (tb[IFLA_ADDRESS] &&
 751                    nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
 752                        return -EINVAL;
 753
 754                if (tb[IFLA_BROADCAST] &&
 755                    nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
 756                        return -EINVAL;
 757        }
 758
 759        return 0;
 760}
 761
 762static int do_setlink(struct net_device *dev, struct ifinfomsg *ifm,
 763                      struct nlattr **tb, char *ifname, int modified)
 764{
 765        const struct net_device_ops *ops = dev->netdev_ops;
 766        int send_addr_notify = 0;
 767        int err;
 768
 769        if (tb[IFLA_NET_NS_PID]) {
 770                struct net *net;
 771                net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
 772                if (IS_ERR(net)) {
 773                        err = PTR_ERR(net);
 774                        goto errout;
 775                }
 776                err = dev_change_net_namespace(dev, net, ifname);
 777                put_net(net);
 778                if (err)
 779                        goto errout;
 780                modified = 1;
 781        }
 782
 783        if (tb[IFLA_MAP]) {
 784                struct rtnl_link_ifmap *u_map;
 785                struct ifmap k_map;
 786
 787                if (!ops->ndo_set_config) {
 788                        err = -EOPNOTSUPP;
 789                        goto errout;
 790                }
 791
 792                if (!netif_device_present(dev)) {
 793                        err = -ENODEV;
 794                        goto errout;
 795                }
 796
 797                u_map = nla_data(tb[IFLA_MAP]);
 798                k_map.mem_start = (unsigned long) u_map->mem_start;
 799                k_map.mem_end = (unsigned long) u_map->mem_end;
 800                k_map.base_addr = (unsigned short) u_map->base_addr;
 801                k_map.irq = (unsigned char) u_map->irq;
 802                k_map.dma = (unsigned char) u_map->dma;
 803                k_map.port = (unsigned char) u_map->port;
 804
 805                err = ops->ndo_set_config(dev, &k_map);
 806                if (err < 0)
 807                        goto errout;
 808
 809                modified = 1;
 810        }
 811
 812        if (tb[IFLA_ADDRESS]) {
 813                struct sockaddr *sa;
 814                int len;
 815
 816                if (!ops->ndo_set_mac_address) {
 817                        err = -EOPNOTSUPP;
 818                        goto errout;
 819                }
 820
 821                if (!netif_device_present(dev)) {
 822                        err = -ENODEV;
 823                        goto errout;
 824                }
 825
 826                len = sizeof(sa_family_t) + dev->addr_len;
 827                sa = kmalloc(len, GFP_KERNEL);
 828                if (!sa) {
 829                        err = -ENOMEM;
 830                        goto errout;
 831                }
 832                sa->sa_family = dev->type;
 833                memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
 834                       dev->addr_len);
 835                err = ops->ndo_set_mac_address(dev, sa);
 836                kfree(sa);
 837                if (err)
 838                        goto errout;
 839                send_addr_notify = 1;
 840                modified = 1;
 841        }
 842
 843        if (tb[IFLA_MTU]) {
 844                err = dev_set_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
 845                if (err < 0)
 846                        goto errout;
 847                modified = 1;
 848        }
 849
 850        /*
 851         * Interface selected by interface index but interface
 852         * name provided implies that a name change has been
 853         * requested.
 854         */
 855        if (ifm->ifi_index > 0 && ifname[0]) {
 856                err = dev_change_name(dev, ifname);
 857                if (err < 0)
 858                        goto errout;
 859                modified = 1;
 860        }
 861
 862        if (tb[IFLA_IFALIAS]) {
 863                err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
 864                                    nla_len(tb[IFLA_IFALIAS]));
 865                if (err < 0)
 866                        goto errout;
 867                modified = 1;
 868        }
 869
 870        if (tb[IFLA_BROADCAST]) {
 871                nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
 872                send_addr_notify = 1;
 873        }
 874
 875        if (ifm->ifi_flags || ifm->ifi_change) {
 876                unsigned int flags = ifm->ifi_flags;
 877
 878                /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
 879                if (ifm->ifi_change)
 880                        flags = (flags & ifm->ifi_change) |
 881                                (dev->flags & ~ifm->ifi_change);
 882                err = dev_change_flags(dev, flags);
 883                if (err < 0)
 884                        goto errout;
 885        }
 886
 887        if (tb[IFLA_TXQLEN])
 888                dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
 889
 890        if (tb[IFLA_OPERSTATE])
 891                set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
 892
 893        if (tb[IFLA_LINKMODE]) {
 894                write_lock_bh(&dev_base_lock);
 895                dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
 896                write_unlock_bh(&dev_base_lock);
 897        }
 898
 899        err = 0;
 900
 901errout:
 902        if (err < 0 && modified && net_ratelimit())
 903                printk(KERN_WARNING "A link change request failed with "
 904                       "some changes comitted already. Interface %s may "
 905                       "have been left with an inconsistent configuration, "
 906                       "please check.\n", dev->name);
 907
 908        if (send_addr_notify)
 909                call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
 910        return err;
 911}
 912
 913static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
 914{
 915        struct net *net = sock_net(skb->sk);
 916        struct ifinfomsg *ifm;
 917        struct net_device *dev;
 918        int err;
 919        struct nlattr *tb[IFLA_MAX+1];
 920        char ifname[IFNAMSIZ];
 921
 922        err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
 923        if (err < 0)
 924                goto errout;
 925
 926        if (tb[IFLA_IFNAME])
 927                nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
 928        else
 929                ifname[0] = '\0';
 930
 931        err = -EINVAL;
 932        ifm = nlmsg_data(nlh);
 933        if (ifm->ifi_index > 0)
 934                dev = dev_get_by_index(net, ifm->ifi_index);
 935        else if (tb[IFLA_IFNAME])
 936                dev = dev_get_by_name(net, ifname);
 937        else
 938                goto errout;
 939
 940        if (dev == NULL) {
 941                err = -ENODEV;
 942                goto errout;
 943        }
 944
 945        if ((err = validate_linkmsg(dev, tb)) < 0)
 946                goto errout_dev;
 947
 948        err = do_setlink(dev, ifm, tb, ifname, 0);
 949errout_dev:
 950        dev_put(dev);
 951errout:
 952        return err;
 953}
 954
 955static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
 956{
 957        struct net *net = sock_net(skb->sk);
 958        const struct rtnl_link_ops *ops;
 959        struct net_device *dev;
 960        struct ifinfomsg *ifm;
 961        char ifname[IFNAMSIZ];
 962        struct nlattr *tb[IFLA_MAX+1];
 963        int err;
 964
 965        err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
 966        if (err < 0)
 967                return err;
 968
 969        if (tb[IFLA_IFNAME])
 970                nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
 971
 972        ifm = nlmsg_data(nlh);
 973        if (ifm->ifi_index > 0)
 974                dev = __dev_get_by_index(net, ifm->ifi_index);
 975        else if (tb[IFLA_IFNAME])
 976                dev = __dev_get_by_name(net, ifname);
 977        else
 978                return -EINVAL;
 979
 980        if (!dev)
 981                return -ENODEV;
 982
 983        ops = dev->rtnl_link_ops;
 984        if (!ops)
 985                return -EOPNOTSUPP;
 986
 987        ops->dellink(dev);
 988        return 0;
 989}
 990
 991struct net_device *rtnl_create_link(struct net *net, char *ifname,
 992                const struct rtnl_link_ops *ops, struct nlattr *tb[])
 993{
 994        int err;
 995        struct net_device *dev;
 996
 997        err = -ENOMEM;
 998        dev = alloc_netdev(ops->priv_size, ifname, ops->setup);
 999        if (!dev)
1000                goto err;
1001
1002        if (strchr(dev->name, '%')) {
1003                err = dev_alloc_name(dev, dev->name);
1004                if (err < 0)
1005                        goto err_free;
1006        }
1007
1008        dev_net_set(dev, net);
1009        dev->rtnl_link_ops = ops;
1010
1011        if (tb[IFLA_MTU])
1012                dev->mtu = nla_get_u32(tb[IFLA_MTU]);
1013        if (tb[IFLA_ADDRESS])
1014                memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
1015                                nla_len(tb[IFLA_ADDRESS]));
1016        if (tb[IFLA_BROADCAST])
1017                memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
1018                                nla_len(tb[IFLA_BROADCAST]));
1019        if (tb[IFLA_TXQLEN])
1020                dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1021        if (tb[IFLA_OPERSTATE])
1022                set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1023        if (tb[IFLA_LINKMODE])
1024                dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
1025
1026        return dev;
1027
1028err_free:
1029        free_netdev(dev);
1030err:
1031        return ERR_PTR(err);
1032}
1033
1034static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
1035{
1036        struct net *net = sock_net(skb->sk);
1037        const struct rtnl_link_ops *ops;
1038        struct net_device *dev;
1039        struct ifinfomsg *ifm;
1040        char kind[MODULE_NAME_LEN];
1041        char ifname[IFNAMSIZ];
1042        struct nlattr *tb[IFLA_MAX+1];
1043        struct nlattr *linkinfo[IFLA_INFO_MAX+1];
1044        int err;
1045
1046#ifdef CONFIG_MODULES
1047replay:
1048#endif
1049        err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1050        if (err < 0)
1051                return err;
1052
1053        if (tb[IFLA_IFNAME])
1054                nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
1055        else
1056                ifname[0] = '\0';
1057
1058        ifm = nlmsg_data(nlh);
1059        if (ifm->ifi_index > 0)
1060                dev = __dev_get_by_index(net, ifm->ifi_index);
1061        else if (ifname[0])
1062                dev = __dev_get_by_name(net, ifname);
1063        else
1064                dev = NULL;
1065
1066        if ((err = validate_linkmsg(dev, tb)) < 0)
1067                return err;
1068
1069        if (tb[IFLA_LINKINFO]) {
1070                err = nla_parse_nested(linkinfo, IFLA_INFO_MAX,
1071                                       tb[IFLA_LINKINFO], ifla_info_policy);
1072                if (err < 0)
1073                        return err;
1074        } else
1075                memset(linkinfo, 0, sizeof(linkinfo));
1076
1077        if (linkinfo[IFLA_INFO_KIND]) {
1078                nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
1079                ops = rtnl_link_ops_get(kind);
1080        } else {
1081                kind[0] = '\0';
1082                ops = NULL;
1083        }
1084
1085        if (1) {
1086                struct nlattr *attr[ops ? ops->maxtype + 1 : 0], **data = NULL;
1087
1088                if (ops) {
1089                        if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
1090                                err = nla_parse_nested(attr, ops->maxtype,
1091                                                       linkinfo[IFLA_INFO_DATA],
1092                                                       ops->policy);
1093                                if (err < 0)
1094                                        return err;
1095                                data = attr;
1096                        }
1097                        if (ops->validate) {
1098                                err = ops->validate(tb, data);
1099                                if (err < 0)
1100                                        return err;
1101                        }
1102                }
1103
1104                if (dev) {
1105                        int modified = 0;
1106
1107                        if (nlh->nlmsg_flags & NLM_F_EXCL)
1108                                return -EEXIST;
1109                        if (nlh->nlmsg_flags & NLM_F_REPLACE)
1110                                return -EOPNOTSUPP;
1111
1112                        if (linkinfo[IFLA_INFO_DATA]) {
1113                                if (!ops || ops != dev->rtnl_link_ops ||
1114                                    !ops->changelink)
1115                                        return -EOPNOTSUPP;
1116
1117                                err = ops->changelink(dev, tb, data);
1118                                if (err < 0)
1119                                        return err;
1120                                modified = 1;
1121                        }
1122
1123                        return do_setlink(dev, ifm, tb, ifname, modified);
1124                }
1125
1126                if (!(nlh->nlmsg_flags & NLM_F_CREATE))
1127                        return -ENODEV;
1128
1129                if (ifm->ifi_index || ifm->ifi_flags || ifm->ifi_change)
1130                        return -EOPNOTSUPP;
1131                if (tb[IFLA_MAP] || tb[IFLA_MASTER] || tb[IFLA_PROTINFO])
1132                        return -EOPNOTSUPP;
1133
1134                if (!ops) {
1135#ifdef CONFIG_MODULES
1136                        if (kind[0]) {
1137                                __rtnl_unlock();
1138                                request_module("rtnl-link-%s", kind);
1139                                rtnl_lock();
1140                                ops = rtnl_link_ops_get(kind);
1141                                if (ops)
1142                                        goto replay;
1143                        }
1144#endif
1145                        return -EOPNOTSUPP;
1146                }
1147
1148                if (!ifname[0])
1149                        snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
1150
1151                dev = rtnl_create_link(net, ifname, ops, tb);
1152
1153                if (IS_ERR(dev))
1154                        err = PTR_ERR(dev);
1155                else if (ops->newlink)
1156                        err = ops->newlink(dev, tb, data);
1157                else
1158                        err = register_netdevice(dev);
1159
1160                if (err < 0 && !IS_ERR(dev))
1161                        free_netdev(dev);
1162                return err;
1163        }
1164}
1165
1166static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
1167{
1168        struct net *net = sock_net(skb->sk);
1169        struct ifinfomsg *ifm;
1170        struct nlattr *tb[IFLA_MAX+1];
1171        struct net_device *dev = NULL;
1172        struct sk_buff *nskb;
1173        int err;
1174
1175        err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy);
1176        if (err < 0)
1177                return err;
1178
1179        ifm = nlmsg_data(nlh);
1180        if (ifm->ifi_index > 0) {
1181                dev = dev_get_by_index(net, ifm->ifi_index);
1182                if (dev == NULL)
1183                        return -ENODEV;
1184        } else
1185                return -EINVAL;
1186
1187        nskb = nlmsg_new(if_nlmsg_size(dev), GFP_KERNEL);
1188        if (nskb == NULL) {
1189                err = -ENOBUFS;
1190                goto errout;
1191        }
1192
1193        err = rtnl_fill_ifinfo(nskb, dev, RTM_NEWLINK, NETLINK_CB(skb).pid,
1194                               nlh->nlmsg_seq, 0, 0);
1195        if (err < 0) {
1196                /* -EMSGSIZE implies BUG in if_nlmsg_size */
1197                WARN_ON(err == -EMSGSIZE);
1198                kfree_skb(nskb);
1199                goto errout;
1200        }
1201        err = rtnl_unicast(nskb, net, NETLINK_CB(skb).pid);
1202errout:
1203        dev_put(dev);
1204
1205        return err;
1206}
1207
1208static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
1209{
1210        int idx;
1211        int s_idx = cb->family;
1212
1213        if (s_idx == 0)
1214                s_idx = 1;
1215        for (idx=1; idx<NPROTO; idx++) {
1216                int type = cb->nlh->nlmsg_type-RTM_BASE;
1217                if (idx < s_idx || idx == PF_PACKET)
1218                        continue;
1219                if (rtnl_msg_handlers[idx] == NULL ||
1220                    rtnl_msg_handlers[idx][type].dumpit == NULL)
1221                        continue;
1222                if (idx > s_idx)
1223                        memset(&cb->args[0], 0, sizeof(cb->args));
1224                if (rtnl_msg_handlers[idx][type].dumpit(skb, cb))
1225                        break;
1226        }
1227        cb->family = idx;
1228
1229        return skb->len;
1230}
1231
1232void rtmsg_ifinfo(int type, struct net_device *dev, unsigned change)
1233{
1234        struct net *net = dev_net(dev);
1235        struct sk_buff *skb;
1236        int err = -ENOBUFS;
1237
1238        skb = nlmsg_new(if_nlmsg_size(dev), GFP_KERNEL);
1239        if (skb == NULL)
1240                goto errout;
1241
1242        err = rtnl_fill_ifinfo(skb, dev, type, 0, 0, change, 0);
1243        if (err < 0) {
1244                /* -EMSGSIZE implies BUG in if_nlmsg_size() */
1245                WARN_ON(err == -EMSGSIZE);
1246                kfree_skb(skb);
1247                goto errout;
1248        }
1249        rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_KERNEL);
1250        return;
1251errout:
1252        if (err < 0)
1253                rtnl_set_sk_err(net, RTNLGRP_LINK, err);
1254}
1255
1256/* Protected by RTNL sempahore.  */
1257static struct rtattr **rta_buf;
1258static int rtattr_max;
1259
1260/* Process one rtnetlink message. */
1261
1262static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
1263{
1264        struct net *net = sock_net(skb->sk);
1265        rtnl_doit_func doit;
1266        int sz_idx, kind;
1267        int min_len;
1268        int family;
1269        int type;
1270        int err;
1271
1272        type = nlh->nlmsg_type;
1273        if (type > RTM_MAX)
1274                return -EOPNOTSUPP;
1275
1276        type -= RTM_BASE;
1277
1278        /* All the messages must have at least 1 byte length */
1279        if (nlh->nlmsg_len < NLMSG_LENGTH(sizeof(struct rtgenmsg)))
1280                return 0;
1281
1282        family = ((struct rtgenmsg*)NLMSG_DATA(nlh))->rtgen_family;
1283        if (family >= NPROTO)
1284                return -EAFNOSUPPORT;
1285
1286        sz_idx = type>>2;
1287        kind = type&3;
1288
1289        if (kind != 2 && security_netlink_recv(skb, CAP_NET_ADMIN))
1290                return -EPERM;
1291
1292        if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
1293                struct sock *rtnl;
1294                rtnl_dumpit_func dumpit;
1295
1296                dumpit = rtnl_get_dumpit(family, type);
1297                if (dumpit == NULL)
1298                        return -EOPNOTSUPP;
1299
1300                __rtnl_unlock();
1301                rtnl = net->rtnl;
1302                err = netlink_dump_start(rtnl, skb, nlh, dumpit, NULL);
1303                rtnl_lock();
1304                return err;
1305        }
1306
1307        memset(rta_buf, 0, (rtattr_max * sizeof(struct rtattr *)));
1308
1309        min_len = rtm_min[sz_idx];
1310        if (nlh->nlmsg_len < min_len)
1311                return -EINVAL;
1312
1313        if (nlh->nlmsg_len > min_len) {
1314                int attrlen = nlh->nlmsg_len - NLMSG_ALIGN(min_len);
1315                struct rtattr *attr = (void*)nlh + NLMSG_ALIGN(min_len);
1316
1317                while (RTA_OK(attr, attrlen)) {
1318                        unsigned flavor = attr->rta_type;
1319                        if (flavor) {
1320                                if (flavor > rta_max[sz_idx])
1321                                        return -EINVAL;
1322                                rta_buf[flavor-1] = attr;
1323                        }
1324                        attr = RTA_NEXT(attr, attrlen);
1325                }
1326        }
1327
1328        doit = rtnl_get_doit(family, type);
1329        if (doit == NULL)
1330                return -EOPNOTSUPP;
1331
1332        return doit(skb, nlh, (void *)&rta_buf[0]);
1333}
1334
1335static void rtnetlink_rcv(struct sk_buff *skb)
1336{
1337        rtnl_lock();
1338        netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
1339        rtnl_unlock();
1340}
1341
1342static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
1343{
1344        struct net_device *dev = ptr;
1345
1346        switch (event) {
1347        case NETDEV_UNREGISTER:
1348                rtmsg_ifinfo(RTM_DELLINK, dev, ~0U);
1349                break;
1350        case NETDEV_REGISTER:
1351                rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U);
1352                break;
1353        case NETDEV_UP:
1354        case NETDEV_DOWN:
1355                rtmsg_ifinfo(RTM_NEWLINK, dev, IFF_UP|IFF_RUNNING);
1356                break;
1357        case NETDEV_CHANGE:
1358        case NETDEV_GOING_DOWN:
1359                break;
1360        default:
1361                rtmsg_ifinfo(RTM_NEWLINK, dev, 0);
1362                break;
1363        }
1364        return NOTIFY_DONE;
1365}
1366
1367static struct notifier_block rtnetlink_dev_notifier = {
1368        .notifier_call  = rtnetlink_event,
1369};
1370
1371
1372static int rtnetlink_net_init(struct net *net)
1373{
1374        struct sock *sk;
1375        sk = netlink_kernel_create(net, NETLINK_ROUTE, RTNLGRP_MAX,
1376                                   rtnetlink_rcv, &rtnl_mutex, THIS_MODULE);
1377        if (!sk)
1378                return -ENOMEM;
1379        net->rtnl = sk;
1380        return 0;
1381}
1382
1383static void rtnetlink_net_exit(struct net *net)
1384{
1385        netlink_kernel_release(net->rtnl);
1386        net->rtnl = NULL;
1387}
1388
1389static struct pernet_operations rtnetlink_net_ops = {
1390        .init = rtnetlink_net_init,
1391        .exit = rtnetlink_net_exit,
1392};
1393
1394void __init rtnetlink_init(void)
1395{
1396        int i;
1397
1398        rtattr_max = 0;
1399        for (i = 0; i < ARRAY_SIZE(rta_max); i++)
1400                if (rta_max[i] > rtattr_max)
1401                        rtattr_max = rta_max[i];
1402        rta_buf = kmalloc(rtattr_max * sizeof(struct rtattr *), GFP_KERNEL);
1403        if (!rta_buf)
1404                panic("rtnetlink_init: cannot allocate rta_buf\n");
1405
1406        if (register_pernet_subsys(&rtnetlink_net_ops))
1407                panic("rtnetlink_init: cannot initialize rtnetlink\n");
1408
1409        netlink_set_nonroot(NETLINK_ROUTE, NL_NONROOT_RECV);
1410        register_netdevice_notifier(&rtnetlink_dev_notifier);
1411
1412        rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink, rtnl_dump_ifinfo);
1413        rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL);
1414        rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL);
1415        rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL);
1416
1417        rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all);
1418        rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all);
1419}
1420
1421EXPORT_SYMBOL(__rta_fill);
1422EXPORT_SYMBOL(rtnetlink_put_metrics);
1423EXPORT_SYMBOL(rtnl_lock);
1424EXPORT_SYMBOL(rtnl_trylock);
1425EXPORT_SYMBOL(rtnl_unlock);
1426EXPORT_SYMBOL(rtnl_is_locked);
1427EXPORT_SYMBOL(rtnl_unicast);
1428EXPORT_SYMBOL(rtnl_notify);
1429EXPORT_SYMBOL(rtnl_set_sk_err);
1430EXPORT_SYMBOL(rtnl_create_link);
1431EXPORT_SYMBOL(ifla_policy);
1432