linux-old/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/config.h>
  20#include <linux/errno.h>
  21#include <linux/types.h>
  22#include <linux/socket.h>
  23#include <linux/kernel.h>
  24#include <linux/major.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
  38#include <asm/uaccess.h>
  39#include <asm/system.h>
  40#include <asm/string.h>
  41
  42#include <linux/inet.h>
  43#include <linux/netdevice.h>
  44#include <net/ip.h>
  45#include <net/protocol.h>
  46#include <net/arp.h>
  47#include <net/route.h>
  48#include <net/udp.h>
  49#include <net/sock.h>
  50#include <net/pkt_sched.h>
  51
  52DECLARE_MUTEX(rtnl_sem);
  53
  54void rtnl_lock(void)
  55{
  56        rtnl_shlock();
  57        rtnl_exlock();
  58}
  59 
  60void rtnl_unlock(void)
  61{
  62        rtnl_exunlock();
  63        rtnl_shunlock();
  64}
  65
  66int rtattr_parse(struct rtattr *tb[], int maxattr, struct rtattr *rta, int len)
  67{
  68        memset(tb, 0, sizeof(struct rtattr*)*maxattr);
  69
  70        while (RTA_OK(rta, len)) {
  71                unsigned flavor = rta->rta_type;
  72                if (flavor && flavor <= maxattr)
  73                        tb[flavor-1] = rta;
  74                rta = RTA_NEXT(rta, len);
  75        }
  76        return 0;
  77}
  78
  79struct sock *rtnl;
  80
  81struct rtnetlink_link * rtnetlink_links[NPROTO];
  82
  83static const int rtm_min[(RTM_MAX+1-RTM_BASE)/4] =
  84{
  85        NLMSG_LENGTH(sizeof(struct ifinfomsg)),
  86        NLMSG_LENGTH(sizeof(struct ifaddrmsg)),
  87        NLMSG_LENGTH(sizeof(struct rtmsg)),
  88        NLMSG_LENGTH(sizeof(struct ndmsg)),
  89        NLMSG_LENGTH(sizeof(struct rtmsg)),
  90        NLMSG_LENGTH(sizeof(struct tcmsg)),
  91        NLMSG_LENGTH(sizeof(struct tcmsg)),
  92        NLMSG_LENGTH(sizeof(struct tcmsg))
  93};
  94
  95static const int rta_max[(RTM_MAX+1-RTM_BASE)/4] =
  96{
  97        IFLA_MAX,
  98        IFA_MAX,
  99        RTA_MAX,
 100        NDA_MAX,
 101        RTA_MAX,
 102        TCA_MAX,
 103        TCA_MAX,
 104        TCA_MAX
 105};
 106
 107void __rta_fill(struct sk_buff *skb, int attrtype, int attrlen, const void *data)
 108{
 109        struct rtattr *rta;
 110        int size = RTA_LENGTH(attrlen);
 111
 112        rta = (struct rtattr*)skb_put(skb, RTA_ALIGN(size));
 113        rta->rta_type = attrtype;
 114        rta->rta_len = size;
 115        memcpy(RTA_DATA(rta), data, attrlen);
 116}
 117
 118int rtnetlink_send(struct sk_buff *skb, u32 pid, unsigned group, int echo)
 119{
 120        int err = 0;
 121
 122        NETLINK_CB(skb).dst_groups = group;
 123        if (echo)
 124                atomic_inc(&skb->users);
 125        netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
 126        if (echo)
 127                err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
 128        return err;
 129}
 130
 131int rtnetlink_put_metrics(struct sk_buff *skb, unsigned *metrics)
 132{
 133        struct rtattr *mx = (struct rtattr*)skb->tail;
 134        int i;
 135
 136        RTA_PUT(skb, RTA_METRICS, 0, NULL);
 137        for (i=0; i<RTAX_MAX; i++) {
 138                if (metrics[i])
 139                        RTA_PUT(skb, i+1, sizeof(unsigned), metrics+i);
 140        }
 141        mx->rta_len = skb->tail - (u8*)mx;
 142        if (mx->rta_len == RTA_LENGTH(0))
 143                skb_trim(skb, (u8*)mx - skb->data);
 144        return 0;
 145
 146rtattr_failure:
 147        skb_trim(skb, (u8*)mx - skb->data);
 148        return -1;
 149}
 150
 151
 152static int rtnetlink_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
 153                                 int type, u32 pid, u32 seq, u32 change)
 154{
 155        struct ifinfomsg *r;
 156        struct nlmsghdr  *nlh;
 157        unsigned char    *b = skb->tail;
 158
 159        nlh = NLMSG_PUT(skb, pid, seq, type, sizeof(*r));
 160        if (pid) nlh->nlmsg_flags |= NLM_F_MULTI;
 161        r = NLMSG_DATA(nlh);
 162        r->ifi_family = AF_UNSPEC;
 163        r->ifi_type = dev->type;
 164        r->ifi_index = dev->ifindex;
 165        r->ifi_flags = dev->flags;
 166        r->ifi_change = change;
 167
 168        if (!netif_running(dev) || !netif_carrier_ok(dev))
 169                r->ifi_flags &= ~IFF_RUNNING;
 170        else
 171                r->ifi_flags |= IFF_RUNNING;
 172
 173        RTA_PUT(skb, IFLA_IFNAME, strlen(dev->name)+1, dev->name);
 174        if (dev->addr_len) {
 175                RTA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
 176                RTA_PUT(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast);
 177        }
 178        if (1) {
 179                unsigned mtu = dev->mtu;
 180                RTA_PUT(skb, IFLA_MTU, sizeof(mtu), &mtu);
 181        }
 182        if (dev->ifindex != dev->iflink)
 183                RTA_PUT(skb, IFLA_LINK, sizeof(int), &dev->iflink);
 184        if (dev->qdisc_sleeping)
 185                RTA_PUT(skb, IFLA_QDISC,
 186                        strlen(dev->qdisc_sleeping->ops->id) + 1,
 187                        dev->qdisc_sleeping->ops->id);
 188        if (dev->master)
 189                RTA_PUT(skb, IFLA_MASTER, sizeof(int), &dev->master->ifindex);
 190        if (dev->get_stats) {
 191                struct net_device_stats *stats = dev->get_stats(dev);
 192                if (stats)
 193                        RTA_PUT(skb, IFLA_STATS, sizeof(*stats), stats);
 194        }
 195        nlh->nlmsg_len = skb->tail - b;
 196        return skb->len;
 197
 198nlmsg_failure:
 199rtattr_failure:
 200        skb_trim(skb, b - skb->data);
 201        return -1;
 202}
 203
 204int rtnetlink_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
 205{
 206        int idx;
 207        int s_idx = cb->args[0];
 208        struct net_device *dev;
 209
 210        read_lock(&dev_base_lock);
 211        for (dev=dev_base, idx=0; dev; dev = dev->next, idx++) {
 212                if (idx < s_idx)
 213                        continue;
 214                if (rtnetlink_fill_ifinfo(skb, dev, RTM_NEWLINK, NETLINK_CB(cb->skb).pid, cb->nlh->nlmsg_seq, 0) <= 0)
 215                        break;
 216        }
 217        read_unlock(&dev_base_lock);
 218        cb->args[0] = idx;
 219
 220        return skb->len;
 221}
 222
 223int rtnetlink_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
 224{
 225        int idx;
 226        int s_idx = cb->family;
 227
 228        if (s_idx == 0)
 229                s_idx = 1;
 230        for (idx=1; idx<NPROTO; idx++) {
 231                int type = cb->nlh->nlmsg_type-RTM_BASE;
 232                if (idx < s_idx || idx == PF_PACKET)
 233                        continue;
 234                if (rtnetlink_links[idx] == NULL ||
 235                    rtnetlink_links[idx][type].dumpit == NULL)
 236                        continue;
 237                if (idx > s_idx)
 238                        memset(&cb->args[0], 0, sizeof(cb->args));
 239                if (rtnetlink_links[idx][type].dumpit(skb, cb))
 240                        break;
 241        }
 242        cb->family = idx;
 243
 244        return skb->len;
 245}
 246
 247void rtmsg_ifinfo(int type, struct net_device *dev, unsigned change)
 248{
 249        struct sk_buff *skb;
 250        int size = NLMSG_GOODSIZE;
 251
 252        skb = alloc_skb(size, GFP_KERNEL);
 253        if (!skb)
 254                return;
 255
 256        if (rtnetlink_fill_ifinfo(skb, dev, type, 0, 0, change) < 0) {
 257                kfree_skb(skb);
 258                return;
 259        }
 260        NETLINK_CB(skb).dst_groups = RTMGRP_LINK;
 261        netlink_broadcast(rtnl, skb, 0, RTMGRP_LINK, GFP_KERNEL);
 262}
 263
 264static int rtnetlink_done(struct netlink_callback *cb)
 265{
 266        return 0;
 267}
 268
 269/* Process one rtnetlink message. */
 270
 271static __inline__ int
 272rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh, int *errp)
 273{
 274        struct rtnetlink_link *link;
 275        struct rtnetlink_link *link_tab;
 276        struct rtattr   *rta[RTATTR_MAX];
 277
 278        int exclusive = 0;
 279        int sz_idx, kind;
 280        int min_len;
 281        int family;
 282        int type;
 283        int err;
 284
 285        /* Only requests are handled by kernel now */
 286        if (!(nlh->nlmsg_flags&NLM_F_REQUEST))
 287                return 0;
 288
 289        type = nlh->nlmsg_type;
 290
 291        /* A control message: ignore them */
 292        if (type < RTM_BASE)
 293                return 0;
 294
 295        /* Unknown message: reply with EINVAL */
 296        if (type > RTM_MAX)
 297                goto err_inval;
 298
 299        type -= RTM_BASE;
 300
 301        /* All the messages must have at least 1 byte length */
 302        if (nlh->nlmsg_len < NLMSG_LENGTH(sizeof(struct rtgenmsg)))
 303                return 0;
 304
 305        family = ((struct rtgenmsg*)NLMSG_DATA(nlh))->rtgen_family;
 306        if (family >= NPROTO) {
 307                *errp = -EAFNOSUPPORT;
 308                return -1;
 309        }
 310
 311        link_tab = rtnetlink_links[family];
 312        if (link_tab == NULL)
 313                link_tab = rtnetlink_links[PF_UNSPEC];
 314        link = &link_tab[type];
 315
 316        sz_idx = type>>2;
 317        kind = type&3;
 318
 319        if (kind != 2 && !cap_raised(NETLINK_CB(skb).eff_cap, CAP_NET_ADMIN)) {
 320                *errp = -EPERM;
 321                return -1;
 322        }
 323
 324        if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
 325                u32 rlen;
 326
 327                if (link->dumpit == NULL)
 328                        link = &(rtnetlink_links[PF_UNSPEC][type]);
 329
 330                if (link->dumpit == NULL)
 331                        goto err_inval;
 332
 333                if ((*errp = netlink_dump_start(rtnl, skb, nlh,
 334                                                link->dumpit,
 335                                                rtnetlink_done)) != 0) {
 336                        return -1;
 337                }
 338                rlen = NLMSG_ALIGN(nlh->nlmsg_len);
 339                if (rlen > skb->len)
 340                        rlen = skb->len;
 341                skb_pull(skb, rlen);
 342                return -1;
 343        }
 344
 345        if (kind != 2) {
 346                if (rtnl_exlock_nowait()) {
 347                        *errp = 0;
 348                        return -1;
 349                }
 350                exclusive = 1;
 351        }
 352
 353        memset(&rta, 0, sizeof(rta));
 354
 355        min_len = rtm_min[sz_idx];
 356        if (nlh->nlmsg_len < min_len)
 357                goto err_inval;
 358
 359        if (nlh->nlmsg_len > min_len) {
 360                int attrlen = nlh->nlmsg_len - NLMSG_ALIGN(min_len);
 361                struct rtattr *attr = (void*)nlh + NLMSG_ALIGN(min_len);
 362
 363                while (RTA_OK(attr, attrlen)) {
 364                        unsigned flavor = attr->rta_type;
 365                        if (flavor) {
 366                                if (flavor > rta_max[sz_idx])
 367                                        goto err_inval;
 368                                rta[flavor-1] = attr;
 369                        }
 370                        attr = RTA_NEXT(attr, attrlen);
 371                }
 372        }
 373
 374        if (link->doit == NULL)
 375                link = &(rtnetlink_links[PF_UNSPEC][type]);
 376        if (link->doit == NULL)
 377                goto err_inval;
 378        err = link->doit(skb, nlh, (void *)&rta);
 379
 380        if (exclusive)
 381                rtnl_exunlock();
 382        *errp = err;
 383        return err;
 384
 385err_inval:
 386        if (exclusive)
 387                rtnl_exunlock();
 388        *errp = -EINVAL;
 389        return -1;
 390}
 391
 392/* 
 393 * Process one packet of messages.
 394 * Malformed skbs with wrong lengths of messages are discarded silently.
 395 */
 396
 397static inline int rtnetlink_rcv_skb(struct sk_buff *skb)
 398{
 399        int err;
 400        struct nlmsghdr * nlh;
 401
 402        while (skb->len >= NLMSG_SPACE(0)) {
 403                u32 rlen;
 404
 405                nlh = (struct nlmsghdr *)skb->data;
 406                if (nlh->nlmsg_len < sizeof(*nlh) || skb->len < nlh->nlmsg_len)
 407                        return 0;
 408                rlen = NLMSG_ALIGN(nlh->nlmsg_len);
 409                if (rlen > skb->len)
 410                        rlen = skb->len;
 411                if (rtnetlink_rcv_msg(skb, nlh, &err)) {
 412                        /* Not error, but we must interrupt processing here:
 413                         *   Note, that in this case we do not pull message
 414                         *   from skb, it will be processed later.
 415                         */
 416                        if (err == 0)
 417                                return -1;
 418                        netlink_ack(skb, nlh, err);
 419                } else if (nlh->nlmsg_flags&NLM_F_ACK)
 420                        netlink_ack(skb, nlh, 0);
 421                skb_pull(skb, rlen);
 422        }
 423
 424        return 0;
 425}
 426
 427/*
 428 *  rtnetlink input queue processing routine:
 429 *      - try to acquire shared lock. If it is failed, defer processing.
 430 *      - feed skbs to rtnetlink_rcv_skb, until it refuse a message,
 431 *        that will occur, when a dump started and/or acquisition of
 432 *        exclusive lock failed.
 433 */
 434
 435static void rtnetlink_rcv(struct sock *sk, int len)
 436{
 437        do {
 438                struct sk_buff *skb;
 439
 440                if (rtnl_shlock_nowait())
 441                        return;
 442
 443                while ((skb = skb_dequeue(&sk->receive_queue)) != NULL) {
 444                        if (rtnetlink_rcv_skb(skb)) {
 445                                if (skb->len)
 446                                        skb_queue_head(&sk->receive_queue, skb);
 447                                else
 448                                        kfree_skb(skb);
 449                                break;
 450                        }
 451                        kfree_skb(skb);
 452                }
 453
 454                up(&rtnl_sem);
 455        } while (rtnl && rtnl->receive_queue.qlen);
 456}
 457
 458static struct rtnetlink_link link_rtnetlink_table[RTM_MAX-RTM_BASE+1] =
 459{
 460        { NULL,                 NULL,                   },
 461        { NULL,                 NULL,                   },
 462        { NULL,                 rtnetlink_dump_ifinfo,  },
 463        { NULL,                 NULL,                   },
 464
 465        { NULL,                 NULL,                   },
 466        { NULL,                 NULL,                   },
 467        { NULL,                 rtnetlink_dump_all,     },
 468        { NULL,                 NULL,                   },
 469
 470        { NULL,                 NULL,                   },
 471        { NULL,                 NULL,                   },
 472        { NULL,                 rtnetlink_dump_all,     },
 473        { NULL,                 NULL,                   },
 474
 475        { neigh_add,            NULL,                   },
 476        { neigh_delete,         NULL,                   },
 477        { NULL,                 neigh_dump_info,        },
 478        { NULL,                 NULL,                   },
 479
 480        { NULL,                 NULL,                   },
 481        { NULL,                 NULL,                   },
 482        { NULL,                 NULL,                   },
 483        { NULL,                 NULL,                   },
 484};
 485
 486
 487static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
 488{
 489        struct net_device *dev = ptr;
 490        switch (event) {
 491        case NETDEV_UNREGISTER:
 492                rtmsg_ifinfo(RTM_DELLINK, dev, ~0U);
 493                break;
 494        case NETDEV_REGISTER:
 495                rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U);
 496                break;
 497        case NETDEV_UP:
 498        case NETDEV_DOWN:
 499                rtmsg_ifinfo(RTM_NEWLINK, dev, IFF_UP|IFF_RUNNING);
 500                break;
 501        case NETDEV_CHANGE:
 502        case NETDEV_GOING_DOWN:
 503                break;
 504        default:
 505                rtmsg_ifinfo(RTM_NEWLINK, dev, 0);
 506                break;
 507        }
 508        return NOTIFY_DONE;
 509}
 510
 511struct notifier_block rtnetlink_dev_notifier = {
 512        rtnetlink_event,
 513        NULL,
 514        0
 515};
 516
 517
 518void __init rtnetlink_init(void)
 519{
 520#ifdef RTNL_DEBUG
 521        printk("Initializing RT netlink socket\n");
 522#endif
 523        rtnl = netlink_kernel_create(NETLINK_ROUTE, rtnetlink_rcv);
 524        if (rtnl == NULL)
 525                panic("rtnetlink_init: cannot initialize rtnetlink\n");
 526        netlink_set_nonroot(NETLINK_ROUTE, NL_NONROOT_RECV);
 527        register_netdevice_notifier(&rtnetlink_dev_notifier);
 528        rtnetlink_links[PF_UNSPEC] = link_rtnetlink_table;
 529        rtnetlink_links[PF_PACKET] = link_rtnetlink_table;
 530}
 531
lxr.linux.no kindly hosted by Redpill Linpro AS, provider of Linux consulting and operations services since 1995.