net/ipv4/sysctl_net_ipv4.c

   1 /*
   2  * sysctl_net_ipv4.c: sysctl interface to net IPV4 subsystem.
   3  *
   4  * Begun April 1, 1996, Mike Shaver.
   5  * Added /proc/sys/net/ipv4 directory entry (empty =) ). [MS]
   6  */
   7
   8 #include <linux/mm.h>
   9 #include <linux/module.h>
  10 #include <linux/sysctl.h>
  11 #include <linux/igmp.h>
  12 #include <linux/inetdevice.h>
  13 #include <linux/seqlock.h>
  14 #include <linux/init.h>
  15 #include <linux/slab.h>
  16 #include <linux/nsproxy.h>
  17 #include <linux/swap.h>
  18 #include <net/snmp.h>
  19 #include <net/icmp.h>
  20 #include <net/ip.h>
  21 #include <net/route.h>
  22 #include <net/tcp.h>
  23 #include <net/udp.h>
  24 #include <net/cipso_ipv4.h>
  25 #include <net/inet_frag.h>
  26 #include <net/ping.h>
  27 #include <net/tcp_memcontrol.h>
  28
  29 static int zero;
  30 static int one = 1;
  31 static int four = 4;
  32 static int thousand = 1000;
  33 static int gso_max_segs = GSO_MAX_SEGS;
  34 static int tcp_retr1_max = 255;
  35 static int ip_local_port_range_min[] = { 1, 1 };
  36 static int ip_local_port_range_max[] = { 65535, 65535 };
  37 static int tcp_adv_win_scale_min = -31;
  38 static int tcp_adv_win_scale_max = 31;
  39 static int tcp_min_snd_mss_min = TCP_MIN_SND_MSS;
  40 static int tcp_min_snd_mss_max = 65535;
  41 static int ip_ttl_min = 1;
  42 static int ip_ttl_max = 255;
  43 static int tcp_syn_retries_min = 1;
  44 static int tcp_syn_retries_max = MAX_TCP_SYNCNT;
  45 static int ip_ping_group_range_min[] = { 0, 0 };
  46 static int ip_ping_group_range_max[] = { GID_T_MAX, GID_T_MAX };
  47 static int one_day_secs = 24 * 3600;
  48
  49 /* Update system visible IP port range */
  50 static void set_local_port_range(struct net *net, int range[2])
  51 {
  52         bool same_parity = !((range[0] ^ range[1]) & 1);
  53
  54         write_seqlock_bh(&net->ipv4.ip_local_ports.lock);
  55         if (same_parity && !net->ipv4.ip_local_ports.warned) {
  56                 net->ipv4.ip_local_ports.warned = true;
  57                 pr_err_ratelimited("ip_local_port_range: prefer different parity for start/end values.\n");
  58         }
  59         net->ipv4.ip_local_ports.range[0] = range[0];
  60         net->ipv4.ip_local_ports.range[1] = range[1];
  61         write_sequnlock_bh(&net->ipv4.ip_local_ports.lock);
  62 }
  63
  64 /* Validate changes from /proc interface. */
  65 static int ipv4_local_port_range(struct ctl_table *table, int write,
  66                                  void __user *buffer,
  67                                  size_t *lenp, loff_t *ppos)
  68 {
  69         struct net *net =
  70                 container_of(table->data, struct net, ipv4.ip_local_ports.range);
  71         int ret;
  72         int range[2];
  73         struct ctl_table tmp = {
  74                 .data = &range,
  75                 .maxlen = sizeof(range),
  76                 .mode = table->mode,
  77                 .extra1 = &ip_local_port_range_min,
  78                 .extra2 = &ip_local_port_range_max,
  79         };
  80
  81         inet_get_local_port_range(net, &range[0], &range[1]);
  82
  83         ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
  84
  85         if (write && ret == 0) {
  86                 if (range[1] < range[0])
  87                         ret = -EINVAL;
  88                 else
  89                         set_local_port_range(net, range);
  90         }
  91
  92         return ret;
  93 }
  94
  95
  96 static void inet_get_ping_group_range_table(struct ctl_table *table, kgid_t *low, kgid_t *high)
  97 {
  98         kgid_t *data = table->data;
  99         struct net *net =
 100                 container_of(table->data, struct net, ipv4.ping_group_range.range);
 101         unsigned int seq;
 102         do {
 103                 seq = read_seqbegin(&net->ipv4.ping_group_range.lock);
 104
 105                 *low = data[0];
 106                 *high = data[1];
 107         } while (read_seqretry(&net->ipv4.ping_group_range.lock, seq));
 108 }
 109
 110 /* Update system visible IP port range */
 111 static void set_ping_group_range(struct ctl_table *table, kgid_t low, kgid_t high)
 112 {
 113         kgid_t *data = table->data;
 114         struct net *net =
 115                 container_of(table->data, struct net, ipv4.ping_group_range.range);
 116         write_seqlock(&net->ipv4.ping_group_range.lock);
 117         data[0] = low;
 118         data[1] = high;
 119         write_sequnlock(&net->ipv4.ping_group_range.lock);
 120 }
 121
 122 /* Validate changes from /proc interface. */
 123 static int ipv4_ping_group_range(struct ctl_table *table, int write,
 124                                  void __user *buffer,
 125                                  size_t *lenp, loff_t *ppos)
 126 {
 127         struct user_namespace *user_ns = current_user_ns();
 128         int ret;
 129         gid_t urange[2];
 130         kgid_t low, high;
 131         struct ctl_table tmp = {
 132                 .data = &urange,
 133                 .maxlen = sizeof(urange),
 134                 .mode = table->mode,
 135                 .extra1 = &ip_ping_group_range_min,
 136                 .extra2 = &ip_ping_group_range_max,
 137         };
 138
 139         inet_get_ping_group_range_table(table, &low, &high);
 140         urange[0] = from_kgid_munged(user_ns, low);
 141         urange[1] = from_kgid_munged(user_ns, high);
 142         ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
 143
 144         if (write && ret == 0) {
 145                 low = make_kgid(user_ns, urange[0]);
 146                 high = make_kgid(user_ns, urange[1]);
 147                 if (!gid_valid(low) || !gid_valid(high))
 148                         return -EINVAL;
 149                 if (urange[1] < urange[0] || gid_lt(high, low)) {
 150                         low = make_kgid(&init_user_ns, 1);
 151                         high = make_kgid(&init_user_ns, 0);
 152                 }
 153                 set_ping_group_range(table, low, high);
 154         }
 155
 156         return ret;
 157 }
 158
 159 static int proc_tcp_congestion_control(struct ctl_table *ctl, int write,
 160                                        void __user *buffer, size_t *lenp, loff_t *ppos)
 161 {
 162         char val[TCP_CA_NAME_MAX];
 163         struct ctl_table tbl = {
 164                 .data = val,
 165                 .maxlen = TCP_CA_NAME_MAX,
 166         };
 167         int ret;
 168
 169         tcp_get_default_congestion_control(val);
 170
 171         ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
 172         if (write && ret == 0)
 173                 ret = tcp_set_default_congestion_control(val);
 174         return ret;
 175 }
 176
 177 static int proc_tcp_available_congestion_control(struct ctl_table *ctl,
 178                                                  int write,
 179                                                  void __user *buffer, size_t *lenp,
 180                                                  loff_t *ppos)
 181 {
 182         struct ctl_table tbl = { .maxlen = TCP_CA_BUF_MAX, };
 183         int ret;
 184
 185         tbl.data = kmalloc(tbl.maxlen, GFP_USER);
 186         if (!tbl.data)
 187                 return -ENOMEM;
 188         tcp_get_available_congestion_control(tbl.data, TCP_CA_BUF_MAX);
 189         ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
 190         kfree(tbl.data);
 191         return ret;
 192 }
 193
 194 static int proc_allowed_congestion_control(struct ctl_table *ctl,
 195                                            int write,
 196                                            void __user *buffer, size_t *lenp,
 197                                            loff_t *ppos)
 198 {
 199         struct ctl_table tbl = { .maxlen = TCP_CA_BUF_MAX };
 200         int ret;
 201
 202         tbl.data = kmalloc(tbl.maxlen, GFP_USER);
 203         if (!tbl.data)
 204                 return -ENOMEM;
 205
 206         tcp_get_allowed_congestion_control(tbl.data, tbl.maxlen);
 207         ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
 208         if (write && ret == 0)
 209                 ret = tcp_set_allowed_congestion_control(tbl.data);
 210         kfree(tbl.data);
 211         return ret;
 212 }
 213
 214 static int proc_tcp_fastopen_key(struct ctl_table *ctl, int write,
 215                                  void __user *buffer, size_t *lenp,
 216                                  loff_t *ppos)
 217 {
 218         struct ctl_table tbl = { .maxlen = (TCP_FASTOPEN_KEY_LENGTH * 2 + 10) };
 219         struct tcp_fastopen_context *ctxt;
 220         u32  user_key[4]; /* 16 bytes, matching TCP_FASTOPEN_KEY_LENGTH */
 221         __le32 key[4];
 222         int ret, i;
 223
 224         tbl.data = kmalloc(tbl.maxlen, GFP_KERNEL);
 225         if (!tbl.data)
 226                 return -ENOMEM;
 227
 228         rcu_read_lock();
 229         ctxt = rcu_dereference(tcp_fastopen_ctx);
 230         if (ctxt)
 231                 memcpy(key, ctxt->key, TCP_FASTOPEN_KEY_LENGTH);
 232         else
 233                 memset(key, 0, sizeof(key));
 234         rcu_read_unlock();
 235
 236         for (i = 0; i < ARRAY_SIZE(key); i++)
 237                 user_key[i] = le32_to_cpu(key[i]);
 238
 239         snprintf(tbl.data, tbl.maxlen, "%08x-%08x-%08x-%08x",
 240                 user_key[0], user_key[1], user_key[2], user_key[3]);
 241         ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
 242
 243         if (write && ret == 0) {
 244                 if (sscanf(tbl.data, "%x-%x-%x-%x", user_key, user_key + 1,
 245                            user_key + 2, user_key + 3) != 4) {
 246                         ret = -EINVAL;
 247                         goto bad_key;
 248                 }
 249                 /* Generate a dummy secret but don't publish it. This
 250                  * is needed so we don't regenerate a new key on the
 251                  * first invocation of tcp_fastopen_cookie_gen
 252                  */
 253                 tcp_fastopen_init_key_once(false);
 254
 255                 for (i = 0; i < ARRAY_SIZE(user_key); i++)
 256                         key[i] = cpu_to_le32(user_key[i]);
 257
 258                 tcp_fastopen_reset_cipher(key, TCP_FASTOPEN_KEY_LENGTH);
 259         }
 260
 261 bad_key:
 262         pr_debug("proc FO key set 0x%x-%x-%x-%x <- 0x%s: %u\n",
 263                  user_key[0], user_key[1], user_key[2], user_key[3],
 264                (char *)tbl.data, ret);
 265         kfree(tbl.data);
 266         return ret;
 267 }
 268
 269 static struct ctl_table ipv4_table[] = {
 270         {
 271                 .procname       = "tcp_timestamps",
 272                 .data           = &sysctl_tcp_timestamps,
 273                 .maxlen         = sizeof(int),
 274                 .mode           = 0644,
 275                 .proc_handler   = proc_dointvec
 276         },
 277         {
 278                 .procname       = "tcp_window_scaling",
 279                 .data           = &sysctl_tcp_window_scaling,
 280                 .maxlen         = sizeof(int),
 281                 .mode           = 0644,
 282                 .proc_handler   = proc_dointvec
 283         },
 284         {
 285                 .procname       = "tcp_sack",
 286                 .data           = &sysctl_tcp_sack,
 287                 .maxlen         = sizeof(int),
 288                 .mode           = 0644,
 289                 .proc_handler   = proc_dointvec
 290         },
 291         {
 292                 .procname       = "tcp_retrans_collapse",
 293                 .data           = &sysctl_tcp_retrans_collapse,
 294                 .maxlen         = sizeof(int),
 295                 .mode           = 0644,
 296                 .proc_handler   = proc_dointvec
 297         },
 298         {
 299                 .procname       = "ip_default_ttl",
 300                 .data           = &sysctl_ip_default_ttl,
 301                 .maxlen         = sizeof(int),
 302                 .mode           = 0644,
 303                 .proc_handler   = proc_dointvec_minmax,
 304                 .extra1         = &ip_ttl_min,
 305                 .extra2         = &ip_ttl_max,
 306         },
 307         {
 308                 .procname       = "tcp_syn_retries",
 309                 .data           = &sysctl_tcp_syn_retries,
 310                 .maxlen         = sizeof(int),
 311                 .mode           = 0644,
 312                 .proc_handler   = proc_dointvec_minmax,
 313                 .extra1         = &tcp_syn_retries_min,
 314                 .extra2         = &tcp_syn_retries_max
 315         },
 316         {
 317                 .procname       = "tcp_synack_retries",
 318                 .data           = &sysctl_tcp_synack_retries,
 319                 .maxlen         = sizeof(int),
 320                 .mode           = 0644,
 321                 .proc_handler   = proc_dointvec
 322         },
 323         {
 324                 .procname       = "tcp_max_orphans",
 325                 .data           = &sysctl_tcp_max_orphans,
 326                 .maxlen         = sizeof(int),
 327                 .mode           = 0644,
 328                 .proc_handler   = proc_dointvec
 329         },
 330         {
 331                 .procname       = "tcp_max_tw_buckets",
 332                 .data           = &tcp_death_row.sysctl_max_tw_buckets,
 333                 .maxlen         = sizeof(int),
 334                 .mode           = 0644,
 335                 .proc_handler   = proc_dointvec
 336         },
 337         {
 338                 .procname       = "ip_early_demux",
 339                 .data           = &sysctl_ip_early_demux,
 340                 .maxlen         = sizeof(int),
 341                 .mode           = 0644,
 342                 .proc_handler   = proc_dointvec
 343         },
 344         {
 345                 .procname       = "ip_dynaddr",
 346                 .data           = &sysctl_ip_dynaddr,
 347                 .maxlen         = sizeof(int),
 348                 .mode           = 0644,
 349                 .proc_handler   = proc_dointvec
 350         },
 351         {
 352                 .procname       = "tcp_keepalive_time",
 353                 .data           = &sysctl_tcp_keepalive_time,
 354                 .maxlen         = sizeof(int),
 355                 .mode           = 0644,
 356                 .proc_handler   = proc_dointvec_jiffies,
 357         },
 358         {
 359                 .procname       = "tcp_keepalive_probes",
 360                 .data           = &sysctl_tcp_keepalive_probes,
 361                 .maxlen         = sizeof(int),
 362                 .mode           = 0644,
 363                 .proc_handler   = proc_dointvec
 364         },
 365         {
 366                 .procname       = "tcp_keepalive_intvl",
 367                 .data           = &sysctl_tcp_keepalive_intvl,
 368                 .maxlen         = sizeof(int),
 369                 .mode           = 0644,
 370                 .proc_handler   = proc_dointvec_jiffies,
 371         },
 372         {
 373                 .procname       = "tcp_retries1",
 374                 .data           = &sysctl_tcp_retries1,
 375                 .maxlen         = sizeof(int),
 376                 .mode           = 0644,
 377                 .proc_handler   = proc_dointvec_minmax,
 378                 .extra2         = &tcp_retr1_max
 379         },
 380         {
 381                 .procname       = "tcp_retries2",
 382                 .data           = &sysctl_tcp_retries2,
 383                 .maxlen         = sizeof(int),
 384                 .mode           = 0644,
 385                 .proc_handler   = proc_dointvec
 386         },
 387         {
 388                 .procname       = "tcp_fin_timeout",
 389                 .data           = &sysctl_tcp_fin_timeout,
 390                 .maxlen         = sizeof(int),
 391                 .mode           = 0644,
 392                 .proc_handler   = proc_dointvec_jiffies,
 393         },
 394 #ifdef CONFIG_SYN_COOKIES
 395         {
 396                 .procname       = "tcp_syncookies",
 397                 .data           = &sysctl_tcp_syncookies,
 398                 .maxlen         = sizeof(int),
 399                 .mode           = 0644,
 400                 .proc_handler   = proc_dointvec
 401         },
 402 #endif
 403         {
 404                 .procname       = "tcp_fastopen",
 405                 .data           = &sysctl_tcp_fastopen,
 406                 .maxlen         = sizeof(int),
 407                 .mode           = 0644,
 408                 .proc_handler   = proc_dointvec,
 409         },
 410         {
 411                 .procname       = "tcp_fastopen_key",
 412                 .mode           = 0600,
 413                 .maxlen         = ((TCP_FASTOPEN_KEY_LENGTH * 2) + 10),
 414                 .proc_handler   = proc_tcp_fastopen_key,
 415         },
 416         {
 417                 .procname       = "tcp_tw_recycle",
 418                 .data           = &tcp_death_row.sysctl_tw_recycle,
 419                 .maxlen         = sizeof(int),
 420                 .mode           = 0644,
 421                 .proc_handler   = proc_dointvec
 422         },
 423         {
 424                 .procname       = "tcp_abort_on_overflow",
 425                 .data           = &sysctl_tcp_abort_on_overflow,
 426                 .maxlen         = sizeof(int),
 427                 .mode           = 0644,
 428                 .proc_handler   = proc_dointvec
 429         },
 430         {
 431                 .procname       = "tcp_stdurg",
 432                 .data           = &sysctl_tcp_stdurg,
 433                 .maxlen         = sizeof(int),
 434                 .mode           = 0644,
 435                 .proc_handler   = proc_dointvec
 436         },
 437         {
 438                 .procname       = "tcp_rfc1337",
 439                 .data           = &sysctl_tcp_rfc1337,
 440                 .maxlen         = sizeof(int),
 441                 .mode           = 0644,
 442                 .proc_handler   = proc_dointvec
 443         },
 444         {
 445                 .procname       = "tcp_max_syn_backlog",
 446                 .data           = &sysctl_max_syn_backlog,
 447                 .maxlen         = sizeof(int),
 448                 .mode           = 0644,
 449                 .proc_handler   = proc_dointvec
 450         },
 451         {
 452                 .procname       = "igmp_max_memberships",
 453                 .data           = &sysctl_igmp_max_memberships,
 454                 .maxlen         = sizeof(int),
 455                 .mode           = 0644,
 456                 .proc_handler   = proc_dointvec
 457         },
 458         {
 459                 .procname       = "igmp_max_msf",
 460                 .data           = &sysctl_igmp_max_msf,
 461                 .maxlen         = sizeof(int),
 462                 .mode           = 0644,
 463                 .proc_handler   = proc_dointvec
 464         },
 465 #ifdef CONFIG_IP_MULTICAST
 466         {
 467                 .procname       = "igmp_qrv",
 468                 .data           = &sysctl_igmp_qrv,
 469                 .maxlen         = sizeof(int),
 470                 .mode           = 0644,
 471                 .proc_handler   = proc_dointvec_minmax,
 472                 .extra1         = &one
 473         },
 474 #endif
 475         {
 476                 .procname       = "inet_peer_threshold",
 477                 .data           = &inet_peer_threshold,
 478                 .maxlen         = sizeof(int),
 479                 .mode           = 0644,
 480                 .proc_handler   = proc_dointvec
 481         },
 482         {
 483                 .procname       = "inet_peer_minttl",
 484                 .data           = &inet_peer_minttl,
 485                 .maxlen         = sizeof(int),
 486                 .mode           = 0644,
 487                 .proc_handler   = proc_dointvec_jiffies,
 488         },
 489         {
 490                 .procname       = "inet_peer_maxttl",
 491                 .data           = &inet_peer_maxttl,
 492                 .maxlen         = sizeof(int),
 493                 .mode           = 0644,
 494                 .proc_handler   = proc_dointvec_jiffies,
 495         },
 496         {
 497                 .procname       = "tcp_orphan_retries",
 498                 .data           = &sysctl_tcp_orphan_retries,
 499                 .maxlen         = sizeof(int),
 500                 .mode           = 0644,
 501                 .proc_handler   = proc_dointvec
 502         },
 503         {
 504                 .procname       = "tcp_fack",
 505                 .data           = &sysctl_tcp_fack,
 506                 .maxlen         = sizeof(int),
 507                 .mode           = 0644,
 508                 .proc_handler   = proc_dointvec
 509         },
 510         {
 511                 .procname       = "tcp_recovery",
 512                 .data           = &sysctl_tcp_recovery,
 513                 .maxlen         = sizeof(int),
 514                 .mode           = 0644,
 515                 .proc_handler   = proc_dointvec,
 516         },
 517         {
 518                 .procname       = "tcp_reordering",
 519                 .data           = &sysctl_tcp_reordering,
 520                 .maxlen         = sizeof(int),
 521                 .mode           = 0644,
 522                 .proc_handler   = proc_dointvec
 523         },
 524         {
 525                 .procname       = "tcp_max_reordering",
 526                 .data           = &sysctl_tcp_max_reordering,
 527                 .maxlen         = sizeof(int),
 528                 .mode           = 0644,
 529                 .proc_handler   = proc_dointvec
 530         },
 531         {
 532                 .procname       = "tcp_dsack",
 533                 .data           = &sysctl_tcp_dsack,
 534                 .maxlen         = sizeof(int),
 535                 .mode           = 0644,
 536                 .proc_handler   = proc_dointvec
 537         },
 538         {
 539                 .procname       = "tcp_mem",
 540                 .maxlen         = sizeof(sysctl_tcp_mem),
 541                 .data           = &sysctl_tcp_mem,
 542                 .mode           = 0644,
 543                 .proc_handler   = proc_doulongvec_minmax,
 544         },
 545         {
 546                 .procname       = "tcp_wmem",
 547                 .data           = &sysctl_tcp_wmem,
 548                 .maxlen         = sizeof(sysctl_tcp_wmem),
 549                 .mode           = 0644,
 550                 .proc_handler   = proc_dointvec_minmax,
 551                 .extra1         = &one,
 552         },
 553         {
 554                 .procname       = "tcp_notsent_lowat",
 555                 .data           = &sysctl_tcp_notsent_lowat,
 556                 .maxlen         = sizeof(sysctl_tcp_notsent_lowat),
 557                 .mode           = 0644,
 558                 .proc_handler   = proc_dointvec,
 559         },
 560         {
 561                 .procname       = "tcp_rmem",
 562                 .data           = &sysctl_tcp_rmem,
 563                 .maxlen         = sizeof(sysctl_tcp_rmem),
 564                 .mode           = 0644,
 565                 .proc_handler   = proc_dointvec_minmax,
 566                 .extra1         = &one,
 567         },
 568         {
 569                 .procname       = "tcp_app_win",
 570                 .data           = &sysctl_tcp_app_win,
 571                 .maxlen         = sizeof(int),
 572                 .mode           = 0644,
 573                 .proc_handler   = proc_dointvec
 574         },
 575         {
 576                 .procname       = "tcp_adv_win_scale",
 577                 .data           = &sysctl_tcp_adv_win_scale,
 578                 .maxlen         = sizeof(int),
 579                 .mode           = 0644,
 580                 .proc_handler   = proc_dointvec_minmax,
 581                 .extra1         = &tcp_adv_win_scale_min,
 582                 .extra2         = &tcp_adv_win_scale_max,
 583         },
 584         {
 585                 .procname       = "tcp_tw_reuse",
 586                 .data           = &sysctl_tcp_tw_reuse,
 587                 .maxlen         = sizeof(int),
 588                 .mode           = 0644,
 589                 .proc_handler   = proc_dointvec
 590         },
 591         {
 592                 .procname       = "tcp_frto",
 593                 .data           = &sysctl_tcp_frto,
 594                 .maxlen         = sizeof(int),
 595                 .mode           = 0644,
 596                 .proc_handler   = proc_dointvec
 597         },
 598         {
 599                 .procname       = "tcp_min_rtt_wlen",
 600                 .data           = &sysctl_tcp_min_rtt_wlen,
 601                 .maxlen         = sizeof(int),
 602                 .mode           = 0644,
 603                 .proc_handler   = proc_dointvec_minmax,
 604                 .extra1         = &zero,
 605                 .extra2         = &one_day_secs
 606         },
 607         {
 608                 .procname       = "tcp_low_latency",
 609                 .data           = &sysctl_tcp_low_latency,
 610                 .maxlen         = sizeof(int),
 611                 .mode           = 0644,
 612                 .proc_handler   = proc_dointvec
 613         },
 614         {
 615                 .procname       = "tcp_no_metrics_save",
 616                 .data           = &sysctl_tcp_nometrics_save,
 617                 .maxlen         = sizeof(int),
 618                 .mode           = 0644,
 619                 .proc_handler   = proc_dointvec,
 620         },
 621         {
 622                 .procname       = "tcp_moderate_rcvbuf",
 623                 .data           = &sysctl_tcp_moderate_rcvbuf,
 624                 .maxlen         = sizeof(int),
 625                 .mode           = 0644,
 626                 .proc_handler   = proc_dointvec,
 627         },
 628         {
 629                 .procname       = "tcp_tso_win_divisor",
 630                 .data           = &sysctl_tcp_tso_win_divisor,
 631                 .maxlen         = sizeof(int),
 632                 .mode           = 0644,
 633                 .proc_handler   = proc_dointvec,
 634         },
 635         {
 636                 .procname       = "tcp_congestion_control",
 637                 .mode           = 0644,
 638                 .maxlen         = TCP_CA_NAME_MAX,
 639                 .proc_handler   = proc_tcp_congestion_control,
 640         },
 641         {
 642                 .procname       = "tcp_workaround_signed_windows",
 643                 .data           = &sysctl_tcp_workaround_signed_windows,
 644                 .maxlen         = sizeof(int),
 645                 .mode           = 0644,
 646                 .proc_handler   = proc_dointvec
 647         },
 648         {
 649                 .procname       = "tcp_limit_output_bytes",
 650                 .data           = &sysctl_tcp_limit_output_bytes,
 651                 .maxlen         = sizeof(int),
 652                 .mode           = 0644,
 653                 .proc_handler   = proc_dointvec
 654         },
 655         {
 656                 .procname       = "tcp_challenge_ack_limit",
 657                 .data           = &sysctl_tcp_challenge_ack_limit,
 658                 .maxlen         = sizeof(int),
 659                 .mode           = 0644,
 660                 .proc_handler   = proc_dointvec
 661         },
 662         {
 663                 .procname       = "tcp_slow_start_after_idle",
 664                 .data           = &sysctl_tcp_slow_start_after_idle,
 665                 .maxlen         = sizeof(int),
 666                 .mode           = 0644,
 667                 .proc_handler   = proc_dointvec
 668         },
 669 #ifdef CONFIG_NETLABEL
 670         {
 671                 .procname       = "cipso_cache_enable",
 672                 .data           = &cipso_v4_cache_enabled,
 673                 .maxlen         = sizeof(int),
 674                 .mode           = 0644,
 675                 .proc_handler   = proc_dointvec,
 676         },
 677         {
 678                 .procname       = "cipso_cache_bucket_size",
 679                 .data           = &cipso_v4_cache_bucketsize,
 680                 .maxlen         = sizeof(int),
 681                 .mode           = 0644,
 682                 .proc_handler   = proc_dointvec,
 683         },
 684         {
 685                 .procname       = "cipso_rbm_optfmt",
 686                 .data           = &cipso_v4_rbm_optfmt,
 687                 .maxlen         = sizeof(int),
 688                 .mode           = 0644,
 689                 .proc_handler   = proc_dointvec,
 690         },
 691         {
 692                 .procname       = "cipso_rbm_strictvalid",
 693                 .data           = &cipso_v4_rbm_strictvalid,
 694                 .maxlen         = sizeof(int),
 695                 .mode           = 0644,
 696                 .proc_handler   = proc_dointvec,
 697         },
 698 #endif /* CONFIG_NETLABEL */
 699         {
 700                 .procname       = "tcp_available_congestion_control",
 701                 .maxlen         = TCP_CA_BUF_MAX,
 702                 .mode           = 0444,
 703                 .proc_handler   = proc_tcp_available_congestion_control,
 704         },
 705         {
 706                 .procname       = "tcp_allowed_congestion_control",
 707                 .maxlen         = TCP_CA_BUF_MAX,
 708                 .mode           = 0644,
 709                 .proc_handler   = proc_allowed_congestion_control,
 710         },
 711         {
 712                 .procname       = "tcp_thin_linear_timeouts",
 713                 .data           = &sysctl_tcp_thin_linear_timeouts,
 714                 .maxlen         = sizeof(int),
 715                 .mode           = 0644,
 716                 .proc_handler   = proc_dointvec
 717         },
 718         {
 719                 .procname       = "tcp_thin_dupack",
 720                 .data           = &sysctl_tcp_thin_dupack,
 721                 .maxlen         = sizeof(int),
 722                 .mode           = 0644,
 723                 .proc_handler   = proc_dointvec
 724         },
 725         {
 726                 .procname       = "tcp_early_retrans",
 727                 .data           = &sysctl_tcp_early_retrans,
 728                 .maxlen         = sizeof(int),
 729                 .mode           = 0644,
 730                 .proc_handler   = proc_dointvec_minmax,
 731                 .extra1         = &zero,
 732                 .extra2         = &four,
 733         },
 734         {
 735                 .procname       = "tcp_min_tso_segs",
 736                 .data           = &sysctl_tcp_min_tso_segs,
 737                 .maxlen         = sizeof(int),
 738                 .mode           = 0644,
 739                 .proc_handler   = proc_dointvec_minmax,
 740                 .extra1         = &one,
 741                 .extra2         = &gso_max_segs,
 742         },
 743         {
 744                 .procname       = "tcp_pacing_ss_ratio",
 745                 .data           = &sysctl_tcp_pacing_ss_ratio,
 746                 .maxlen         = sizeof(int),
 747                 .mode           = 0644,
 748                 .proc_handler   = proc_dointvec_minmax,
 749                 .extra1         = &zero,
 750                 .extra2         = &thousand,
 751         },
 752         {
 753                 .procname       = "tcp_pacing_ca_ratio",
 754                 .data           = &sysctl_tcp_pacing_ca_ratio,
 755                 .maxlen         = sizeof(int),
 756                 .mode           = 0644,
 757                 .proc_handler   = proc_dointvec_minmax,
 758                 .extra1         = &zero,
 759                 .extra2         = &thousand,
 760         },
 761         {
 762                 .procname       = "tcp_autocorking",
 763                 .data           = &sysctl_tcp_autocorking,
 764                 .maxlen         = sizeof(int),
 765                 .mode           = 0644,
 766                 .proc_handler   = proc_dointvec_minmax,
 767                 .extra1         = &zero,
 768                 .extra2         = &one,
 769         },
 770         {
 771                 .procname       = "tcp_invalid_ratelimit",
 772                 .data           = &sysctl_tcp_invalid_ratelimit,
 773                 .maxlen         = sizeof(int),
 774                 .mode           = 0644,
 775                 .proc_handler   = proc_dointvec_ms_jiffies,
 776         },
 777         {
 778                 .procname       = "icmp_msgs_per_sec",
 779                 .data           = &sysctl_icmp_msgs_per_sec,
 780                 .maxlen         = sizeof(int),
 781                 .mode           = 0644,
 782                 .proc_handler   = proc_dointvec_minmax,
 783                 .extra1         = &zero,
 784         },
 785         {
 786                 .procname       = "icmp_msgs_burst",
 787                 .data           = &sysctl_icmp_msgs_burst,
 788                 .maxlen         = sizeof(int),
 789                 .mode           = 0644,
 790                 .proc_handler   = proc_dointvec_minmax,
 791                 .extra1         = &zero,
 792         },
 793         {
 794                 .procname       = "udp_mem",
 795                 .data           = &sysctl_udp_mem,
 796                 .maxlen         = sizeof(sysctl_udp_mem),
 797                 .mode           = 0644,
 798                 .proc_handler   = proc_doulongvec_minmax,
 799         },
 800         {
 801                 .procname       = "udp_rmem_min",
 802                 .data           = &sysctl_udp_rmem_min,
 803                 .maxlen         = sizeof(sysctl_udp_rmem_min),
 804                 .mode           = 0644,
 805                 .proc_handler   = proc_dointvec_minmax,
 806                 .extra1         = &one
 807         },
 808         {
 809                 .procname       = "udp_wmem_min",
 810                 .data           = &sysctl_udp_wmem_min,
 811                 .maxlen         = sizeof(sysctl_udp_wmem_min),
 812                 .mode           = 0644,
 813                 .proc_handler   = proc_dointvec_minmax,
 814                 .extra1         = &one
 815         },
 816         { }
 817 };
 818
 819 static struct ctl_table ipv4_net_table[] = {
 820         {
 821                 .procname       = "icmp_echo_ignore_all",
 822                 .data           = &init_net.ipv4.sysctl_icmp_echo_ignore_all,
 823                 .maxlen         = sizeof(int),
 824                 .mode           = 0644,
 825                 .proc_handler   = proc_dointvec
 826         },
 827         {
 828                 .procname       = "icmp_echo_ignore_broadcasts",
 829                 .data           = &init_net.ipv4.sysctl_icmp_echo_ignore_broadcasts,
 830                 .maxlen         = sizeof(int),
 831                 .mode           = 0644,
 832                 .proc_handler   = proc_dointvec
 833         },
 834         {
 835                 .procname       = "icmp_ignore_bogus_error_responses",
 836                 .data           = &init_net.ipv4.sysctl_icmp_ignore_bogus_error_responses,
 837                 .maxlen         = sizeof(int),
 838                 .mode           = 0644,
 839                 .proc_handler   = proc_dointvec
 840         },
 841         {
 842                 .procname       = "icmp_errors_use_inbound_ifaddr",
 843                 .data           = &init_net.ipv4.sysctl_icmp_errors_use_inbound_ifaddr,
 844                 .maxlen         = sizeof(int),
 845                 .mode           = 0644,
 846                 .proc_handler   = proc_dointvec
 847         },
 848         {
 849                 .procname       = "icmp_ratelimit",
 850                 .data           = &init_net.ipv4.sysctl_icmp_ratelimit,
 851                 .maxlen         = sizeof(int),
 852                 .mode           = 0644,
 853                 .proc_handler   = proc_dointvec_ms_jiffies,
 854         },
 855         {
 856                 .procname       = "icmp_ratemask",
 857                 .data           = &init_net.ipv4.sysctl_icmp_ratemask,
 858                 .maxlen         = sizeof(int),
 859                 .mode           = 0644,
 860                 .proc_handler   = proc_dointvec
 861         },
 862         {
 863                 .procname       = "ping_group_range",
 864                 .data           = &init_net.ipv4.ping_group_range.range,
 865                 .maxlen         = sizeof(gid_t)*2,
 866                 .mode           = 0644,
 867                 .proc_handler   = ipv4_ping_group_range,
 868         },
 869         {
 870                 .procname       = "tcp_ecn",
 871                 .data           = &init_net.ipv4.sysctl_tcp_ecn,
 872                 .maxlen         = sizeof(int),
 873                 .mode           = 0644,
 874                 .proc_handler   = proc_dointvec
 875         },
 876         {
 877                 .procname       = "tcp_ecn_fallback",
 878                 .data           = &init_net.ipv4.sysctl_tcp_ecn_fallback,
 879                 .maxlen         = sizeof(int),
 880                 .mode           = 0644,
 881                 .proc_handler   = proc_dointvec
 882         },
 883         {
 884                 .procname       = "ip_local_port_range",
 885                 .maxlen         = sizeof(init_net.ipv4.ip_local_ports.range),
 886                 .data           = &init_net.ipv4.ip_local_ports.range,
 887                 .mode           = 0644,
 888                 .proc_handler   = ipv4_local_port_range,
 889         },
 890         {
 891                 .procname       = "ip_local_reserved_ports",
 892                 .data           = &init_net.ipv4.sysctl_local_reserved_ports,
 893                 .maxlen         = 65536,
 894                 .mode           = 0644,
 895                 .proc_handler   = proc_do_large_bitmap,
 896         },
 897         {
 898                 .procname       = "ip_no_pmtu_disc",
 899                 .data           = &init_net.ipv4.sysctl_ip_no_pmtu_disc,
 900                 .maxlen         = sizeof(int),
 901                 .mode           = 0644,
 902                 .proc_handler   = proc_dointvec
 903         },
 904         {
 905                 .procname       = "ip_forward_use_pmtu",
 906                 .data           = &init_net.ipv4.sysctl_ip_fwd_use_pmtu,
 907                 .maxlen         = sizeof(int),
 908                 .mode           = 0644,
 909                 .proc_handler   = proc_dointvec,
 910         },
 911         {
 912                 .procname       = "ip_nonlocal_bind",
 913                 .data           = &init_net.ipv4.sysctl_ip_nonlocal_bind,
 914                 .maxlen         = sizeof(int),
 915                 .mode           = 0644,
 916                 .proc_handler   = proc_dointvec
 917         },
 918         {
 919                 .procname       = "fwmark_reflect",
 920                 .data           = &init_net.ipv4.sysctl_fwmark_reflect,
 921                 .maxlen         = sizeof(int),
 922                 .mode           = 0644,
 923                 .proc_handler   = proc_dointvec,
 924         },
 925         {
 926                 .procname       = "tcp_fwmark_accept",
 927                 .data           = &init_net.ipv4.sysctl_tcp_fwmark_accept,
 928                 .maxlen         = sizeof(int),
 929                 .mode           = 0644,
 930                 .proc_handler   = proc_dointvec,
 931         },
 932         {
 933                 .procname       = "tcp_mtu_probing",
 934                 .data           = &init_net.ipv4.sysctl_tcp_mtu_probing,
 935                 .maxlen         = sizeof(int),
 936                 .mode           = 0644,
 937                 .proc_handler   = proc_dointvec,
 938         },
 939         {
 940                 .procname       = "tcp_base_mss",
 941                 .data           = &init_net.ipv4.sysctl_tcp_base_mss,
 942                 .maxlen         = sizeof(int),
 943                 .mode           = 0644,
 944                 .proc_handler   = proc_dointvec,
 945         },
 946         {
 947                 .procname       = "tcp_min_snd_mss",
 948                 .data           = &init_net.ipv4.sysctl_tcp_min_snd_mss,
 949                 .maxlen         = sizeof(int),
 950                 .mode           = 0644,
 951                 .proc_handler   = proc_dointvec_minmax,
 952                 .extra1         = &tcp_min_snd_mss_min,
 953                 .extra2         = &tcp_min_snd_mss_max,
 954         },
 955         {
 956                 .procname       = "tcp_probe_threshold",
 957                 .data           = &init_net.ipv4.sysctl_tcp_probe_threshold,
 958                 .maxlen         = sizeof(int),
 959                 .mode           = 0644,
 960                 .proc_handler   = proc_dointvec,
 961         },
 962         {
 963                 .procname       = "tcp_probe_interval",
 964                 .data           = &init_net.ipv4.sysctl_tcp_probe_interval,
 965                 .maxlen         = sizeof(int),
 966                 .mode           = 0644,
 967                 .proc_handler   = proc_dointvec,
 968         },
 969         {
 970                 .procname       = "igmp_link_local_mcast_reports",
 971                 .data           = &sysctl_igmp_llm_reports,
 972                 .maxlen         = sizeof(int),
 973                 .mode           = 0644,
 974                 .proc_handler   = proc_dointvec
 975         },
 976         { }
 977 };
 978
 979 static __net_init int ipv4_sysctl_init_net(struct net *net)
 980 {
 981         struct ctl_table *table;
 982
 983         table = ipv4_net_table;
 984         if (!net_eq(net, &init_net)) {
 985                 int i;
 986
 987                 table = kmemdup(table, sizeof(ipv4_net_table), GFP_KERNEL);
 988                 if (!table)
 989                         goto err_alloc;
 990
 991                 /* Update the variables to point into the current struct net */
 992                 for (i = 0; i < ARRAY_SIZE(ipv4_net_table) - 1; i++)
 993                         table[i].data += (void *)net - (void *)&init_net;
 994         }
 995
 996         net->ipv4.ipv4_hdr = register_net_sysctl(net, "net/ipv4", table);
 997         if (!net->ipv4.ipv4_hdr)
 998                 goto err_reg;
 999
1000         net->ipv4.sysctl_local_reserved_ports = kzalloc(65536 / 8, GFP_KERNEL);
1001         if (!net->ipv4.sysctl_local_reserved_ports)
1002                 goto err_ports;
1003
1004         return 0;
1005
1006 err_ports:
1007         unregister_net_sysctl_table(net->ipv4.ipv4_hdr);
1008 err_reg:
1009         if (!net_eq(net, &init_net))
1010                 kfree(table);
1011 err_alloc:
1012         return -ENOMEM;
1013 }
1014
1015 static __net_exit void ipv4_sysctl_exit_net(struct net *net)
1016 {
1017         struct ctl_table *table;
1018
1019         kfree(net->ipv4.sysctl_local_reserved_ports);
1020         table = net->ipv4.ipv4_hdr->ctl_table_arg;
1021         unregister_net_sysctl_table(net->ipv4.ipv4_hdr);
1022         kfree(table);
1023 }
1024
1025 static __net_initdata struct pernet_operations ipv4_sysctl_ops = {
1026         .init = ipv4_sysctl_init_net,
1027         .exit = ipv4_sysctl_exit_net,
1028 };
1029
1030 static __init int sysctl_ipv4_init(void)
1031 {
1032         struct ctl_table_header *hdr;
1033
1034         hdr = register_net_sysctl(&init_net, "net/ipv4", ipv4_table);
1035         if (!hdr)
1036                 return -ENOMEM;
1037
1038         if (register_pernet_subsys(&ipv4_sysctl_ops)) {
1039                 unregister_net_sysctl_table(hdr);
1040                 return -ENOMEM;
1041         }
1042
1043         return 0;
1044 }
1045
1046 __initcall(sysctl_ipv4_init);