i386/linux/linux-2.3.21/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/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/tcp.h>
  49 #include <net/udp.h>
  50 #include <net/sock.h>
  51 #include <net/pkt_sched.h>
  52
  53 DECLARE_MUTEX(rtnl_sem);
  54
  55 void rtnl_lock(void)
  56 {
  57         rtnl_shlock();
  58         rtnl_exlock();
  59 }
  60
  61 void rtnl_unlock(void)
  62 {
  63         rtnl_exunlock();
  64         rtnl_shunlock();
  65 }
  66
  67 int rtattr_parse(struct rtattr *tb[], int maxattr, struct rtattr *rta, int len)
  68 {
  69         memset(tb, 0, sizeof(struct rtattr*)*maxattr);
  70
  71         while (RTA_OK(rta, len)) {
  72                 unsigned flavor = rta->rta_type;
  73                 if (flavor && flavor <= maxattr)
  74                         tb[flavor-1] = rta;
  75                 rta = RTA_NEXT(rta, len);
  76         }
  77         return 0;
  78 }
  79
  80 #ifdef CONFIG_RTNETLINK
  81 struct sock *rtnl;
  82
  83 struct rtnetlink_link * rtnetlink_links[NPROTO];
  84
  85 #define _S      1       /* superuser privileges required */
  86 #define _X      2       /* exclusive access to tables required */
  87 #define _G      4       /* GET request */
  88
  89 static const int rtm_min[(RTM_MAX+1-RTM_BASE)/4] =
  90 {
  91         NLMSG_LENGTH(sizeof(struct ifinfomsg)),
  92         NLMSG_LENGTH(sizeof(struct ifaddrmsg)),
  93         NLMSG_LENGTH(sizeof(struct rtmsg)),
  94         NLMSG_LENGTH(sizeof(struct ndmsg)),
  95         NLMSG_LENGTH(sizeof(struct rtmsg)),
  96         NLMSG_LENGTH(sizeof(struct tcmsg)),
  97         NLMSG_LENGTH(sizeof(struct tcmsg)),
  98         NLMSG_LENGTH(sizeof(struct tcmsg))
  99 };
 100
 101 static const int rta_max[(RTM_MAX+1-RTM_BASE)/4] =
 102 {
 103         IFLA_MAX,
 104         IFA_MAX,
 105         RTA_MAX,
 106         NDA_MAX,
 107         RTA_MAX,
 108         TCA_MAX,
 109         TCA_MAX,
 110         TCA_MAX
 111 };
 112
 113 void __rta_fill(struct sk_buff *skb, int attrtype, int attrlen, const void *data)
 114 {
 115         struct rtattr *rta;
 116         int size = RTA_LENGTH(attrlen);
 117
 118         rta = (struct rtattr*)skb_put(skb, RTA_ALIGN(size));
 119         rta->rta_type = attrtype;
 120         rta->rta_len = size;
 121         memcpy(RTA_DATA(rta), data, attrlen);
 122 }
 123
 124 int rtnetlink_send(struct sk_buff *skb, u32 pid, unsigned group, int echo)
 125 {
 126         int err = 0;
 127
 128         NETLINK_CB(skb).dst_groups = group;
 129         if (echo)
 130                 atomic_inc(&skb->users);
 131         netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
 132         if (echo)
 133                 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
 134         return err;
 135 }
 136
 137 int rtnetlink_put_metrics(struct sk_buff *skb, unsigned *metrics)
 138 {
 139         struct rtattr *mx = (struct rtattr*)skb->tail;
 140         int i;
 141
 142         RTA_PUT(skb, RTA_METRICS, 0, NULL);
 143         for (i=0; i<RTAX_MAX; i++) {
 144                 if (metrics[i])
 145                         RTA_PUT(skb, i+1, sizeof(unsigned), metrics+i);
 146         }
 147         mx->rta_len = skb->tail - (u8*)mx;
 148         if (mx->rta_len == RTA_LENGTH(0))
 149                 skb_trim(skb, (u8*)mx - skb->data);
 150         return 0;
 151
 152 rtattr_failure:
 153         skb_trim(skb, (u8*)mx - skb->data);
 154         return -1;
 155 }
 156
 157
 158 static int rtnetlink_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
 159                                  int type, u32 pid, u32 seq, u32 change)
 160 {
 161         struct ifinfomsg *r;
 162         struct nlmsghdr  *nlh;
 163         unsigned char    *b = skb->tail;
 164
 165         nlh = NLMSG_PUT(skb, pid, seq, type, sizeof(*r));
 166         if (pid) nlh->nlmsg_flags |= NLM_F_MULTI;
 167         r = NLMSG_DATA(nlh);
 168         r->ifi_family = AF_UNSPEC;
 169         r->ifi_type = dev->type;
 170         r->ifi_index = dev->ifindex;
 171         r->ifi_flags = dev->flags;
 172         r->ifi_change = change;
 173
 174         RTA_PUT(skb, IFLA_IFNAME, strlen(dev->name)+1, dev->name);
 175         if (dev->addr_len) {
 176                 RTA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
 177                 RTA_PUT(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast);
 178         }
 179         if (1) {
 180                 unsigned mtu = dev->mtu;
 181                 RTA_PUT(skb, IFLA_MTU, sizeof(mtu), &mtu);
 182         }
 183         if (dev->ifindex != dev->iflink)
 184                 RTA_PUT(skb, IFLA_LINK, sizeof(int), &dev->iflink);
 185         if (dev->qdisc_sleeping)
 186                 RTA_PUT(skb, IFLA_QDISC,
 187                         strlen(dev->qdisc_sleeping->ops->id) + 1,
 188                         dev->qdisc_sleeping->ops->id);
 189         if (dev->get_stats) {
 190                 struct net_device_stats *stats = dev->get_stats(dev);
 191                 if (stats)
 192                         RTA_PUT(skb, IFLA_STATS, sizeof(*stats), stats);
 193         }
 194         nlh->nlmsg_len = skb->tail - b;
 195         return skb->len;
 196
 197 nlmsg_failure:
 198 rtattr_failure:
 199         skb_trim(skb, b - skb->data);
 200         return -1;
 201 }
 202
 203 int rtnetlink_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
 204 {
 205         int idx;
 206         int s_idx = cb->args[0];
 207         struct net_device *dev;
 208
 209         read_lock(&dev_base_lock);
 210         for (dev=dev_base, idx=0; dev; dev = dev->next, idx++) {
 211                 if (idx < s_idx)
 212                         continue;
 213                 if (rtnetlink_fill_ifinfo(skb, dev, RTM_NEWLINK, NETLINK_CB(cb->skb).pid, cb->nlh->nlmsg_seq, 0) <= 0)
 214                         break;
 215         }
 216         read_unlock(&dev_base_lock);
 217         cb->args[0] = idx;
 218
 219         return skb->len;
 220 }
 221
 222 int rtnetlink_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
 223 {
 224         int idx;
 225         int s_idx = cb->family;
 226
 227         if (s_idx == 0)
 228                 s_idx = 1;
 229         for (idx=1; idx<NPROTO; idx++) {
 230                 int type = cb->nlh->nlmsg_type-RTM_BASE;
 231                 if (idx < s_idx || idx == PF_PACKET)
 232                         continue;
 233                 if (rtnetlink_links[idx] == NULL ||
 234                     rtnetlink_links[idx][type].dumpit == NULL)
 235                         continue;
 236                 if (idx > s_idx)
 237                         memset(&cb->args[0], 0, sizeof(cb->args));
 238                 if (rtnetlink_links[idx][type].dumpit(skb, cb))
 239                         break;
 240         }
 241         cb->family = idx;
 242
 243         return skb->len;
 244 }
 245
 246 void rtmsg_ifinfo(int type, struct net_device *dev)
 247 {
 248         struct sk_buff *skb;
 249         int size = NLMSG_GOODSIZE;
 250
 251         skb = alloc_skb(size, GFP_KERNEL);
 252         if (!skb)
 253                 return;
 254
 255         if (rtnetlink_fill_ifinfo(skb, dev, type, 0, 0, ~0U) < 0) {
 256                 kfree_skb(skb);
 257                 return;
 258         }
 259         NETLINK_CB(skb).dst_groups = RTMGRP_LINK;
 260         netlink_broadcast(rtnl, skb, 0, RTMGRP_LINK, GFP_KERNEL);
 261 }
 262
 263 static int rtnetlink_done(struct netlink_callback *cb)
 264 {
 265         return 0;
 266 }
 267
 268 /* Process one rtnetlink message. */
 269
 270 extern __inline__ int
 271 rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh, int *errp)
 272 {
 273         struct rtnetlink_link *link;
 274         struct rtnetlink_link *link_tab;
 275         struct rtattr   *rta[RTATTR_MAX];
 276
 277         int exclusive = 0;
 278         int sz_idx, kind;
 279         int min_len;
 280         int family;
 281         int type;
 282         int err;
 283
 284         /* Only requests are handled by kernel now */
 285         if (!(nlh->nlmsg_flags&NLM_F_REQUEST))
 286                 return 0;
 287
 288         type = nlh->nlmsg_type;
 289
 290         /* A control message: ignore them */
 291         if (type < RTM_BASE)
 292                 return 0;
 293
 294         /* Unknown message: reply with EINVAL */
 295         if (type > RTM_MAX)
 296                 goto err_inval;
 297
 298         type -= RTM_BASE;
 299
 300         /* All the messages must have at least 1 byte length */
 301         if (nlh->nlmsg_len < NLMSG_LENGTH(sizeof(struct rtgenmsg)))
 302                 return 0;
 303
 304         family = ((struct rtgenmsg*)NLMSG_DATA(nlh))->rtgen_family;
 305         if (family > NPROTO) {
 306                 *errp = -EAFNOSUPPORT;
 307                 return -1;
 308         }
 309
 310         link_tab = rtnetlink_links[family];
 311         if (link_tab == NULL)
 312                 link_tab = rtnetlink_links[PF_UNSPEC];
 313         link = &link_tab[type];
 314
 315         sz_idx = type>>2;
 316         kind = type&3;
 317
 318         if (kind != 2 && !cap_raised(NETLINK_CB(skb).eff_cap, CAP_NET_ADMIN)) {
 319                 *errp = -EPERM;
 320                 return -1;
 321         }
 322
 323         if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
 324                 u32 rlen;
 325
 326                 if (link->dumpit == NULL)
 327                         link = &(rtnetlink_links[PF_UNSPEC][type]);
 328
 329                 if (link->dumpit == NULL)
 330                         goto err_inval;
 331
 332                 if ((*errp = netlink_dump_start(rtnl, skb, nlh,
 333                                                 link->dumpit,
 334                                                 rtnetlink_done)) != 0) {
 335                         return -1;
 336                 }
 337                 rlen = NLMSG_ALIGN(nlh->nlmsg_len);
 338                 if (rlen > skb->len)
 339                         rlen = skb->len;
 340                 skb_pull(skb, rlen);
 341                 return -1;
 342         }
 343
 344         if (kind != 2) {
 345                 if (rtnl_exlock_nowait()) {
 346                         *errp = 0;
 347                         return -1;
 348                 }
 349                 exclusive = 1;
 350         }
 351
 352         memset(&rta, 0, sizeof(rta));
 353
 354         min_len = rtm_min[sz_idx];
 355         if (nlh->nlmsg_len < min_len)
 356                 goto err_inval;
 357
 358         if (nlh->nlmsg_len > min_len) {
 359                 int attrlen = nlh->nlmsg_len - NLMSG_ALIGN(min_len);
 360                 struct rtattr *attr = (void*)nlh + NLMSG_ALIGN(min_len);
 361
 362                 while (RTA_OK(attr, attrlen)) {
 363                         unsigned flavor = attr->rta_type;
 364                         if (flavor) {
 365                                 if (flavor > rta_max[sz_idx])
 366                                         goto err_inval;
 367                                 rta[flavor-1] = attr;
 368                         }
 369                         attr = RTA_NEXT(attr, attrlen);
 370                 }
 371         }
 372
 373         if (link->doit == NULL)
 374                 link = &(rtnetlink_links[PF_UNSPEC][type]);
 375         if (link->doit == NULL)
 376                 goto err_inval;
 377         err = link->doit(skb, nlh, (void *)&rta);
 378
 379         if (exclusive)
 380                 rtnl_exunlock();
 381         *errp = err;
 382         return err;
 383
 384 err_inval:
 385         if (exclusive)
 386                 rtnl_exunlock();
 387         *errp = -EINVAL;
 388         return -1;
 389 }
 390
 391 /*
 392  * Process one packet of messages.
 393  * Malformed skbs with wrong lengths of messages are discarded silently.
 394  */
 395
 396 extern __inline__ int rtnetlink_rcv_skb(struct sk_buff *skb)
 397 {
 398         int err;
 399         struct nlmsghdr * nlh;
 400
 401         while (skb->len >= NLMSG_SPACE(0)) {
 402                 u32 rlen;
 403
 404                 nlh = (struct nlmsghdr *)skb->data;
 405                 if (nlh->nlmsg_len < sizeof(*nlh) || skb->len < nlh->nlmsg_len)
 406                         return 0;
 407                 rlen = NLMSG_ALIGN(nlh->nlmsg_len);
 408                 if (rlen > skb->len)
 409                         rlen = skb->len;
 410                 if (rtnetlink_rcv_msg(skb, nlh, &err)) {
 411                         /* Not error, but we must interrupt processing here:
 412                          *   Note, that in this case we do not pull message
 413                          *   from skb, it will be processed later.
 414                          */
 415                         if (err == 0)
 416                                 return -1;
 417                         netlink_ack(skb, nlh, err);
 418                 } else if (nlh->nlmsg_flags&NLM_F_ACK)
 419                         netlink_ack(skb, nlh, 0);
 420                 skb_pull(skb, rlen);
 421         }
 422
 423         return 0;
 424 }
 425
 426 /*
 427  *  rtnetlink input queue processing routine:
 428  *      - try to acquire shared lock. If it is failed, defer processing.
 429  *      - feed skbs to rtnetlink_rcv_skb, until it refuse a message,
 430  *        that will occur, when a dump started and/or acquisition of
 431  *        exclusive lock failed.
 432  */
 433
 434 static void rtnetlink_rcv(struct sock *sk, int len)
 435 {
 436         do {
 437                 struct sk_buff *skb;
 438
 439                 if (rtnl_shlock_nowait())
 440                         return;
 441
 442                 while ((skb = skb_dequeue(&sk->receive_queue)) != NULL) {
 443                         if (rtnetlink_rcv_skb(skb)) {
 444                                 if (skb->len)
 445                                         skb_queue_head(&sk->receive_queue, skb);
 446                                 else
 447                                         kfree_skb(skb);
 448                                 break;
 449                         }
 450                         kfree_skb(skb);
 451                 }
 452
 453                 up(&rtnl_sem);
 454         } while (rtnl && rtnl->receive_queue.qlen);
 455 }
 456
 457 static struct rtnetlink_link link_rtnetlink_table[RTM_MAX-RTM_BASE+1] =
 458 {
 459         { NULL,                 NULL,                   },
 460         { NULL,                 NULL,                   },
 461         { NULL,                 rtnetlink_dump_ifinfo,  },
 462         { NULL,                 NULL,                   },
 463
 464         { NULL,                 NULL,                   },
 465         { NULL,                 NULL,                   },
 466         { NULL,                 rtnetlink_dump_all,     },
 467         { NULL,                 NULL,                   },
 468
 469         { NULL,                 NULL,                   },
 470         { NULL,                 NULL,                   },
 471         { NULL,                 rtnetlink_dump_all,     },
 472         { NULL,                 NULL,                   },
 473
 474         { neigh_add,            NULL,                   },
 475         { neigh_delete,         NULL,                   },
 476         { NULL,                 neigh_dump_info,        },
 477         { NULL,                 NULL,                   },
 478
 479         { NULL,                 NULL,                   },
 480         { NULL,                 NULL,                   },
 481         { NULL,                 NULL,                   },
 482         { NULL,                 NULL,                   },
 483 };
 484
 485
 486 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
 487 {
 488         struct net_device *dev = ptr;
 489         switch (event) {
 490         case NETDEV_UNREGISTER:
 491                 rtmsg_ifinfo(RTM_DELLINK, dev);
 492                 break;
 493         default:
 494                 rtmsg_ifinfo(RTM_NEWLINK, dev);
 495                 break;
 496         }
 497         return NOTIFY_DONE;
 498 }
 499
 500 struct notifier_block rtnetlink_dev_notifier = {
 501         rtnetlink_event,
 502         NULL,
 503         0
 504 };
 505
 506
 507 void __init rtnetlink_init(void)
 508 {
 509 #ifdef RTNL_DEBUG
 510         printk("Initializing RT netlink socket\n");
 511 #endif
 512         rtnl = netlink_kernel_create(NETLINK_ROUTE, rtnetlink_rcv);
 513         if (rtnl == NULL)
 514                 panic("rtnetlink_init: cannot initialize rtnetlink\n");
 515         register_netdevice_notifier(&rtnetlink_dev_notifier);
 516         rtnetlink_links[PF_UNSPEC] = link_rtnetlink_table;
 517         rtnetlink_links[PF_PACKET] = link_rtnetlink_table;
 518 }
 519
 520
 521
 522 #endif