linux/net/ipv4/fib_frontend.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 *              IPv4 Forwarding Information Base: FIB frontend.
   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
  16#include <linux/module.h>
  17#include <asm/uaccess.h>
  18#include <linux/bitops.h>
  19#include <linux/capability.h>
  20#include <linux/types.h>
  21#include <linux/kernel.h>
  22#include <linux/mm.h>
  23#include <linux/string.h>
  24#include <linux/socket.h>
  25#include <linux/sockios.h>
  26#include <linux/errno.h>
  27#include <linux/in.h>
  28#include <linux/inet.h>
  29#include <linux/inetdevice.h>
  30#include <linux/netdevice.h>
  31#include <linux/if_addr.h>
  32#include <linux/if_arp.h>
  33#include <linux/skbuff.h>
  34#include <linux/cache.h>
  35#include <linux/init.h>
  36#include <linux/list.h>
  37#include <linux/slab.h>
  38
  39#include <net/ip.h>
  40#include <net/protocol.h>
  41#include <net/route.h>
  42#include <net/tcp.h>
  43#include <net/sock.h>
  44#include <net/arp.h>
  45#include <net/ip_fib.h>
  46#include <net/rtnetlink.h>
  47#include <net/xfrm.h>
  48
  49#ifndef CONFIG_IP_MULTIPLE_TABLES
  50
  51static int __net_init fib4_rules_init(struct net *net)
  52{
  53        struct fib_table *local_table, *main_table;
  54
  55        local_table = fib_trie_table(RT_TABLE_LOCAL);
  56        if (local_table == NULL)
  57                return -ENOMEM;
  58
  59        main_table  = fib_trie_table(RT_TABLE_MAIN);
  60        if (main_table == NULL)
  61                goto fail;
  62
  63        hlist_add_head_rcu(&local_table->tb_hlist,
  64                                &net->ipv4.fib_table_hash[TABLE_LOCAL_INDEX]);
  65        hlist_add_head_rcu(&main_table->tb_hlist,
  66                                &net->ipv4.fib_table_hash[TABLE_MAIN_INDEX]);
  67        return 0;
  68
  69fail:
  70        kfree(local_table);
  71        return -ENOMEM;
  72}
  73#else
  74
  75struct fib_table *fib_new_table(struct net *net, u32 id)
  76{
  77        struct fib_table *tb;
  78        unsigned int h;
  79
  80        if (id == 0)
  81                id = RT_TABLE_MAIN;
  82        tb = fib_get_table(net, id);
  83        if (tb)
  84                return tb;
  85
  86        tb = fib_trie_table(id);
  87        if (!tb)
  88                return NULL;
  89
  90        switch (id) {
  91        case RT_TABLE_LOCAL:
  92                net->ipv4.fib_local = tb;
  93                break;
  94
  95        case RT_TABLE_MAIN:
  96                net->ipv4.fib_main = tb;
  97                break;
  98
  99        case RT_TABLE_DEFAULT:
 100                net->ipv4.fib_default = tb;
 101                break;
 102
 103        default:
 104                break;
 105        }
 106
 107        h = id & (FIB_TABLE_HASHSZ - 1);
 108        hlist_add_head_rcu(&tb->tb_hlist, &net->ipv4.fib_table_hash[h]);
 109        return tb;
 110}
 111
 112struct fib_table *fib_get_table(struct net *net, u32 id)
 113{
 114        struct fib_table *tb;
 115        struct hlist_node *node;
 116        struct hlist_head *head;
 117        unsigned int h;
 118
 119        if (id == 0)
 120                id = RT_TABLE_MAIN;
 121        h = id & (FIB_TABLE_HASHSZ - 1);
 122
 123        rcu_read_lock();
 124        head = &net->ipv4.fib_table_hash[h];
 125        hlist_for_each_entry_rcu(tb, node, head, tb_hlist) {
 126                if (tb->tb_id == id) {
 127                        rcu_read_unlock();
 128                        return tb;
 129                }
 130        }
 131        rcu_read_unlock();
 132        return NULL;
 133}
 134#endif /* CONFIG_IP_MULTIPLE_TABLES */
 135
 136static void fib_flush(struct net *net)
 137{
 138        int flushed = 0;
 139        struct fib_table *tb;
 140        struct hlist_node *node;
 141        struct hlist_head *head;
 142        unsigned int h;
 143
 144        for (h = 0; h < FIB_TABLE_HASHSZ; h++) {
 145                head = &net->ipv4.fib_table_hash[h];
 146                hlist_for_each_entry(tb, node, head, tb_hlist)
 147                        flushed += fib_table_flush(tb);
 148        }
 149
 150        if (flushed)
 151                rt_cache_flush(net);
 152}
 153
 154/*
 155 * Find address type as if only "dev" was present in the system. If
 156 * on_dev is NULL then all interfaces are taken into consideration.
 157 */
 158static inline unsigned int __inet_dev_addr_type(struct net *net,
 159                                                const struct net_device *dev,
 160                                                __be32 addr)
 161{
 162        struct flowi4           fl4 = { .daddr = addr };
 163        struct fib_result       res;
 164        unsigned int ret = RTN_BROADCAST;
 165        struct fib_table *local_table;
 166
 167        if (ipv4_is_zeronet(addr) || ipv4_is_lbcast(addr))
 168                return RTN_BROADCAST;
 169        if (ipv4_is_multicast(addr))
 170                return RTN_MULTICAST;
 171
 172        local_table = fib_get_table(net, RT_TABLE_LOCAL);
 173        if (local_table) {
 174                ret = RTN_UNICAST;
 175                rcu_read_lock();
 176                if (!fib_table_lookup(local_table, &fl4, &res, FIB_LOOKUP_NOREF)) {
 177                        if (!dev || dev == res.fi->fib_dev)
 178                                ret = res.type;
 179                }
 180                rcu_read_unlock();
 181        }
 182        return ret;
 183}
 184
 185unsigned int inet_addr_type(struct net *net, __be32 addr)
 186{
 187        return __inet_dev_addr_type(net, NULL, addr);
 188}
 189EXPORT_SYMBOL(inet_addr_type);
 190
 191unsigned int inet_dev_addr_type(struct net *net, const struct net_device *dev,
 192                                __be32 addr)
 193{
 194        return __inet_dev_addr_type(net, dev, addr);
 195}
 196EXPORT_SYMBOL(inet_dev_addr_type);
 197
 198__be32 fib_compute_spec_dst(struct sk_buff *skb)
 199{
 200        struct net_device *dev = skb->dev;
 201        struct in_device *in_dev;
 202        struct fib_result res;
 203        struct rtable *rt;
 204        struct flowi4 fl4;
 205        struct net *net;
 206        int scope;
 207
 208        rt = skb_rtable(skb);
 209        if ((rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST | RTCF_LOCAL)) ==
 210            RTCF_LOCAL)
 211                return ip_hdr(skb)->daddr;
 212
 213        in_dev = __in_dev_get_rcu(dev);
 214        BUG_ON(!in_dev);
 215
 216        net = dev_net(dev);
 217
 218        scope = RT_SCOPE_UNIVERSE;
 219        if (!ipv4_is_zeronet(ip_hdr(skb)->saddr)) {
 220                fl4.flowi4_oif = 0;
 221                fl4.flowi4_iif = net->loopback_dev->ifindex;
 222                fl4.daddr = ip_hdr(skb)->saddr;
 223                fl4.saddr = 0;
 224                fl4.flowi4_tos = RT_TOS(ip_hdr(skb)->tos);
 225                fl4.flowi4_scope = scope;
 226                fl4.flowi4_mark = IN_DEV_SRC_VMARK(in_dev) ? skb->mark : 0;
 227                if (!fib_lookup(net, &fl4, &res))
 228                        return FIB_RES_PREFSRC(net, res);
 229        } else {
 230                scope = RT_SCOPE_LINK;
 231        }
 232
 233        return inet_select_addr(dev, ip_hdr(skb)->saddr, scope);
 234}
 235
 236/* Given (packet source, input interface) and optional (dst, oif, tos):
 237 * - (main) check, that source is valid i.e. not broadcast or our local
 238 *   address.
 239 * - figure out what "logical" interface this packet arrived
 240 *   and calculate "specific destination" address.
 241 * - check, that packet arrived from expected physical interface.
 242 * called with rcu_read_lock()
 243 */
 244static int __fib_validate_source(struct sk_buff *skb, __be32 src, __be32 dst,
 245                                 u8 tos, int oif, struct net_device *dev,
 246                                 int rpf, struct in_device *idev, u32 *itag)
 247{
 248        int ret, no_addr, accept_local;
 249        struct fib_result res;
 250        struct flowi4 fl4;
 251        struct net *net;
 252        bool dev_match;
 253
 254        fl4.flowi4_oif = 0;
 255        fl4.flowi4_iif = oif;
 256        fl4.daddr = src;
 257        fl4.saddr = dst;
 258        fl4.flowi4_tos = tos;
 259        fl4.flowi4_scope = RT_SCOPE_UNIVERSE;
 260
 261        no_addr = idev->ifa_list == NULL;
 262
 263        accept_local = IN_DEV_ACCEPT_LOCAL(idev);
 264        fl4.flowi4_mark = IN_DEV_SRC_VMARK(idev) ? skb->mark : 0;
 265
 266        net = dev_net(dev);
 267        if (fib_lookup(net, &fl4, &res))
 268                goto last_resort;
 269        if (res.type != RTN_UNICAST) {
 270                if (res.type != RTN_LOCAL || !accept_local)
 271                        goto e_inval;
 272        }
 273        fib_combine_itag(itag, &res);
 274        dev_match = false;
 275
 276#ifdef CONFIG_IP_ROUTE_MULTIPATH
 277        for (ret = 0; ret < res.fi->fib_nhs; ret++) {
 278                struct fib_nh *nh = &res.fi->fib_nh[ret];
 279
 280                if (nh->nh_dev == dev) {
 281                        dev_match = true;
 282                        break;
 283                }
 284        }
 285#else
 286        if (FIB_RES_DEV(res) == dev)
 287                dev_match = true;
 288#endif
 289        if (dev_match) {
 290                ret = FIB_RES_NH(res).nh_scope >= RT_SCOPE_HOST;
 291                return ret;
 292        }
 293        if (no_addr)
 294                goto last_resort;
 295        if (rpf == 1)
 296                goto e_rpf;
 297        fl4.flowi4_oif = dev->ifindex;
 298
 299        ret = 0;
 300        if (fib_lookup(net, &fl4, &res) == 0) {
 301                if (res.type == RTN_UNICAST)
 302                        ret = FIB_RES_NH(res).nh_scope >= RT_SCOPE_HOST;
 303        }
 304        return ret;
 305
 306last_resort:
 307        if (rpf)
 308                goto e_rpf;
 309        *itag = 0;
 310        return 0;
 311
 312e_inval:
 313        return -EINVAL;
 314e_rpf:
 315        return -EXDEV;
 316}
 317
 318/* Ignore rp_filter for packets protected by IPsec. */
 319int fib_validate_source(struct sk_buff *skb, __be32 src, __be32 dst,
 320                        u8 tos, int oif, struct net_device *dev,
 321                        struct in_device *idev, u32 *itag)
 322{
 323        int r = secpath_exists(skb) ? 0 : IN_DEV_RPFILTER(idev);
 324
 325        if (!r && !fib_num_tclassid_users(dev_net(dev)) &&
 326            (dev->ifindex != oif || !IN_DEV_TX_REDIRECTS(idev))) {
 327                *itag = 0;
 328                return 0;
 329        }
 330        return __fib_validate_source(skb, src, dst, tos, oif, dev, r, idev, itag);
 331}
 332
 333static inline __be32 sk_extract_addr(struct sockaddr *addr)
 334{
 335        return ((struct sockaddr_in *) addr)->sin_addr.s_addr;
 336}
 337
 338static int put_rtax(struct nlattr *mx, int len, int type, u32 value)
 339{
 340        struct nlattr *nla;
 341
 342        nla = (struct nlattr *) ((char *) mx + len);
 343        nla->nla_type = type;
 344        nla->nla_len = nla_attr_size(4);
 345        *(u32 *) nla_data(nla) = value;
 346
 347        return len + nla_total_size(4);
 348}
 349
 350static int rtentry_to_fib_config(struct net *net, int cmd, struct rtentry *rt,
 351                                 struct fib_config *cfg)
 352{
 353        __be32 addr;
 354        int plen;
 355
 356        memset(cfg, 0, sizeof(*cfg));
 357        cfg->fc_nlinfo.nl_net = net;
 358
 359        if (rt->rt_dst.sa_family != AF_INET)
 360                return -EAFNOSUPPORT;
 361
 362        /*
 363         * Check mask for validity:
 364         * a) it must be contiguous.
 365         * b) destination must have all host bits clear.
 366         * c) if application forgot to set correct family (AF_INET),
 367         *    reject request unless it is absolutely clear i.e.
 368         *    both family and mask are zero.
 369         */
 370        plen = 32;
 371        addr = sk_extract_addr(&rt->rt_dst);
 372        if (!(rt->rt_flags & RTF_HOST)) {
 373                __be32 mask = sk_extract_addr(&rt->rt_genmask);
 374
 375                if (rt->rt_genmask.sa_family != AF_INET) {
 376                        if (mask || rt->rt_genmask.sa_family)
 377                                return -EAFNOSUPPORT;
 378                }
 379
 380                if (bad_mask(mask, addr))
 381                        return -EINVAL;
 382
 383                plen = inet_mask_len(mask);
 384        }
 385
 386        cfg->fc_dst_len = plen;
 387        cfg->fc_dst = addr;
 388
 389        if (cmd != SIOCDELRT) {
 390                cfg->fc_nlflags = NLM_F_CREATE;
 391                cfg->fc_protocol = RTPROT_BOOT;
 392        }
 393
 394        if (rt->rt_metric)
 395                cfg->fc_priority = rt->rt_metric - 1;
 396
 397        if (rt->rt_flags & RTF_REJECT) {
 398                cfg->fc_scope = RT_SCOPE_HOST;
 399                cfg->fc_type = RTN_UNREACHABLE;
 400                return 0;
 401        }
 402
 403        cfg->fc_scope = RT_SCOPE_NOWHERE;
 404        cfg->fc_type = RTN_UNICAST;
 405
 406        if (rt->rt_dev) {
 407                char *colon;
 408                struct net_device *dev;
 409                char devname[IFNAMSIZ];
 410
 411                if (copy_from_user(devname, rt->rt_dev, IFNAMSIZ-1))
 412                        return -EFAULT;
 413
 414                devname[IFNAMSIZ-1] = 0;
 415                colon = strchr(devname, ':');
 416                if (colon)
 417                        *colon = 0;
 418                dev = __dev_get_by_name(net, devname);
 419                if (!dev)
 420                        return -ENODEV;
 421                cfg->fc_oif = dev->ifindex;
 422                if (colon) {
 423                        struct in_ifaddr *ifa;
 424                        struct in_device *in_dev = __in_dev_get_rtnl(dev);
 425                        if (!in_dev)
 426                                return -ENODEV;
 427                        *colon = ':';
 428                        for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next)
 429                                if (strcmp(ifa->ifa_label, devname) == 0)
 430                                        break;
 431                        if (ifa == NULL)
 432                                return -ENODEV;
 433                        cfg->fc_prefsrc = ifa->ifa_local;
 434                }
 435        }
 436
 437        addr = sk_extract_addr(&rt->rt_gateway);
 438        if (rt->rt_gateway.sa_family == AF_INET && addr) {
 439                cfg->fc_gw = addr;
 440                if (rt->rt_flags & RTF_GATEWAY &&
 441                    inet_addr_type(net, addr) == RTN_UNICAST)
 442                        cfg->fc_scope = RT_SCOPE_UNIVERSE;
 443        }
 444
 445        if (cmd == SIOCDELRT)
 446                return 0;
 447
 448        if (rt->rt_flags & RTF_GATEWAY && !cfg->fc_gw)
 449                return -EINVAL;
 450
 451        if (cfg->fc_scope == RT_SCOPE_NOWHERE)
 452                cfg->fc_scope = RT_SCOPE_LINK;
 453
 454        if (rt->rt_flags & (RTF_MTU | RTF_WINDOW | RTF_IRTT)) {
 455                struct nlattr *mx;
 456                int len = 0;
 457
 458                mx = kzalloc(3 * nla_total_size(4), GFP_KERNEL);
 459                if (mx == NULL)
 460                        return -ENOMEM;
 461
 462                if (rt->rt_flags & RTF_MTU)
 463                        len = put_rtax(mx, len, RTAX_ADVMSS, rt->rt_mtu - 40);
 464
 465                if (rt->rt_flags & RTF_WINDOW)
 466                        len = put_rtax(mx, len, RTAX_WINDOW, rt->rt_window);
 467
 468                if (rt->rt_flags & RTF_IRTT)
 469                        len = put_rtax(mx, len, RTAX_RTT, rt->rt_irtt << 3);
 470
 471                cfg->fc_mx = mx;
 472                cfg->fc_mx_len = len;
 473        }
 474
 475        return 0;
 476}
 477
 478/*
 479 * Handle IP routing ioctl calls.
 480 * These are used to manipulate the routing tables
 481 */
 482int ip_rt_ioctl(struct net *net, unsigned int cmd, void __user *arg)
 483{
 484        struct fib_config cfg;
 485        struct rtentry rt;
 486        int err;
 487
 488        switch (cmd) {
 489        case SIOCADDRT:         /* Add a route */
 490        case SIOCDELRT:         /* Delete a route */
 491                if (!capable(CAP_NET_ADMIN))
 492                        return -EPERM;
 493
 494                if (copy_from_user(&rt, arg, sizeof(rt)))
 495                        return -EFAULT;
 496
 497                rtnl_lock();
 498                err = rtentry_to_fib_config(net, cmd, &rt, &cfg);
 499                if (err == 0) {
 500                        struct fib_table *tb;
 501
 502                        if (cmd == SIOCDELRT) {
 503                                tb = fib_get_table(net, cfg.fc_table);
 504                                if (tb)
 505                                        err = fib_table_delete(tb, &cfg);
 506                                else
 507                                        err = -ESRCH;
 508                        } else {
 509                                tb = fib_new_table(net, cfg.fc_table);
 510                                if (tb)
 511                                        err = fib_table_insert(tb, &cfg);
 512                                else
 513                                        err = -ENOBUFS;
 514                        }
 515
 516                        /* allocated by rtentry_to_fib_config() */
 517                        kfree(cfg.fc_mx);
 518                }
 519                rtnl_unlock();
 520                return err;
 521        }
 522        return -EINVAL;
 523}
 524
 525const struct nla_policy rtm_ipv4_policy[RTA_MAX + 1] = {
 526        [RTA_DST]               = { .type = NLA_U32 },
 527        [RTA_SRC]               = { .type = NLA_U32 },
 528        [RTA_IIF]               = { .type = NLA_U32 },
 529        [RTA_OIF]               = { .type = NLA_U32 },
 530        [RTA_GATEWAY]           = { .type = NLA_U32 },
 531        [RTA_PRIORITY]          = { .type = NLA_U32 },
 532        [RTA_PREFSRC]           = { .type = NLA_U32 },
 533        [RTA_METRICS]           = { .type = NLA_NESTED },
 534        [RTA_MULTIPATH]         = { .len = sizeof(struct rtnexthop) },
 535        [RTA_FLOW]              = { .type = NLA_U32 },
 536};
 537
 538static int rtm_to_fib_config(struct net *net, struct sk_buff *skb,
 539                             struct nlmsghdr *nlh, struct fib_config *cfg)
 540{
 541        struct nlattr *attr;
 542        int err, remaining;
 543        struct rtmsg *rtm;
 544
 545        err = nlmsg_validate(nlh, sizeof(*rtm), RTA_MAX, rtm_ipv4_policy);
 546        if (err < 0)
 547                goto errout;
 548
 549        memset(cfg, 0, sizeof(*cfg));
 550
 551        rtm = nlmsg_data(nlh);
 552        cfg->fc_dst_len = rtm->rtm_dst_len;
 553        cfg->fc_tos = rtm->rtm_tos;
 554        cfg->fc_table = rtm->rtm_table;
 555        cfg->fc_protocol = rtm->rtm_protocol;
 556        cfg->fc_scope = rtm->rtm_scope;
 557        cfg->fc_type = rtm->rtm_type;
 558        cfg->fc_flags = rtm->rtm_flags;
 559        cfg->fc_nlflags = nlh->nlmsg_flags;
 560
 561        cfg->fc_nlinfo.pid = NETLINK_CB(skb).pid;
 562        cfg->fc_nlinfo.nlh = nlh;
 563        cfg->fc_nlinfo.nl_net = net;
 564
 565        if (cfg->fc_type > RTN_MAX) {
 566                err = -EINVAL;
 567                goto errout;
 568        }
 569
 570        nlmsg_for_each_attr(attr, nlh, sizeof(struct rtmsg), remaining) {
 571                switch (nla_type(attr)) {
 572                case RTA_DST:
 573                        cfg->fc_dst = nla_get_be32(attr);
 574                        break;
 575                case RTA_OIF:
 576                        cfg->fc_oif = nla_get_u32(attr);
 577                        break;
 578                case RTA_GATEWAY:
 579                        cfg->fc_gw = nla_get_be32(attr);
 580                        break;
 581                case RTA_PRIORITY:
 582                        cfg->fc_priority = nla_get_u32(attr);
 583                        break;
 584                case RTA_PREFSRC:
 585                        cfg->fc_prefsrc = nla_get_be32(attr);
 586                        break;
 587                case RTA_METRICS:
 588                        cfg->fc_mx = nla_data(attr);
 589                        cfg->fc_mx_len = nla_len(attr);
 590                        break;
 591                case RTA_MULTIPATH:
 592                        cfg->fc_mp = nla_data(attr);
 593                        cfg->fc_mp_len = nla_len(attr);
 594                        break;
 595                case RTA_FLOW:
 596                        cfg->fc_flow = nla_get_u32(attr);
 597                        break;
 598                case RTA_TABLE:
 599                        cfg->fc_table = nla_get_u32(attr);
 600                        break;
 601                }
 602        }
 603
 604        return 0;
 605errout:
 606        return err;
 607}
 608
 609static int inet_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
 610{
 611        struct net *net = sock_net(skb->sk);
 612        struct fib_config cfg;
 613        struct fib_table *tb;
 614        int err;
 615
 616        err = rtm_to_fib_config(net, skb, nlh, &cfg);
 617        if (err < 0)
 618                goto errout;
 619
 620        tb = fib_get_table(net, cfg.fc_table);
 621        if (tb == NULL) {
 622                err = -ESRCH;
 623                goto errout;
 624        }
 625
 626        err = fib_table_delete(tb, &cfg);
 627errout:
 628        return err;
 629}
 630
 631static int inet_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
 632{
 633        struct net *net = sock_net(skb->sk);
 634        struct fib_config cfg;
 635        struct fib_table *tb;
 636        int err;
 637
 638        err = rtm_to_fib_config(net, skb, nlh, &cfg);
 639        if (err < 0)
 640                goto errout;
 641
 642        tb = fib_new_table(net, cfg.fc_table);
 643        if (tb == NULL) {
 644                err = -ENOBUFS;
 645                goto errout;
 646        }
 647
 648        err = fib_table_insert(tb, &cfg);
 649errout:
 650        return err;
 651}
 652
 653static int inet_dump_fib(struct sk_buff *skb, struct netlink_callback *cb)
 654{
 655        struct net *net = sock_net(skb->sk);
 656        unsigned int h, s_h;
 657        unsigned int e = 0, s_e;
 658        struct fib_table *tb;
 659        struct hlist_node *node;
 660        struct hlist_head *head;
 661        int dumped = 0;
 662
 663        if (nlmsg_len(cb->nlh) >= sizeof(struct rtmsg) &&
 664            ((struct rtmsg *) nlmsg_data(cb->nlh))->rtm_flags & RTM_F_CLONED)
 665                return ip_rt_dump(skb, cb);
 666
 667        s_h = cb->args[0];
 668        s_e = cb->args[1];
 669
 670        for (h = s_h; h < FIB_TABLE_HASHSZ; h++, s_e = 0) {
 671                e = 0;
 672                head = &net->ipv4.fib_table_hash[h];
 673                hlist_for_each_entry(tb, node, head, tb_hlist) {
 674                        if (e < s_e)
 675                                goto next;
 676                        if (dumped)
 677                                memset(&cb->args[2], 0, sizeof(cb->args) -
 678                                                 2 * sizeof(cb->args[0]));
 679                        if (fib_table_dump(tb, skb, cb) < 0)
 680                                goto out;
 681                        dumped = 1;
 682next:
 683                        e++;
 684                }
 685        }
 686out:
 687        cb->args[1] = e;
 688        cb->args[0] = h;
 689
 690        return skb->len;
 691}
 692
 693/* Prepare and feed intra-kernel routing request.
 694 * Really, it should be netlink message, but :-( netlink
 695 * can be not configured, so that we feed it directly
 696 * to fib engine. It is legal, because all events occur
 697 * only when netlink is already locked.
 698 */
 699static void fib_magic(int cmd, int type, __be32 dst, int dst_len, struct in_ifaddr *ifa)
 700{
 701        struct net *net = dev_net(ifa->ifa_dev->dev);
 702        struct fib_table *tb;
 703        struct fib_config cfg = {
 704                .fc_protocol = RTPROT_KERNEL,
 705                .fc_type = type,
 706                .fc_dst = dst,
 707                .fc_dst_len = dst_len,
 708                .fc_prefsrc = ifa->ifa_local,
 709                .fc_oif = ifa->ifa_dev->dev->ifindex,
 710                .fc_nlflags = NLM_F_CREATE | NLM_F_APPEND,
 711                .fc_nlinfo = {
 712                        .nl_net = net,
 713                },
 714        };
 715
 716        if (type == RTN_UNICAST)
 717                tb = fib_new_table(net, RT_TABLE_MAIN);
 718        else
 719                tb = fib_new_table(net, RT_TABLE_LOCAL);
 720
 721        if (tb == NULL)
 722                return;
 723
 724        cfg.fc_table = tb->tb_id;
 725
 726        if (type != RTN_LOCAL)
 727                cfg.fc_scope = RT_SCOPE_LINK;
 728        else
 729                cfg.fc_scope = RT_SCOPE_HOST;
 730
 731        if (cmd == RTM_NEWROUTE)
 732                fib_table_insert(tb, &cfg);
 733        else
 734                fib_table_delete(tb, &cfg);
 735}
 736
 737void fib_add_ifaddr(struct in_ifaddr *ifa)
 738{
 739        struct in_device *in_dev = ifa->ifa_dev;
 740        struct net_device *dev = in_dev->dev;
 741        struct in_ifaddr *prim = ifa;
 742        __be32 mask = ifa->ifa_mask;
 743        __be32 addr = ifa->ifa_local;
 744        __be32 prefix = ifa->ifa_address & mask;
 745
 746        if (ifa->ifa_flags & IFA_F_SECONDARY) {
 747                prim = inet_ifa_byprefix(in_dev, prefix, mask);
 748                if (prim == NULL) {
 749                        pr_warn("%s: bug: prim == NULL\n", __func__);
 750                        return;
 751                }
 752        }
 753
 754        fib_magic(RTM_NEWROUTE, RTN_LOCAL, addr, 32, prim);
 755
 756        if (!(dev->flags & IFF_UP))
 757                return;
 758
 759        /* Add broadcast address, if it is explicitly assigned. */
 760        if (ifa->ifa_broadcast && ifa->ifa_broadcast != htonl(0xFFFFFFFF))
 761                fib_magic(RTM_NEWROUTE, RTN_BROADCAST, ifa->ifa_broadcast, 32, prim);
 762
 763        if (!ipv4_is_zeronet(prefix) && !(ifa->ifa_flags & IFA_F_SECONDARY) &&
 764            (prefix != addr || ifa->ifa_prefixlen < 32)) {
 765                fib_magic(RTM_NEWROUTE,
 766                          dev->flags & IFF_LOOPBACK ? RTN_LOCAL : RTN_UNICAST,
 767                          prefix, ifa->ifa_prefixlen, prim);
 768
 769                /* Add network specific broadcasts, when it takes a sense */
 770                if (ifa->ifa_prefixlen < 31) {
 771                        fib_magic(RTM_NEWROUTE, RTN_BROADCAST, prefix, 32, prim);
 772                        fib_magic(RTM_NEWROUTE, RTN_BROADCAST, prefix | ~mask,
 773                                  32, prim);
 774                }
 775        }
 776}
 777
 778/* Delete primary or secondary address.
 779 * Optionally, on secondary address promotion consider the addresses
 780 * from subnet iprim as deleted, even if they are in device list.
 781 * In this case the secondary ifa can be in device list.
 782 */
 783void fib_del_ifaddr(struct in_ifaddr *ifa, struct in_ifaddr *iprim)
 784{
 785        struct in_device *in_dev = ifa->ifa_dev;
 786        struct net_device *dev = in_dev->dev;
 787        struct in_ifaddr *ifa1;
 788        struct in_ifaddr *prim = ifa, *prim1 = NULL;
 789        __be32 brd = ifa->ifa_address | ~ifa->ifa_mask;
 790        __be32 any = ifa->ifa_address & ifa->ifa_mask;
 791#define LOCAL_OK        1
 792#define BRD_OK          2
 793#define BRD0_OK         4
 794#define BRD1_OK         8
 795        unsigned int ok = 0;
 796        int subnet = 0;         /* Primary network */
 797        int gone = 1;           /* Address is missing */
 798        int same_prefsrc = 0;   /* Another primary with same IP */
 799
 800        if (ifa->ifa_flags & IFA_F_SECONDARY) {
 801                prim = inet_ifa_byprefix(in_dev, any, ifa->ifa_mask);
 802                if (prim == NULL) {
 803                        pr_warn("%s: bug: prim == NULL\n", __func__);
 804                        return;
 805                }
 806                if (iprim && iprim != prim) {
 807                        pr_warn("%s: bug: iprim != prim\n", __func__);
 808                        return;
 809                }
 810        } else if (!ipv4_is_zeronet(any) &&
 811                   (any != ifa->ifa_local || ifa->ifa_prefixlen < 32)) {
 812                fib_magic(RTM_DELROUTE,
 813                          dev->flags & IFF_LOOPBACK ? RTN_LOCAL : RTN_UNICAST,
 814                          any, ifa->ifa_prefixlen, prim);
 815                subnet = 1;
 816        }
 817
 818        /* Deletion is more complicated than add.
 819         * We should take care of not to delete too much :-)
 820         *
 821         * Scan address list to be sure that addresses are really gone.
 822         */
 823
 824        for (ifa1 = in_dev->ifa_list; ifa1; ifa1 = ifa1->ifa_next) {
 825                if (ifa1 == ifa) {
 826                        /* promotion, keep the IP */
 827                        gone = 0;
 828                        continue;
 829                }
 830                /* Ignore IFAs from our subnet */
 831                if (iprim && ifa1->ifa_mask == iprim->ifa_mask &&
 832                    inet_ifa_match(ifa1->ifa_address, iprim))
 833                        continue;
 834
 835                /* Ignore ifa1 if it uses different primary IP (prefsrc) */
 836                if (ifa1->ifa_flags & IFA_F_SECONDARY) {
 837                        /* Another address from our subnet? */
 838                        if (ifa1->ifa_mask == prim->ifa_mask &&
 839                            inet_ifa_match(ifa1->ifa_address, prim))
 840                                prim1 = prim;
 841                        else {
 842                                /* We reached the secondaries, so
 843                                 * same_prefsrc should be determined.
 844                                 */
 845                                if (!same_prefsrc)
 846                                        continue;
 847                                /* Search new prim1 if ifa1 is not
 848                                 * using the current prim1
 849                                 */
 850                                if (!prim1 ||
 851                                    ifa1->ifa_mask != prim1->ifa_mask ||
 852                                    !inet_ifa_match(ifa1->ifa_address, prim1))
 853                                        prim1 = inet_ifa_byprefix(in_dev,
 854                                                        ifa1->ifa_address,
 855                                                        ifa1->ifa_mask);
 856                                if (!prim1)
 857                                        continue;
 858                                if (prim1->ifa_local != prim->ifa_local)
 859                                        continue;
 860                        }
 861                } else {
 862                        if (prim->ifa_local != ifa1->ifa_local)
 863                                continue;
 864                        prim1 = ifa1;
 865                        if (prim != prim1)
 866                                same_prefsrc = 1;
 867                }
 868                if (ifa->ifa_local == ifa1->ifa_local)
 869                        ok |= LOCAL_OK;
 870                if (ifa->ifa_broadcast == ifa1->ifa_broadcast)
 871                        ok |= BRD_OK;
 872                if (brd == ifa1->ifa_broadcast)
 873                        ok |= BRD1_OK;
 874                if (any == ifa1->ifa_broadcast)
 875                        ok |= BRD0_OK;
 876                /* primary has network specific broadcasts */
 877                if (prim1 == ifa1 && ifa1->ifa_prefixlen < 31) {
 878                        __be32 brd1 = ifa1->ifa_address | ~ifa1->ifa_mask;
 879                        __be32 any1 = ifa1->ifa_address & ifa1->ifa_mask;
 880
 881                        if (!ipv4_is_zeronet(any1)) {
 882                                if (ifa->ifa_broadcast == brd1 ||
 883                                    ifa->ifa_broadcast == any1)
 884                                        ok |= BRD_OK;
 885                                if (brd == brd1 || brd == any1)
 886                                        ok |= BRD1_OK;
 887                                if (any == brd1 || any == any1)
 888                                        ok |= BRD0_OK;
 889                        }
 890                }
 891        }
 892
 893        if (!(ok & BRD_OK))
 894                fib_magic(RTM_DELROUTE, RTN_BROADCAST, ifa->ifa_broadcast, 32, prim);
 895        if (subnet && ifa->ifa_prefixlen < 31) {
 896                if (!(ok & BRD1_OK))
 897                        fib_magic(RTM_DELROUTE, RTN_BROADCAST, brd, 32, prim);
 898                if (!(ok & BRD0_OK))
 899                        fib_magic(RTM_DELROUTE, RTN_BROADCAST, any, 32, prim);
 900        }
 901        if (!(ok & LOCAL_OK)) {
 902                fib_magic(RTM_DELROUTE, RTN_LOCAL, ifa->ifa_local, 32, prim);
 903
 904                /* Check, that this local address finally disappeared. */
 905                if (gone &&
 906                    inet_addr_type(dev_net(dev), ifa->ifa_local) != RTN_LOCAL) {
 907                        /* And the last, but not the least thing.
 908                         * We must flush stray FIB entries.
 909                         *
 910                         * First of all, we scan fib_info list searching
 911                         * for stray nexthop entries, then ignite fib_flush.
 912                         */
 913                        if (fib_sync_down_addr(dev_net(dev), ifa->ifa_local))
 914                                fib_flush(dev_net(dev));
 915                }
 916        }
 917#undef LOCAL_OK
 918#undef BRD_OK
 919#undef BRD0_OK
 920#undef BRD1_OK
 921}
 922
 923static void nl_fib_lookup(struct fib_result_nl *frn, struct fib_table *tb)
 924{
 925
 926        struct fib_result       res;
 927        struct flowi4           fl4 = {
 928                .flowi4_mark = frn->fl_mark,
 929                .daddr = frn->fl_addr,
 930                .flowi4_tos = frn->fl_tos,
 931                .flowi4_scope = frn->fl_scope,
 932        };
 933
 934        frn->err = -ENOENT;
 935        if (tb) {
 936                local_bh_disable();
 937
 938                frn->tb_id = tb->tb_id;
 939                rcu_read_lock();
 940                frn->err = fib_table_lookup(tb, &fl4, &res, FIB_LOOKUP_NOREF);
 941
 942                if (!frn->err) {
 943                        frn->prefixlen = res.prefixlen;
 944                        frn->nh_sel = res.nh_sel;
 945                        frn->type = res.type;
 946                        frn->scope = res.scope;
 947                }
 948                rcu_read_unlock();
 949                local_bh_enable();
 950        }
 951}
 952
 953static void nl_fib_input(struct sk_buff *skb)
 954{
 955        struct net *net;
 956        struct fib_result_nl *frn;
 957        struct nlmsghdr *nlh;
 958        struct fib_table *tb;
 959        u32 pid;
 960
 961        net = sock_net(skb->sk);
 962        nlh = nlmsg_hdr(skb);
 963        if (skb->len < NLMSG_SPACE(0) || skb->len < nlh->nlmsg_len ||
 964            nlh->nlmsg_len < NLMSG_LENGTH(sizeof(*frn)))
 965                return;
 966
 967        skb = skb_clone(skb, GFP_KERNEL);
 968        if (skb == NULL)
 969                return;
 970        nlh = nlmsg_hdr(skb);
 971
 972        frn = (struct fib_result_nl *) NLMSG_DATA(nlh);
 973        tb = fib_get_table(net, frn->tb_id_in);
 974
 975        nl_fib_lookup(frn, tb);
 976
 977        pid = NETLINK_CB(skb).pid;      /* pid of sending process */
 978        NETLINK_CB(skb).pid = 0;        /* from kernel */
 979        NETLINK_CB(skb).dst_group = 0;  /* unicast */
 980        netlink_unicast(net->ipv4.fibnl, skb, pid, MSG_DONTWAIT);
 981}
 982
 983static int __net_init nl_fib_lookup_init(struct net *net)
 984{
 985        struct sock *sk;
 986        struct netlink_kernel_cfg cfg = {
 987                .input  = nl_fib_input,
 988        };
 989
 990        sk = netlink_kernel_create(net, NETLINK_FIB_LOOKUP, THIS_MODULE, &cfg);
 991        if (sk == NULL)
 992                return -EAFNOSUPPORT;
 993        net->ipv4.fibnl = sk;
 994        return 0;
 995}
 996
 997static void nl_fib_lookup_exit(struct net *net)
 998{
 999        netlink_kernel_release(net->ipv4.fibnl);
1000        net->ipv4.fibnl = NULL;
1001}
1002
1003static void fib_disable_ip(struct net_device *dev, int force)
1004{
1005        if (fib_sync_down_dev(dev, force))
1006                fib_flush(dev_net(dev));
1007        rt_cache_flush(dev_net(dev));
1008        arp_ifdown(dev);
1009}
1010
1011static int fib_inetaddr_event(struct notifier_block *this, unsigned long event, void *ptr)
1012{
1013        struct in_ifaddr *ifa = (struct in_ifaddr *)ptr;
1014        struct net_device *dev = ifa->ifa_dev->dev;
1015        struct net *net = dev_net(dev);
1016
1017        switch (event) {
1018        case NETDEV_UP:
1019                fib_add_ifaddr(ifa);
1020#ifdef CONFIG_IP_ROUTE_MULTIPATH
1021                fib_sync_up(dev);
1022#endif
1023                atomic_inc(&net->ipv4.dev_addr_genid);
1024                rt_cache_flush(dev_net(dev));
1025                break;
1026        case NETDEV_DOWN:
1027                fib_del_ifaddr(ifa, NULL);
1028                atomic_inc(&net->ipv4.dev_addr_genid);
1029                if (ifa->ifa_dev->ifa_list == NULL) {
1030                        /* Last address was deleted from this interface.
1031                         * Disable IP.
1032                         */
1033                        fib_disable_ip(dev, 1);
1034                } else {
1035                        rt_cache_flush(dev_net(dev));
1036                }
1037                break;
1038        }
1039        return NOTIFY_DONE;
1040}
1041
1042static int fib_netdev_event(struct notifier_block *this, unsigned long event, void *ptr)
1043{
1044        struct net_device *dev = ptr;
1045        struct in_device *in_dev = __in_dev_get_rtnl(dev);
1046        struct net *net = dev_net(dev);
1047
1048        if (event == NETDEV_UNREGISTER) {
1049                fib_disable_ip(dev, 2);
1050                rt_flush_dev(dev);
1051                return NOTIFY_DONE;
1052        }
1053
1054        if (!in_dev)
1055                return NOTIFY_DONE;
1056
1057        switch (event) {
1058        case NETDEV_UP:
1059                for_ifa(in_dev) {
1060                        fib_add_ifaddr(ifa);
1061                } endfor_ifa(in_dev);
1062#ifdef CONFIG_IP_ROUTE_MULTIPATH
1063                fib_sync_up(dev);
1064#endif
1065                atomic_inc(&net->ipv4.dev_addr_genid);
1066                rt_cache_flush(dev_net(dev));
1067                break;
1068        case NETDEV_DOWN:
1069                fib_disable_ip(dev, 0);
1070                break;
1071        case NETDEV_CHANGEMTU:
1072        case NETDEV_CHANGE:
1073                rt_cache_flush(dev_net(dev));
1074                break;
1075        case NETDEV_UNREGISTER_BATCH:
1076                break;
1077        }
1078        return NOTIFY_DONE;
1079}
1080
1081static struct notifier_block fib_inetaddr_notifier = {
1082        .notifier_call = fib_inetaddr_event,
1083};
1084
1085static struct notifier_block fib_netdev_notifier = {
1086        .notifier_call = fib_netdev_event,
1087};
1088
1089static int __net_init ip_fib_net_init(struct net *net)
1090{
1091        int err;
1092        size_t size = sizeof(struct hlist_head) * FIB_TABLE_HASHSZ;
1093
1094        /* Avoid false sharing : Use at least a full cache line */
1095        size = max_t(size_t, size, L1_CACHE_BYTES);
1096
1097        net->ipv4.fib_table_hash = kzalloc(size, GFP_KERNEL);
1098        if (net->ipv4.fib_table_hash == NULL)
1099                return -ENOMEM;
1100
1101        err = fib4_rules_init(net);
1102        if (err < 0)
1103                goto fail;
1104        return 0;
1105
1106fail:
1107        kfree(net->ipv4.fib_table_hash);
1108        return err;
1109}
1110
1111static void ip_fib_net_exit(struct net *net)
1112{
1113        unsigned int i;
1114
1115#ifdef CONFIG_IP_MULTIPLE_TABLES
1116        fib4_rules_exit(net);
1117#endif
1118
1119        rtnl_lock();
1120        for (i = 0; i < FIB_TABLE_HASHSZ; i++) {
1121                struct fib_table *tb;
1122                struct hlist_head *head;
1123                struct hlist_node *node, *tmp;
1124
1125                head = &net->ipv4.fib_table_hash[i];
1126                hlist_for_each_entry_safe(tb, node, tmp, head, tb_hlist) {
1127                        hlist_del(node);
1128                        fib_table_flush(tb);
1129                        fib_free_table(tb);
1130                }
1131        }
1132        rtnl_unlock();
1133        kfree(net->ipv4.fib_table_hash);
1134}
1135
1136static int __net_init fib_net_init(struct net *net)
1137{
1138        int error;
1139
1140#ifdef CONFIG_IP_ROUTE_CLASSID
1141        net->ipv4.fib_num_tclassid_users = 0;
1142#endif
1143        error = ip_fib_net_init(net);
1144        if (error < 0)
1145                goto out;
1146        error = nl_fib_lookup_init(net);
1147        if (error < 0)
1148                goto out_nlfl;
1149        error = fib_proc_init(net);
1150        if (error < 0)
1151                goto out_proc;
1152out:
1153        return error;
1154
1155out_proc:
1156        nl_fib_lookup_exit(net);
1157out_nlfl:
1158        ip_fib_net_exit(net);
1159        goto out;
1160}
1161
1162static void __net_exit fib_net_exit(struct net *net)
1163{
1164        fib_proc_exit(net);
1165        nl_fib_lookup_exit(net);
1166        ip_fib_net_exit(net);
1167}
1168
1169static struct pernet_operations fib_net_ops = {
1170        .init = fib_net_init,
1171        .exit = fib_net_exit,
1172};
1173
1174void __init ip_fib_init(void)
1175{
1176        rtnl_register(PF_INET, RTM_NEWROUTE, inet_rtm_newroute, NULL, NULL);
1177        rtnl_register(PF_INET, RTM_DELROUTE, inet_rtm_delroute, NULL, NULL);
1178        rtnl_register(PF_INET, RTM_GETROUTE, NULL, inet_dump_fib, NULL);
1179
1180        register_pernet_subsys(&fib_net_ops);
1181        register_netdevice_notifier(&fib_netdev_notifier);
1182        register_inetaddr_notifier(&fib_inetaddr_notifier);
1183
1184        fib_trie_init();
1185}
1186
lxr.linux.no kindly hosted by Redpill Linpro AS, provider of Linux consulting and operations services since 1995.