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
  28 static int zero;
  29 static int one = 1;
  30 static int four = 4;
  31 static int thousand = 1000;
  32 static int gso_max_segs = GSO_MAX_SEGS;
  33 static int tcp_retr1_max = 255;
  34 static int ip_local_port_range_min[] = { 1, 1 };
  35 static int ip_local_port_range_max[] = { 65535, 65535 };
  36 static int tcp_adv_win_scale_min = -31;
  37 static int tcp_adv_win_scale_max = 31;
  38 static int ip_ttl_min = 1;
  39 static int ip_ttl_max = 255;
  40 static int tcp_syn_retries_min = 1;
  41 static int tcp_syn_retries_max = MAX_TCP_SYNCNT;
  42 static int ip_ping_group_range_min[] = { 0, 0 };
  43 static int ip_ping_group_range_max[] = { GID_T_MAX, GID_T_MAX };
  44
  45 /* Update system visible IP port range */
  46 static void set_local_port_range(struct net *net, int range[2])
  47 {
  48         bool same_parity = !((range[0] ^ range[1]) & 1);
  49
  50         write_seqlock_bh(&net->ipv4.ip_local_ports.lock);
  51         if (same_parity && !net->ipv4.ip_local_ports.warned) {
  52                 net->ipv4.ip_local_ports.warned = true;
  53                 pr_err_ratelimited("ip_local_port_range: prefer different parity for start/end values.\n");
  54         }
  55         net->ipv4.ip_local_ports.range[0] = range[0];
  56         net->ipv4.ip_local_ports.range[1] = range[1];
  57         write_sequnlock_bh(&net->ipv4.ip_local_ports.lock);
  58 }
  59
  60 /* Validate changes from /proc interface. */
  61 static int ipv4_local_port_range(struct ctl_table *table, int write,
  62                                  void __user *buffer,
  63                                  size_t *lenp, loff_t *ppos)
  64 {
  65         struct net *net =
  66                 container_of(table->data, struct net, ipv4.ip_local_ports.range);
  67         int ret;
  68         int range[2];
  69         struct ctl_table tmp = {
  70                 .data = &range,
  71                 .maxlen = sizeof(range),
  72                 .mode = table->mode,
  73                 .extra1 = &ip_local_port_range_min,
  74                 .extra2 = &ip_local_port_range_max,
  75         };
  76
  77         inet_get_local_port_range(net, &range[0], &range[1]);
  78
  79         ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
  80
  81         if (write && ret == 0) {
  82                 if (range[1] < range[0])
  83                         ret = -EINVAL;
  84                 else
  85                         set_local_port_range(net, range);
  86         }
  87
  88         return ret;
  89 }
  90
  91
  92 static void inet_get_ping_group_range_table(struct ctl_table *table, kgid_t *low, kgid_t *high)
  93 {
  94         kgid_t *data = table->data;
  95         struct net *net =
  96                 container_of(table->data, struct net, ipv4.ping_group_range.range);
  97         unsigned int seq;
  98         do {
  99                 seq = read_seqbegin(&net->ipv4.ping_group_range.lock);
 100
 101                 *low = data[0];
 102                 *high = data[1];
 103         } while (read_seqretry(&net->ipv4.ping_group_range.lock, seq));
 104 }
 105
 106 /* Update system visible IP port range */
 107 static void set_ping_group_range(struct ctl_table *table, kgid_t low, kgid_t high)
 108 {
 109         kgid_t *data = table->data;
 110         struct net *net =
 111                 container_of(table->data, struct net, ipv4.ping_group_range.range);
 112         write_seqlock(&net->ipv4.ping_group_range.lock);
 113         data[0] = low;
 114         data[1] = high;
 115         write_sequnlock(&net->ipv4.ping_group_range.lock);
 116 }
 117
 118 /* Validate changes from /proc interface. */
 119 static int ipv4_ping_group_range(struct ctl_table *table, int write,
 120                                  void __user *buffer,
 121                                  size_t *lenp, loff_t *ppos)
 122 {
 123         struct user_namespace *user_ns = current_user_ns();
 124         int ret;
 125         gid_t urange[2];
 126         kgid_t low, high;
 127         struct ctl_table tmp = {
 128                 .data = &urange,
 129                 .maxlen = sizeof(urange),
 130                 .mode = table->mode,
 131                 .extra1 = &ip_ping_group_range_min,
 132                 .extra2 = &ip_ping_group_range_max,
 133         };
 134
 135         inet_get_ping_group_range_table(table, &low, &high);
 136         urange[0] = from_kgid_munged(user_ns, low);
 137         urange[1] = from_kgid_munged(user_ns, high);
 138         ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
 139
 140         if (write && ret == 0) {
 141                 low = make_kgid(user_ns, urange[0]);
 142                 high = make_kgid(user_ns, urange[1]);
 143                 if (!gid_valid(low) || !gid_valid(high))
 144                         return -EINVAL;
 145                 if (urange[1] < urange[0] || gid_lt(high, low)) {
 146                         low = make_kgid(&init_user_ns, 1);
 147                         high = make_kgid(&init_user_ns, 0);
 148                 }
 149                 set_ping_group_range(table, low, high);
 150         }
 151
 152         return ret;
 153 }
 154
 155 static int proc_tcp_congestion_control(struct ctl_table *ctl, int write,
 156                                        void __user *buffer, size_t *lenp, loff_t *ppos)
 157 {
 158         char val[TCP_CA_NAME_MAX];
 159         struct ctl_table tbl = {
 160                 .data = val,
 161                 .maxlen = TCP_CA_NAME_MAX,
 162         };
 163         int ret;
 164
 165         tcp_get_default_congestion_control(val);
 166
 167         ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
 168         if (write && ret == 0)
 169                 ret = tcp_set_default_congestion_control(val);
 170         return ret;
 171 }
 172
 173 static int proc_tcp_available_congestion_control(struct ctl_table *ctl,
 174                                                  int write,
 175                                                  void __user *buffer, size_t *lenp,
 176                                                  loff_t *ppos)
 177 {
 178         struct ctl_table tbl = { .maxlen = TCP_CA_BUF_MAX, };
 179         int ret;
 180
 181         tbl.data = kmalloc(tbl.maxlen, GFP_USER);
 182         if (!tbl.data)
 183                 return -ENOMEM;
 184         tcp_get_available_congestion_control(tbl.data, TCP_CA_BUF_MAX);
 185         ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
 186         kfree(tbl.data);
 187         return ret;
 188 }
 189
 190 static int proc_allowed_congestion_control(struct ctl_table *ctl,
 191                                            int write,
 192                                            void __user *buffer, size_t *lenp,
 193                                            loff_t *ppos)
 194 {
 195         struct ctl_table tbl = { .maxlen = TCP_CA_BUF_MAX };
 196         int ret;
 197
 198         tbl.data = kmalloc(tbl.maxlen, GFP_USER);
 199         if (!tbl.data)
 200                 return -ENOMEM;
 201
 202         tcp_get_allowed_congestion_control(tbl.data, tbl.maxlen);
 203         ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
 204         if (write && ret == 0)
 205                 ret = tcp_set_allowed_congestion_control(tbl.data);
 206         kfree(tbl.data);
 207         return ret;
 208 }
 209
 210 static int proc_tcp_fastopen_key(struct ctl_table *ctl, int write,
 211                                  void __user *buffer, size_t *lenp,
 212                                  loff_t *ppos)
 213 {
 214         struct ctl_table tbl = { .maxlen = (TCP_FASTOPEN_KEY_LENGTH * 2 + 10) };
 215         struct tcp_fastopen_context *ctxt;
 216         u32  user_key[4]; /* 16 bytes, matching TCP_FASTOPEN_KEY_LENGTH */
 217         __le32 key[4];
 218         int ret, i;
 219
 220         tbl.data = kmalloc(tbl.maxlen, GFP_KERNEL);
 221         if (!tbl.data)
 222                 return -ENOMEM;
 223
 224         rcu_read_lock();
 225         ctxt = rcu_dereference(tcp_fastopen_ctx);
 226         if (ctxt)
 227                 memcpy(key, ctxt->key, TCP_FASTOPEN_KEY_LENGTH);
 228         else
 229                 memset(key, 0, sizeof(key));
 230         rcu_read_unlock();
 231
 232         for (i = 0; i < ARRAY_SIZE(key); i++)
 233                 user_key[i] = le32_to_cpu(key[i]);
 234
 235         snprintf(tbl.data, tbl.maxlen, "%08x-%08x-%08x-%08x",
 236                 user_key[0], user_key[1], user_key[2], user_key[3]);
 237         ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
 238
 239         if (write && ret == 0) {
 240                 if (sscanf(tbl.data, "%x-%x-%x-%x", user_key, user_key + 1,
 241                            user_key + 2, user_key + 3) != 4) {
 242                         ret = -EINVAL;
 243                         goto bad_key;
 244                 }
 245                 /* Generate a dummy secret but don't publish it. This
 246                  * is needed so we don't regenerate a new key on the
 247                  * first invocation of tcp_fastopen_cookie_gen
 248                  */
 249                 tcp_fastopen_init_key_once(false);
 250
 251                 for (i = 0; i < ARRAY_SIZE(user_key); i++)
 252                         key[i] = cpu_to_le32(user_key[i]);
 253
 254                 tcp_fastopen_reset_cipher(key, TCP_FASTOPEN_KEY_LENGTH);
 255         }
 256
 257 bad_key:
 258         pr_debug("proc FO key set 0x%x-%x-%x-%x <- 0x%s: %u\n",
 259                  user_key[0], user_key[1], user_key[2], user_key[3],
 260                (char *)tbl.data, ret);
 261         kfree(tbl.data);
 262         return ret;
 263 }
 264
 265 static struct ctl_table ipv4_table[] = {
 266         {
 267                 .procname       = "tcp_timestamps",
 268                 .data           = &sysctl_tcp_timestamps,
 269                 .maxlen         = sizeof(int),
 270                 .mode           = 0644,
 271                 .proc_handler   = proc_dointvec
 272         },
 273         {
 274                 .procname       = "tcp_window_scaling",
 275                 .data           = &sysctl_tcp_window_scaling,
 276                 .maxlen         = sizeof(int),
 277                 .mode           = 0644,
 278                 .proc_handler   = proc_dointvec
 279         },
 280         {
 281                 .procname       = "tcp_sack",
 282                 .data           = &sysctl_tcp_sack,
 283                 .maxlen         = sizeof(int),
 284                 .mode           = 0644,
 285                 .proc_handler   = proc_dointvec
 286         },
 287         {
 288                 .procname       = "tcp_retrans_collapse",
 289                 .data           = &sysctl_tcp_retrans_collapse,
 290                 .maxlen         = sizeof(int),
 291                 .mode           = 0644,
 292                 .proc_handler   = proc_dointvec
 293         },
 294         {
 295                 .procname       = "tcp_max_orphans",
 296                 .data           = &sysctl_tcp_max_orphans,
 297                 .maxlen         = sizeof(int),
 298                 .mode           = 0644,
 299                 .proc_handler   = proc_dointvec
 300         },
 301         {
 302                 .procname       = "tcp_max_tw_buckets",
 303                 .data           = &tcp_death_row.sysctl_max_tw_buckets,
 304                 .maxlen         = sizeof(int),
 305                 .mode           = 0644,
 306                 .proc_handler   = proc_dointvec
 307         },
 308         {
 309                 .procname       = "tcp_fastopen",
 310                 .data           = &sysctl_tcp_fastopen,
 311                 .maxlen         = sizeof(int),
 312                 .mode           = 0644,
 313                 .proc_handler   = proc_dointvec,
 314         },
 315         {
 316                 .procname       = "tcp_fastopen_key",
 317                 .mode           = 0600,
 318                 .maxlen         = ((TCP_FASTOPEN_KEY_LENGTH * 2) + 10),
 319                 .proc_handler   = proc_tcp_fastopen_key,
 320         },
 321         {
 322                 .procname       = "tcp_tw_recycle",
 323                 .data           = &tcp_death_row.sysctl_tw_recycle,
 324                 .maxlen         = sizeof(int),
 325                 .mode           = 0644,
 326                 .proc_handler   = proc_dointvec
 327         },
 328         {
 329                 .procname       = "tcp_abort_on_overflow",
 330                 .data           = &sysctl_tcp_abort_on_overflow,
 331                 .maxlen         = sizeof(int),
 332                 .mode           = 0644,
 333                 .proc_handler   = proc_dointvec
 334         },
 335         {
 336                 .procname       = "tcp_stdurg",
 337                 .data           = &sysctl_tcp_stdurg,
 338                 .maxlen         = sizeof(int),
 339                 .mode           = 0644,
 340                 .proc_handler   = proc_dointvec
 341         },
 342         {
 343                 .procname       = "tcp_rfc1337",
 344                 .data           = &sysctl_tcp_rfc1337,
 345                 .maxlen         = sizeof(int),
 346                 .mode           = 0644,
 347                 .proc_handler   = proc_dointvec
 348         },
 349         {
 350                 .procname       = "tcp_max_syn_backlog",
 351                 .data           = &sysctl_max_syn_backlog,
 352                 .maxlen         = sizeof(int),
 353                 .mode           = 0644,
 354                 .proc_handler   = proc_dointvec
 355         },
 356         {
 357                 .procname       = "inet_peer_threshold",
 358                 .data           = &inet_peer_threshold,
 359                 .maxlen         = sizeof(int),
 360                 .mode           = 0644,
 361                 .proc_handler   = proc_dointvec
 362         },
 363         {
 364                 .procname       = "inet_peer_minttl",
 365                 .data           = &inet_peer_minttl,
 366                 .maxlen         = sizeof(int),
 367                 .mode           = 0644,
 368                 .proc_handler   = proc_dointvec_jiffies,
 369         },
 370         {
 371                 .procname       = "inet_peer_maxttl",
 372                 .data           = &inet_peer_maxttl,
 373                 .maxlen         = sizeof(int),
 374                 .mode           = 0644,
 375                 .proc_handler   = proc_dointvec_jiffies,
 376         },
 377         {
 378                 .procname       = "tcp_fack",
 379                 .data           = &sysctl_tcp_fack,
 380                 .maxlen         = sizeof(int),
 381                 .mode           = 0644,
 382                 .proc_handler   = proc_dointvec
 383         },
 384         {
 385                 .procname       = "tcp_recovery",
 386                 .data           = &sysctl_tcp_recovery,
 387                 .maxlen         = sizeof(int),
 388                 .mode           = 0644,
 389                 .proc_handler   = proc_dointvec,
 390         },
 391         {
 392                 .procname       = "tcp_max_reordering",
 393                 .data           = &sysctl_tcp_max_reordering,
 394                 .maxlen         = sizeof(int),
 395                 .mode           = 0644,
 396                 .proc_handler   = proc_dointvec
 397         },
 398         {
 399                 .procname       = "tcp_dsack",
 400                 .data           = &sysctl_tcp_dsack,
 401                 .maxlen         = sizeof(int),
 402                 .mode           = 0644,
 403                 .proc_handler   = proc_dointvec
 404         },
 405         {
 406                 .procname       = "tcp_mem",
 407                 .maxlen         = sizeof(sysctl_tcp_mem),
 408                 .data           = &sysctl_tcp_mem,
 409                 .mode           = 0644,
 410                 .proc_handler   = proc_doulongvec_minmax,
 411         },
 412         {
 413                 .procname       = "tcp_wmem",
 414                 .data           = &sysctl_tcp_wmem,
 415                 .maxlen         = sizeof(sysctl_tcp_wmem),
 416                 .mode           = 0644,
 417                 .proc_handler   = proc_dointvec_minmax,
 418                 .extra1         = &one,
 419         },
 420         {
 421                 .procname       = "tcp_rmem",
 422                 .data           = &sysctl_tcp_rmem,
 423                 .maxlen         = sizeof(sysctl_tcp_rmem),
 424                 .mode           = 0644,
 425                 .proc_handler   = proc_dointvec_minmax,
 426                 .extra1         = &one,
 427         },
 428         {
 429                 .procname       = "tcp_app_win",
 430                 .data           = &sysctl_tcp_app_win,
 431                 .maxlen         = sizeof(int),
 432                 .mode           = 0644,
 433                 .proc_handler   = proc_dointvec
 434         },
 435         {
 436                 .procname       = "tcp_adv_win_scale",
 437                 .data           = &sysctl_tcp_adv_win_scale,
 438                 .maxlen         = sizeof(int),
 439                 .mode           = 0644,
 440                 .proc_handler   = proc_dointvec_minmax,
 441                 .extra1         = &tcp_adv_win_scale_min,
 442                 .extra2         = &tcp_adv_win_scale_max,
 443         },
 444         {
 445                 .procname       = "tcp_tw_reuse",
 446                 .data           = &sysctl_tcp_tw_reuse,
 447                 .maxlen         = sizeof(int),
 448                 .mode           = 0644,
 449                 .proc_handler   = proc_dointvec
 450         },
 451         {
 452                 .procname       = "tcp_frto",
 453                 .data           = &sysctl_tcp_frto,
 454                 .maxlen         = sizeof(int),
 455                 .mode           = 0644,
 456                 .proc_handler   = proc_dointvec
 457         },
 458         {
 459                 .procname       = "tcp_min_rtt_wlen",
 460                 .data           = &sysctl_tcp_min_rtt_wlen,
 461                 .maxlen         = sizeof(int),
 462                 .mode           = 0644,
 463                 .proc_handler   = proc_dointvec
 464         },
 465         {
 466                 .procname       = "tcp_low_latency",
 467                 .data           = &sysctl_tcp_low_latency,
 468                 .maxlen         = sizeof(int),
 469                 .mode           = 0644,
 470                 .proc_handler   = proc_dointvec
 471         },
 472         {
 473                 .procname       = "tcp_no_metrics_save",
 474                 .data           = &sysctl_tcp_nometrics_save,
 475                 .maxlen         = sizeof(int),
 476                 .mode           = 0644,
 477                 .proc_handler   = proc_dointvec,
 478         },
 479         {
 480                 .procname       = "tcp_moderate_rcvbuf",
 481                 .data           = &sysctl_tcp_moderate_rcvbuf,
 482                 .maxlen         = sizeof(int),
 483                 .mode           = 0644,
 484                 .proc_handler   = proc_dointvec,
 485         },
 486         {
 487                 .procname       = "tcp_tso_win_divisor",
 488                 .data           = &sysctl_tcp_tso_win_divisor,
 489                 .maxlen         = sizeof(int),
 490                 .mode           = 0644,
 491                 .proc_handler   = proc_dointvec,
 492         },
 493         {
 494                 .procname       = "tcp_congestion_control",
 495                 .mode           = 0644,
 496                 .maxlen         = TCP_CA_NAME_MAX,
 497                 .proc_handler   = proc_tcp_congestion_control,
 498         },
 499         {
 500                 .procname       = "tcp_workaround_signed_windows",
 501                 .data           = &sysctl_tcp_workaround_signed_windows,
 502                 .maxlen         = sizeof(int),
 503                 .mode           = 0644,
 504                 .proc_handler   = proc_dointvec
 505         },
 506         {
 507                 .procname       = "tcp_limit_output_bytes",
 508                 .data           = &sysctl_tcp_limit_output_bytes,
 509                 .maxlen         = sizeof(int),
 510                 .mode           = 0644,
 511                 .proc_handler   = proc_dointvec
 512         },
 513         {
 514                 .procname       = "tcp_challenge_ack_limit",
 515                 .data           = &sysctl_tcp_challenge_ack_limit,
 516                 .maxlen         = sizeof(int),
 517                 .mode           = 0644,
 518                 .proc_handler   = proc_dointvec
 519         },
 520         {
 521                 .procname       = "tcp_slow_start_after_idle",
 522                 .data           = &sysctl_tcp_slow_start_after_idle,
 523                 .maxlen         = sizeof(int),
 524                 .mode           = 0644,
 525                 .proc_handler   = proc_dointvec
 526         },
 527 #ifdef CONFIG_NETLABEL
 528         {
 529                 .procname       = "cipso_cache_enable",
 530                 .data           = &cipso_v4_cache_enabled,
 531                 .maxlen         = sizeof(int),
 532                 .mode           = 0644,
 533                 .proc_handler   = proc_dointvec,
 534         },
 535         {
 536                 .procname       = "cipso_cache_bucket_size",
 537                 .data           = &cipso_v4_cache_bucketsize,
 538                 .maxlen         = sizeof(int),
 539                 .mode           = 0644,
 540                 .proc_handler   = proc_dointvec,
 541         },
 542         {
 543                 .procname       = "cipso_rbm_optfmt",
 544                 .data           = &cipso_v4_rbm_optfmt,
 545                 .maxlen         = sizeof(int),
 546                 .mode           = 0644,
 547                 .proc_handler   = proc_dointvec,
 548         },
 549         {
 550                 .procname       = "cipso_rbm_strictvalid",
 551                 .data           = &cipso_v4_rbm_strictvalid,
 552                 .maxlen         = sizeof(int),
 553                 .mode           = 0644,
 554                 .proc_handler   = proc_dointvec,
 555         },
 556 #endif /* CONFIG_NETLABEL */
 557         {
 558                 .procname       = "tcp_available_congestion_control",
 559                 .maxlen         = TCP_CA_BUF_MAX,
 560                 .mode           = 0444,
 561                 .proc_handler   = proc_tcp_available_congestion_control,
 562         },
 563         {
 564                 .procname       = "tcp_allowed_congestion_control",
 565                 .maxlen         = TCP_CA_BUF_MAX,
 566                 .mode           = 0644,
 567                 .proc_handler   = proc_allowed_congestion_control,
 568         },
 569         {
 570                 .procname       = "tcp_thin_linear_timeouts",
 571                 .data           = &sysctl_tcp_thin_linear_timeouts,
 572                 .maxlen         = sizeof(int),
 573                 .mode           = 0644,
 574                 .proc_handler   = proc_dointvec
 575         },
 576         {
 577                 .procname       = "tcp_thin_dupack",
 578                 .data           = &sysctl_tcp_thin_dupack,
 579                 .maxlen         = sizeof(int),
 580                 .mode           = 0644,
 581                 .proc_handler   = proc_dointvec
 582         },
 583         {
 584                 .procname       = "tcp_early_retrans",
 585                 .data           = &sysctl_tcp_early_retrans,
 586                 .maxlen         = sizeof(int),
 587                 .mode           = 0644,
 588                 .proc_handler   = proc_dointvec_minmax,
 589                 .extra1         = &zero,
 590                 .extra2         = &four,
 591         },
 592         {
 593                 .procname       = "tcp_min_tso_segs",
 594                 .data           = &sysctl_tcp_min_tso_segs,
 595                 .maxlen         = sizeof(int),
 596                 .mode           = 0644,
 597                 .proc_handler   = proc_dointvec_minmax,
 598                 .extra1         = &one,
 599                 .extra2         = &gso_max_segs,
 600         },
 601         {
 602                 .procname       = "tcp_pacing_ss_ratio",
 603                 .data           = &sysctl_tcp_pacing_ss_ratio,
 604                 .maxlen         = sizeof(int),
 605                 .mode           = 0644,
 606                 .proc_handler   = proc_dointvec_minmax,
 607                 .extra1         = &zero,
 608                 .extra2         = &thousand,
 609         },
 610         {
 611                 .procname       = "tcp_pacing_ca_ratio",
 612                 .data           = &sysctl_tcp_pacing_ca_ratio,
 613                 .maxlen         = sizeof(int),
 614                 .mode           = 0644,
 615                 .proc_handler   = proc_dointvec_minmax,
 616                 .extra1         = &zero,
 617                 .extra2         = &thousand,
 618         },
 619         {
 620                 .procname       = "tcp_autocorking",
 621                 .data           = &sysctl_tcp_autocorking,
 622                 .maxlen         = sizeof(int),
 623                 .mode           = 0644,
 624                 .proc_handler   = proc_dointvec_minmax,
 625                 .extra1         = &zero,
 626                 .extra2         = &one,
 627         },
 628         {
 629                 .procname       = "tcp_invalid_ratelimit",
 630                 .data           = &sysctl_tcp_invalid_ratelimit,
 631                 .maxlen         = sizeof(int),
 632                 .mode           = 0644,
 633                 .proc_handler   = proc_dointvec_ms_jiffies,
 634         },
 635         {
 636                 .procname       = "icmp_msgs_per_sec",
 637                 .data           = &sysctl_icmp_msgs_per_sec,
 638                 .maxlen         = sizeof(int),
 639                 .mode           = 0644,
 640                 .proc_handler   = proc_dointvec_minmax,
 641                 .extra1         = &zero,
 642         },
 643         {
 644                 .procname       = "icmp_msgs_burst",
 645                 .data           = &sysctl_icmp_msgs_burst,
 646                 .maxlen         = sizeof(int),
 647                 .mode           = 0644,
 648                 .proc_handler   = proc_dointvec_minmax,
 649                 .extra1         = &zero,
 650         },
 651         {
 652                 .procname       = "udp_mem",
 653                 .data           = &sysctl_udp_mem,
 654                 .maxlen         = sizeof(sysctl_udp_mem),
 655                 .mode           = 0644,
 656                 .proc_handler   = proc_doulongvec_minmax,
 657         },
 658         {
 659                 .procname       = "udp_rmem_min",
 660                 .data           = &sysctl_udp_rmem_min,
 661                 .maxlen         = sizeof(sysctl_udp_rmem_min),
 662                 .mode           = 0644,
 663                 .proc_handler   = proc_dointvec_minmax,
 664                 .extra1         = &one
 665         },
 666         {
 667                 .procname       = "udp_wmem_min",
 668                 .data           = &sysctl_udp_wmem_min,
 669                 .maxlen         = sizeof(sysctl_udp_wmem_min),
 670                 .mode           = 0644,
 671                 .proc_handler   = proc_dointvec_minmax,
 672                 .extra1         = &one
 673         },
 674         { }
 675 };
 676
 677 static struct ctl_table ipv4_net_table[] = {
 678         {
 679                 .procname       = "icmp_echo_ignore_all",
 680                 .data           = &init_net.ipv4.sysctl_icmp_echo_ignore_all,
 681                 .maxlen         = sizeof(int),
 682                 .mode           = 0644,
 683                 .proc_handler   = proc_dointvec
 684         },
 685         {
 686                 .procname       = "icmp_echo_ignore_broadcasts",
 687                 .data           = &init_net.ipv4.sysctl_icmp_echo_ignore_broadcasts,
 688                 .maxlen         = sizeof(int),
 689                 .mode           = 0644,
 690                 .proc_handler   = proc_dointvec
 691         },
 692         {
 693                 .procname       = "icmp_ignore_bogus_error_responses",
 694                 .data           = &init_net.ipv4.sysctl_icmp_ignore_bogus_error_responses,
 695                 .maxlen         = sizeof(int),
 696                 .mode           = 0644,
 697                 .proc_handler   = proc_dointvec
 698         },
 699         {
 700                 .procname       = "icmp_errors_use_inbound_ifaddr",
 701                 .data           = &init_net.ipv4.sysctl_icmp_errors_use_inbound_ifaddr,
 702                 .maxlen         = sizeof(int),
 703                 .mode           = 0644,
 704                 .proc_handler   = proc_dointvec
 705         },
 706         {
 707                 .procname       = "icmp_ratelimit",
 708                 .data           = &init_net.ipv4.sysctl_icmp_ratelimit,
 709                 .maxlen         = sizeof(int),
 710                 .mode           = 0644,
 711                 .proc_handler   = proc_dointvec_ms_jiffies,
 712         },
 713         {
 714                 .procname       = "icmp_ratemask",
 715                 .data           = &init_net.ipv4.sysctl_icmp_ratemask,
 716                 .maxlen         = sizeof(int),
 717                 .mode           = 0644,
 718                 .proc_handler   = proc_dointvec
 719         },
 720         {
 721                 .procname       = "ping_group_range",
 722                 .data           = &init_net.ipv4.ping_group_range.range,
 723                 .maxlen         = sizeof(gid_t)*2,
 724                 .mode           = 0644,
 725                 .proc_handler   = ipv4_ping_group_range,
 726         },
 727         {
 728                 .procname       = "tcp_ecn",
 729                 .data           = &init_net.ipv4.sysctl_tcp_ecn,
 730                 .maxlen         = sizeof(int),
 731                 .mode           = 0644,
 732                 .proc_handler   = proc_dointvec
 733         },
 734         {
 735                 .procname       = "tcp_ecn_fallback",
 736                 .data           = &init_net.ipv4.sysctl_tcp_ecn_fallback,
 737                 .maxlen         = sizeof(int),
 738                 .mode           = 0644,
 739                 .proc_handler   = proc_dointvec
 740         },
 741         {
 742                 .procname       = "ip_dynaddr",
 743                 .data           = &init_net.ipv4.sysctl_ip_dynaddr,
 744                 .maxlen         = sizeof(int),
 745                 .mode           = 0644,
 746                 .proc_handler   = proc_dointvec
 747         },
 748         {
 749                 .procname       = "ip_early_demux",
 750                 .data           = &init_net.ipv4.sysctl_ip_early_demux,
 751                 .maxlen         = sizeof(int),
 752                 .mode           = 0644,
 753                 .proc_handler   = proc_dointvec
 754         },
 755         {
 756                 .procname       = "ip_default_ttl",
 757                 .data           = &init_net.ipv4.sysctl_ip_default_ttl,
 758                 .maxlen         = sizeof(int),
 759                 .mode           = 0644,
 760                 .proc_handler   = proc_dointvec_minmax,
 761                 .extra1         = &ip_ttl_min,
 762                 .extra2         = &ip_ttl_max,
 763         },
 764         {
 765                 .procname       = "ip_local_port_range",
 766                 .maxlen         = sizeof(init_net.ipv4.ip_local_ports.range),
 767                 .data           = &init_net.ipv4.ip_local_ports.range,
 768                 .mode           = 0644,
 769                 .proc_handler   = ipv4_local_port_range,
 770         },
 771         {
 772                 .procname       = "ip_local_reserved_ports",
 773                 .data           = &init_net.ipv4.sysctl_local_reserved_ports,
 774                 .maxlen         = 65536,
 775                 .mode           = 0644,
 776                 .proc_handler   = proc_do_large_bitmap,
 777         },
 778         {
 779                 .procname       = "ip_no_pmtu_disc",
 780                 .data           = &init_net.ipv4.sysctl_ip_no_pmtu_disc,
 781                 .maxlen         = sizeof(int),
 782                 .mode           = 0644,
 783                 .proc_handler   = proc_dointvec
 784         },
 785         {
 786                 .procname       = "ip_forward_use_pmtu",
 787                 .data           = &init_net.ipv4.sysctl_ip_fwd_use_pmtu,
 788                 .maxlen         = sizeof(int),
 789                 .mode           = 0644,
 790                 .proc_handler   = proc_dointvec,
 791         },
 792         {
 793                 .procname       = "ip_nonlocal_bind",
 794                 .data           = &init_net.ipv4.sysctl_ip_nonlocal_bind,
 795                 .maxlen         = sizeof(int),
 796                 .mode           = 0644,
 797                 .proc_handler   = proc_dointvec
 798         },
 799         {
 800                 .procname       = "fwmark_reflect",
 801                 .data           = &init_net.ipv4.sysctl_fwmark_reflect,
 802                 .maxlen         = sizeof(int),
 803                 .mode           = 0644,
 804                 .proc_handler   = proc_dointvec,
 805         },
 806         {
 807                 .procname       = "tcp_fwmark_accept",
 808                 .data           = &init_net.ipv4.sysctl_tcp_fwmark_accept,
 809                 .maxlen         = sizeof(int),
 810                 .mode           = 0644,
 811                 .proc_handler   = proc_dointvec,
 812         },
 813 #ifdef CONFIG_NET_L3_MASTER_DEV
 814         {
 815                 .procname       = "tcp_l3mdev_accept",
 816                 .data           = &init_net.ipv4.sysctl_tcp_l3mdev_accept,
 817                 .maxlen         = sizeof(int),
 818                 .mode           = 0644,
 819                 .proc_handler   = proc_dointvec_minmax,
 820                 .extra1         = &zero,
 821                 .extra2         = &one,
 822         },
 823 #endif
 824         {
 825                 .procname       = "tcp_mtu_probing",
 826                 .data           = &init_net.ipv4.sysctl_tcp_mtu_probing,
 827                 .maxlen         = sizeof(int),
 828                 .mode           = 0644,
 829                 .proc_handler   = proc_dointvec,
 830         },
 831         {
 832                 .procname       = "tcp_base_mss",
 833                 .data           = &init_net.ipv4.sysctl_tcp_base_mss,
 834                 .maxlen         = sizeof(int),
 835                 .mode           = 0644,
 836                 .proc_handler   = proc_dointvec,
 837         },
 838         {
 839                 .procname       = "tcp_probe_threshold",
 840                 .data           = &init_net.ipv4.sysctl_tcp_probe_threshold,
 841                 .maxlen         = sizeof(int),
 842                 .mode           = 0644,
 843                 .proc_handler   = proc_dointvec,
 844         },
 845         {
 846                 .procname       = "tcp_probe_interval",
 847                 .data           = &init_net.ipv4.sysctl_tcp_probe_interval,
 848                 .maxlen         = sizeof(int),
 849                 .mode           = 0644,
 850                 .proc_handler   = proc_dointvec,
 851         },
 852         {
 853                 .procname       = "igmp_link_local_mcast_reports",
 854                 .data           = &init_net.ipv4.sysctl_igmp_llm_reports,
 855                 .maxlen         = sizeof(int),
 856                 .mode           = 0644,
 857                 .proc_handler   = proc_dointvec
 858         },
 859         {
 860                 .procname       = "igmp_max_memberships",
 861                 .data           = &init_net.ipv4.sysctl_igmp_max_memberships,
 862                 .maxlen         = sizeof(int),
 863                 .mode           = 0644,
 864                 .proc_handler   = proc_dointvec
 865         },
 866         {
 867                 .procname       = "igmp_max_msf",
 868                 .data           = &init_net.ipv4.sysctl_igmp_max_msf,
 869                 .maxlen         = sizeof(int),
 870                 .mode           = 0644,
 871                 .proc_handler   = proc_dointvec
 872         },
 873 #ifdef CONFIG_IP_MULTICAST
 874         {
 875                 .procname       = "igmp_qrv",
 876                 .data           = &init_net.ipv4.sysctl_igmp_qrv,
 877                 .maxlen         = sizeof(int),
 878                 .mode           = 0644,
 879                 .proc_handler   = proc_dointvec_minmax,
 880                 .extra1         = &one
 881         },
 882 #endif
 883         {
 884                 .procname       = "tcp_keepalive_time",
 885                 .data           = &init_net.ipv4.sysctl_tcp_keepalive_time,
 886                 .maxlen         = sizeof(int),
 887                 .mode           = 0644,
 888                 .proc_handler   = proc_dointvec_jiffies,
 889         },
 890         {
 891                 .procname       = "tcp_keepalive_probes",
 892                 .data           = &init_net.ipv4.sysctl_tcp_keepalive_probes,
 893                 .maxlen         = sizeof(int),
 894                 .mode           = 0644,
 895                 .proc_handler   = proc_dointvec
 896         },
 897         {
 898                 .procname       = "tcp_keepalive_intvl",
 899                 .data           = &init_net.ipv4.sysctl_tcp_keepalive_intvl,
 900                 .maxlen         = sizeof(int),
 901                 .mode           = 0644,
 902                 .proc_handler   = proc_dointvec_jiffies,
 903         },
 904         {
 905                 .procname       = "tcp_syn_retries",
 906                 .data           = &init_net.ipv4.sysctl_tcp_syn_retries,
 907                 .maxlen         = sizeof(int),
 908                 .mode           = 0644,
 909                 .proc_handler   = proc_dointvec_minmax,
 910                 .extra1         = &tcp_syn_retries_min,
 911                 .extra2         = &tcp_syn_retries_max
 912         },
 913         {
 914                 .procname       = "tcp_synack_retries",
 915                 .data           = &init_net.ipv4.sysctl_tcp_synack_retries,
 916                 .maxlen         = sizeof(int),
 917                 .mode           = 0644,
 918                 .proc_handler   = proc_dointvec
 919         },
 920 #ifdef CONFIG_SYN_COOKIES
 921         {
 922                 .procname       = "tcp_syncookies",
 923                 .data           = &init_net.ipv4.sysctl_tcp_syncookies,
 924                 .maxlen         = sizeof(int),
 925                 .mode           = 0644,
 926                 .proc_handler   = proc_dointvec
 927         },
 928 #endif
 929         {
 930                 .procname       = "tcp_reordering",
 931                 .data           = &init_net.ipv4.sysctl_tcp_reordering,
 932                 .maxlen         = sizeof(int),
 933                 .mode           = 0644,
 934                 .proc_handler   = proc_dointvec
 935         },
 936         {
 937                 .procname       = "tcp_retries1",
 938                 .data           = &init_net.ipv4.sysctl_tcp_retries1,
 939                 .maxlen         = sizeof(int),
 940                 .mode           = 0644,
 941                 .proc_handler   = proc_dointvec_minmax,
 942                 .extra2         = &tcp_retr1_max
 943         },
 944         {
 945                 .procname       = "tcp_retries2",
 946                 .data           = &init_net.ipv4.sysctl_tcp_retries2,
 947                 .maxlen         = sizeof(int),
 948                 .mode           = 0644,
 949                 .proc_handler   = proc_dointvec
 950         },
 951         {
 952                 .procname       = "tcp_orphan_retries",
 953                 .data           = &init_net.ipv4.sysctl_tcp_orphan_retries,
 954                 .maxlen         = sizeof(int),
 955                 .mode           = 0644,
 956                 .proc_handler   = proc_dointvec
 957         },
 958         {
 959                 .procname       = "tcp_fin_timeout",
 960                 .data           = &init_net.ipv4.sysctl_tcp_fin_timeout,
 961                 .maxlen         = sizeof(int),
 962                 .mode           = 0644,
 963                 .proc_handler   = proc_dointvec_jiffies,
 964         },
 965         {
 966                 .procname       = "tcp_notsent_lowat",
 967                 .data           = &init_net.ipv4.sysctl_tcp_notsent_lowat,
 968                 .maxlen         = sizeof(unsigned int),
 969                 .mode           = 0644,
 970                 .proc_handler   = proc_douintvec,
 971         },
 972 #ifdef CONFIG_IP_ROUTE_MULTIPATH
 973         {
 974                 .procname       = "fib_multipath_use_neigh",
 975                 .data           = &init_net.ipv4.sysctl_fib_multipath_use_neigh,
 976                 .maxlen         = sizeof(int),
 977                 .mode           = 0644,
 978                 .proc_handler   = proc_dointvec_minmax,
 979                 .extra1         = &zero,
 980                 .extra2         = &one,
 981         },
 982 #endif
 983         { }
 984 };
 985
 986 static __net_init int ipv4_sysctl_init_net(struct net *net)
 987 {
 988         struct ctl_table *table;
 989
 990         table = ipv4_net_table;
 991         if (!net_eq(net, &init_net)) {
 992                 int i;
 993
 994                 table = kmemdup(table, sizeof(ipv4_net_table), GFP_KERNEL);
 995                 if (!table)
 996                         goto err_alloc;
 997
 998                 /* Update the variables to point into the current struct net */
 999                 for (i = 0; i < ARRAY_SIZE(ipv4_net_table) - 1; i++)
1000                         table[i].data += (void *)net - (void *)&init_net;
1001         }
1002
1003         net->ipv4.ipv4_hdr = register_net_sysctl(net, "net/ipv4", table);
1004         if (!net->ipv4.ipv4_hdr)
1005                 goto err_reg;
1006
1007         net->ipv4.sysctl_local_reserved_ports = kzalloc(65536 / 8, GFP_KERNEL);
1008         if (!net->ipv4.sysctl_local_reserved_ports)
1009                 goto err_ports;
1010
1011         return 0;
1012
1013 err_ports:
1014         unregister_net_sysctl_table(net->ipv4.ipv4_hdr);
1015 err_reg:
1016         if (!net_eq(net, &init_net))
1017                 kfree(table);
1018 err_alloc:
1019         return -ENOMEM;
1020 }
1021
1022 static __net_exit void ipv4_sysctl_exit_net(struct net *net)
1023 {
1024         struct ctl_table *table;
1025
1026         kfree(net->ipv4.sysctl_local_reserved_ports);
1027         table = net->ipv4.ipv4_hdr->ctl_table_arg;
1028         unregister_net_sysctl_table(net->ipv4.ipv4_hdr);
1029         kfree(table);
1030 }
1031
1032 static __net_initdata struct pernet_operations ipv4_sysctl_ops = {
1033         .init = ipv4_sysctl_init_net,
1034         .exit = ipv4_sysctl_exit_net,
1035 };
1036
1037 static __init int sysctl_ipv4_init(void)
1038 {
1039         struct ctl_table_header *hdr;
1040
1041         hdr = register_net_sysctl(&init_net, "net/ipv4", ipv4_table);
1042         if (!hdr)
1043                 return -ENOMEM;
1044
1045         if (register_pernet_subsys(&ipv4_sysctl_ops)) {
1046                 unregister_net_sysctl_table(hdr);
1047                 return -ENOMEM;
1048         }
1049
1050         return 0;
1051 }
1052
1053 __initcall(sysctl_ipv4_init);