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 tcp_min_snd_mss_min = TCP_MIN_SND_MSS;
  39 static int tcp_min_snd_mss_max = 65535;
  40 static int ip_ttl_min = 1;
  41 static int ip_ttl_max = 255;
  42 static int tcp_syn_retries_min = 1;
  43 static int tcp_syn_retries_max = MAX_TCP_SYNCNT;
  44 static int ip_ping_group_range_min[] = { 0, 0 };
  45 static int ip_ping_group_range_max[] = { GID_T_MAX, GID_T_MAX };
  46 static int one_day_secs = 24 * 3600;
  47
  48 /* Update system visible IP port range */
  49 static void set_local_port_range(struct net *net, int range[2])
  50 {
  51         bool same_parity = !((range[0] ^ range[1]) & 1);
  52
  53         write_seqlock_bh(&net->ipv4.ip_local_ports.lock);
  54         if (same_parity && !net->ipv4.ip_local_ports.warned) {
  55                 net->ipv4.ip_local_ports.warned = true;
  56                 pr_err_ratelimited("ip_local_port_range: prefer different parity for start/end values.\n");
  57         }
  58         net->ipv4.ip_local_ports.range[0] = range[0];
  59         net->ipv4.ip_local_ports.range[1] = range[1];
  60         write_sequnlock_bh(&net->ipv4.ip_local_ports.lock);
  61 }
  62
  63 /* Validate changes from /proc interface. */
  64 static int ipv4_local_port_range(struct ctl_table *table, int write,
  65                                  void __user *buffer,
  66                                  size_t *lenp, loff_t *ppos)
  67 {
  68         struct net *net =
  69                 container_of(table->data, struct net, ipv4.ip_local_ports.range);
  70         int ret;
  71         int range[2];
  72         struct ctl_table tmp = {
  73                 .data = &range,
  74                 .maxlen = sizeof(range),
  75                 .mode = table->mode,
  76                 .extra1 = &ip_local_port_range_min,
  77                 .extra2 = &ip_local_port_range_max,
  78         };
  79
  80         inet_get_local_port_range(net, &range[0], &range[1]);
  81
  82         ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
  83
  84         if (write && ret == 0) {
  85                 if (range[1] < range[0])
  86                         ret = -EINVAL;
  87                 else
  88                         set_local_port_range(net, range);
  89         }
  90
  91         return ret;
  92 }
  93
  94
  95 static void inet_get_ping_group_range_table(struct ctl_table *table, kgid_t *low, kgid_t *high)
  96 {
  97         kgid_t *data = table->data;
  98         struct net *net =
  99                 container_of(table->data, struct net, ipv4.ping_group_range.range);
 100         unsigned int seq;
 101         do {
 102                 seq = read_seqbegin(&net->ipv4.ping_group_range.lock);
 103
 104                 *low = data[0];
 105                 *high = data[1];
 106         } while (read_seqretry(&net->ipv4.ping_group_range.lock, seq));
 107 }
 108
 109 /* Update system visible IP port range */
 110 static void set_ping_group_range(struct ctl_table *table, kgid_t low, kgid_t high)
 111 {
 112         kgid_t *data = table->data;
 113         struct net *net =
 114                 container_of(table->data, struct net, ipv4.ping_group_range.range);
 115         write_seqlock(&net->ipv4.ping_group_range.lock);
 116         data[0] = low;
 117         data[1] = high;
 118         write_sequnlock(&net->ipv4.ping_group_range.lock);
 119 }
 120
 121 /* Validate changes from /proc interface. */
 122 static int ipv4_ping_group_range(struct ctl_table *table, int write,
 123                                  void __user *buffer,
 124                                  size_t *lenp, loff_t *ppos)
 125 {
 126         struct user_namespace *user_ns = current_user_ns();
 127         int ret;
 128         gid_t urange[2];
 129         kgid_t low, high;
 130         struct ctl_table tmp = {
 131                 .data = &urange,
 132                 .maxlen = sizeof(urange),
 133                 .mode = table->mode,
 134                 .extra1 = &ip_ping_group_range_min,
 135                 .extra2 = &ip_ping_group_range_max,
 136         };
 137
 138         inet_get_ping_group_range_table(table, &low, &high);
 139         urange[0] = from_kgid_munged(user_ns, low);
 140         urange[1] = from_kgid_munged(user_ns, high);
 141         ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
 142
 143         if (write && ret == 0) {
 144                 low = make_kgid(user_ns, urange[0]);
 145                 high = make_kgid(user_ns, urange[1]);
 146                 if (!gid_valid(low) || !gid_valid(high))
 147                         return -EINVAL;
 148                 if (urange[1] < urange[0] || gid_lt(high, low)) {
 149                         low = make_kgid(&init_user_ns, 1);
 150                         high = make_kgid(&init_user_ns, 0);
 151                 }
 152                 set_ping_group_range(table, low, high);
 153         }
 154
 155         return ret;
 156 }
 157
 158 static int proc_tcp_congestion_control(struct ctl_table *ctl, int write,
 159                                        void __user *buffer, size_t *lenp, loff_t *ppos)
 160 {
 161         char val[TCP_CA_NAME_MAX];
 162         struct ctl_table tbl = {
 163                 .data = val,
 164                 .maxlen = TCP_CA_NAME_MAX,
 165         };
 166         int ret;
 167
 168         tcp_get_default_congestion_control(val);
 169
 170         ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
 171         if (write && ret == 0)
 172                 ret = tcp_set_default_congestion_control(val);
 173         return ret;
 174 }
 175
 176 static int proc_tcp_available_congestion_control(struct ctl_table *ctl,
 177                                                  int write,
 178                                                  void __user *buffer, size_t *lenp,
 179                                                  loff_t *ppos)
 180 {
 181         struct ctl_table tbl = { .maxlen = TCP_CA_BUF_MAX, };
 182         int ret;
 183
 184         tbl.data = kmalloc(tbl.maxlen, GFP_USER);
 185         if (!tbl.data)
 186                 return -ENOMEM;
 187         tcp_get_available_congestion_control(tbl.data, TCP_CA_BUF_MAX);
 188         ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
 189         kfree(tbl.data);
 190         return ret;
 191 }
 192
 193 static int proc_allowed_congestion_control(struct ctl_table *ctl,
 194                                            int write,
 195                                            void __user *buffer, size_t *lenp,
 196                                            loff_t *ppos)
 197 {
 198         struct ctl_table tbl = { .maxlen = TCP_CA_BUF_MAX };
 199         int ret;
 200
 201         tbl.data = kmalloc(tbl.maxlen, GFP_USER);
 202         if (!tbl.data)
 203                 return -ENOMEM;
 204
 205         tcp_get_allowed_congestion_control(tbl.data, tbl.maxlen);
 206         ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
 207         if (write && ret == 0)
 208                 ret = tcp_set_allowed_congestion_control(tbl.data);
 209         kfree(tbl.data);
 210         return ret;
 211 }
 212
 213 static int proc_tcp_fastopen_key(struct ctl_table *ctl, int write,
 214                                  void __user *buffer, size_t *lenp,
 215                                  loff_t *ppos)
 216 {
 217         struct ctl_table tbl = { .maxlen = (TCP_FASTOPEN_KEY_LENGTH * 2 + 10) };
 218         struct tcp_fastopen_context *ctxt;
 219         u32  user_key[4]; /* 16 bytes, matching TCP_FASTOPEN_KEY_LENGTH */
 220         __le32 key[4];
 221         int ret, i;
 222
 223         tbl.data = kmalloc(tbl.maxlen, GFP_KERNEL);
 224         if (!tbl.data)
 225                 return -ENOMEM;
 226
 227         rcu_read_lock();
 228         ctxt = rcu_dereference(tcp_fastopen_ctx);
 229         if (ctxt)
 230                 memcpy(key, ctxt->key, TCP_FASTOPEN_KEY_LENGTH);
 231         else
 232                 memset(key, 0, sizeof(key));
 233         rcu_read_unlock();
 234
 235         for (i = 0; i < ARRAY_SIZE(key); i++)
 236                 user_key[i] = le32_to_cpu(key[i]);
 237
 238         snprintf(tbl.data, tbl.maxlen, "%08x-%08x-%08x-%08x",
 239                 user_key[0], user_key[1], user_key[2], user_key[3]);
 240         ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
 241
 242         if (write && ret == 0) {
 243                 if (sscanf(tbl.data, "%x-%x-%x-%x", user_key, user_key + 1,
 244                            user_key + 2, user_key + 3) != 4) {
 245                         ret = -EINVAL;
 246                         goto bad_key;
 247                 }
 248                 /* Generate a dummy secret but don't publish it. This
 249                  * is needed so we don't regenerate a new key on the
 250                  * first invocation of tcp_fastopen_cookie_gen
 251                  */
 252                 tcp_fastopen_init_key_once(false);
 253
 254                 for (i = 0; i < ARRAY_SIZE(user_key); i++)
 255                         key[i] = cpu_to_le32(user_key[i]);
 256
 257                 tcp_fastopen_reset_cipher(key, TCP_FASTOPEN_KEY_LENGTH);
 258         }
 259
 260 bad_key:
 261         pr_debug("proc FO key set 0x%x-%x-%x-%x <- 0x%s: %u\n",
 262                  user_key[0], user_key[1], user_key[2], user_key[3],
 263                (char *)tbl.data, ret);
 264         kfree(tbl.data);
 265         return ret;
 266 }
 267
 268 static struct ctl_table ipv4_table[] = {
 269         {
 270                 .procname       = "tcp_timestamps",
 271                 .data           = &sysctl_tcp_timestamps,
 272                 .maxlen         = sizeof(int),
 273                 .mode           = 0644,
 274                 .proc_handler   = proc_dointvec
 275         },
 276         {
 277                 .procname       = "tcp_window_scaling",
 278                 .data           = &sysctl_tcp_window_scaling,
 279                 .maxlen         = sizeof(int),
 280                 .mode           = 0644,
 281                 .proc_handler   = proc_dointvec
 282         },
 283         {
 284                 .procname       = "tcp_sack",
 285                 .data           = &sysctl_tcp_sack,
 286                 .maxlen         = sizeof(int),
 287                 .mode           = 0644,
 288                 .proc_handler   = proc_dointvec
 289         },
 290         {
 291                 .procname       = "tcp_retrans_collapse",
 292                 .data           = &sysctl_tcp_retrans_collapse,
 293                 .maxlen         = sizeof(int),
 294                 .mode           = 0644,
 295                 .proc_handler   = proc_dointvec
 296         },
 297         {
 298                 .procname       = "tcp_max_orphans",
 299                 .data           = &sysctl_tcp_max_orphans,
 300                 .maxlen         = sizeof(int),
 301                 .mode           = 0644,
 302                 .proc_handler   = proc_dointvec
 303         },
 304         {
 305                 .procname       = "tcp_max_tw_buckets",
 306                 .data           = &tcp_death_row.sysctl_max_tw_buckets,
 307                 .maxlen         = sizeof(int),
 308                 .mode           = 0644,
 309                 .proc_handler   = proc_dointvec
 310         },
 311         {
 312                 .procname       = "tcp_fastopen",
 313                 .data           = &sysctl_tcp_fastopen,
 314                 .maxlen         = sizeof(int),
 315                 .mode           = 0644,
 316                 .proc_handler   = proc_dointvec,
 317         },
 318         {
 319                 .procname       = "tcp_fastopen_key",
 320                 .mode           = 0600,
 321                 .maxlen         = ((TCP_FASTOPEN_KEY_LENGTH * 2) + 10),
 322                 .proc_handler   = proc_tcp_fastopen_key,
 323         },
 324         {
 325                 .procname       = "tcp_tw_recycle",
 326                 .data           = &tcp_death_row.sysctl_tw_recycle,
 327                 .maxlen         = sizeof(int),
 328                 .mode           = 0644,
 329                 .proc_handler   = proc_dointvec
 330         },
 331         {
 332                 .procname       = "tcp_abort_on_overflow",
 333                 .data           = &sysctl_tcp_abort_on_overflow,
 334                 .maxlen         = sizeof(int),
 335                 .mode           = 0644,
 336                 .proc_handler   = proc_dointvec
 337         },
 338         {
 339                 .procname       = "tcp_stdurg",
 340                 .data           = &sysctl_tcp_stdurg,
 341                 .maxlen         = sizeof(int),
 342                 .mode           = 0644,
 343                 .proc_handler   = proc_dointvec
 344         },
 345         {
 346                 .procname       = "tcp_rfc1337",
 347                 .data           = &sysctl_tcp_rfc1337,
 348                 .maxlen         = sizeof(int),
 349                 .mode           = 0644,
 350                 .proc_handler   = proc_dointvec
 351         },
 352         {
 353                 .procname       = "tcp_max_syn_backlog",
 354                 .data           = &sysctl_max_syn_backlog,
 355                 .maxlen         = sizeof(int),
 356                 .mode           = 0644,
 357                 .proc_handler   = proc_dointvec
 358         },
 359         {
 360                 .procname       = "inet_peer_threshold",
 361                 .data           = &inet_peer_threshold,
 362                 .maxlen         = sizeof(int),
 363                 .mode           = 0644,
 364                 .proc_handler   = proc_dointvec
 365         },
 366         {
 367                 .procname       = "inet_peer_minttl",
 368                 .data           = &inet_peer_minttl,
 369                 .maxlen         = sizeof(int),
 370                 .mode           = 0644,
 371                 .proc_handler   = proc_dointvec_jiffies,
 372         },
 373         {
 374                 .procname       = "inet_peer_maxttl",
 375                 .data           = &inet_peer_maxttl,
 376                 .maxlen         = sizeof(int),
 377                 .mode           = 0644,
 378                 .proc_handler   = proc_dointvec_jiffies,
 379         },
 380         {
 381                 .procname       = "tcp_fack",
 382                 .data           = &sysctl_tcp_fack,
 383                 .maxlen         = sizeof(int),
 384                 .mode           = 0644,
 385                 .proc_handler   = proc_dointvec
 386         },
 387         {
 388                 .procname       = "tcp_recovery",
 389                 .data           = &sysctl_tcp_recovery,
 390                 .maxlen         = sizeof(int),
 391                 .mode           = 0644,
 392                 .proc_handler   = proc_dointvec,
 393         },
 394         {
 395                 .procname       = "tcp_max_reordering",
 396                 .data           = &sysctl_tcp_max_reordering,
 397                 .maxlen         = sizeof(int),
 398                 .mode           = 0644,
 399                 .proc_handler   = proc_dointvec
 400         },
 401         {
 402                 .procname       = "tcp_dsack",
 403                 .data           = &sysctl_tcp_dsack,
 404                 .maxlen         = sizeof(int),
 405                 .mode           = 0644,
 406                 .proc_handler   = proc_dointvec
 407         },
 408         {
 409                 .procname       = "tcp_mem",
 410                 .maxlen         = sizeof(sysctl_tcp_mem),
 411                 .data           = &sysctl_tcp_mem,
 412                 .mode           = 0644,
 413                 .proc_handler   = proc_doulongvec_minmax,
 414         },
 415         {
 416                 .procname       = "tcp_wmem",
 417                 .data           = &sysctl_tcp_wmem,
 418                 .maxlen         = sizeof(sysctl_tcp_wmem),
 419                 .mode           = 0644,
 420                 .proc_handler   = proc_dointvec_minmax,
 421                 .extra1         = &one,
 422         },
 423         {
 424                 .procname       = "tcp_rmem",
 425                 .data           = &sysctl_tcp_rmem,
 426                 .maxlen         = sizeof(sysctl_tcp_rmem),
 427                 .mode           = 0644,
 428                 .proc_handler   = proc_dointvec_minmax,
 429                 .extra1         = &one,
 430         },
 431         {
 432                 .procname       = "tcp_app_win",
 433                 .data           = &sysctl_tcp_app_win,
 434                 .maxlen         = sizeof(int),
 435                 .mode           = 0644,
 436                 .proc_handler   = proc_dointvec
 437         },
 438         {
 439                 .procname       = "tcp_adv_win_scale",
 440                 .data           = &sysctl_tcp_adv_win_scale,
 441                 .maxlen         = sizeof(int),
 442                 .mode           = 0644,
 443                 .proc_handler   = proc_dointvec_minmax,
 444                 .extra1         = &tcp_adv_win_scale_min,
 445                 .extra2         = &tcp_adv_win_scale_max,
 446         },
 447         {
 448                 .procname       = "tcp_tw_reuse",
 449                 .data           = &sysctl_tcp_tw_reuse,
 450                 .maxlen         = sizeof(int),
 451                 .mode           = 0644,
 452                 .proc_handler   = proc_dointvec
 453         },
 454         {
 455                 .procname       = "tcp_frto",
 456                 .data           = &sysctl_tcp_frto,
 457                 .maxlen         = sizeof(int),
 458                 .mode           = 0644,
 459                 .proc_handler   = proc_dointvec
 460         },
 461         {
 462                 .procname       = "tcp_min_rtt_wlen",
 463                 .data           = &sysctl_tcp_min_rtt_wlen,
 464                 .maxlen         = sizeof(int),
 465                 .mode           = 0644,
 466                 .proc_handler   = proc_dointvec_minmax,
 467                 .extra1         = &zero,
 468                 .extra2         = &one_day_secs
 469         },
 470         {
 471                 .procname       = "tcp_low_latency",
 472                 .data           = &sysctl_tcp_low_latency,
 473                 .maxlen         = sizeof(int),
 474                 .mode           = 0644,
 475                 .proc_handler   = proc_dointvec
 476         },
 477         {
 478                 .procname       = "tcp_no_metrics_save",
 479                 .data           = &sysctl_tcp_nometrics_save,
 480                 .maxlen         = sizeof(int),
 481                 .mode           = 0644,
 482                 .proc_handler   = proc_dointvec,
 483         },
 484         {
 485                 .procname       = "tcp_moderate_rcvbuf",
 486                 .data           = &sysctl_tcp_moderate_rcvbuf,
 487                 .maxlen         = sizeof(int),
 488                 .mode           = 0644,
 489                 .proc_handler   = proc_dointvec,
 490         },
 491         {
 492                 .procname       = "tcp_tso_win_divisor",
 493                 .data           = &sysctl_tcp_tso_win_divisor,
 494                 .maxlen         = sizeof(int),
 495                 .mode           = 0644,
 496                 .proc_handler   = proc_dointvec,
 497         },
 498         {
 499                 .procname       = "tcp_congestion_control",
 500                 .mode           = 0644,
 501                 .maxlen         = TCP_CA_NAME_MAX,
 502                 .proc_handler   = proc_tcp_congestion_control,
 503         },
 504         {
 505                 .procname       = "tcp_workaround_signed_windows",
 506                 .data           = &sysctl_tcp_workaround_signed_windows,
 507                 .maxlen         = sizeof(int),
 508                 .mode           = 0644,
 509                 .proc_handler   = proc_dointvec
 510         },
 511         {
 512                 .procname       = "tcp_limit_output_bytes",
 513                 .data           = &sysctl_tcp_limit_output_bytes,
 514                 .maxlen         = sizeof(int),
 515                 .mode           = 0644,
 516                 .proc_handler   = proc_dointvec
 517         },
 518         {
 519                 .procname       = "tcp_challenge_ack_limit",
 520                 .data           = &sysctl_tcp_challenge_ack_limit,
 521                 .maxlen         = sizeof(int),
 522                 .mode           = 0644,
 523                 .proc_handler   = proc_dointvec
 524         },
 525         {
 526                 .procname       = "tcp_slow_start_after_idle",
 527                 .data           = &sysctl_tcp_slow_start_after_idle,
 528                 .maxlen         = sizeof(int),
 529                 .mode           = 0644,
 530                 .proc_handler   = proc_dointvec
 531         },
 532 #ifdef CONFIG_NETLABEL
 533         {
 534                 .procname       = "cipso_cache_enable",
 535                 .data           = &cipso_v4_cache_enabled,
 536                 .maxlen         = sizeof(int),
 537                 .mode           = 0644,
 538                 .proc_handler   = proc_dointvec,
 539         },
 540         {
 541                 .procname       = "cipso_cache_bucket_size",
 542                 .data           = &cipso_v4_cache_bucketsize,
 543                 .maxlen         = sizeof(int),
 544                 .mode           = 0644,
 545                 .proc_handler   = proc_dointvec,
 546         },
 547         {
 548                 .procname       = "cipso_rbm_optfmt",
 549                 .data           = &cipso_v4_rbm_optfmt,
 550                 .maxlen         = sizeof(int),
 551                 .mode           = 0644,
 552                 .proc_handler   = proc_dointvec,
 553         },
 554         {
 555                 .procname       = "cipso_rbm_strictvalid",
 556                 .data           = &cipso_v4_rbm_strictvalid,
 557                 .maxlen         = sizeof(int),
 558                 .mode           = 0644,
 559                 .proc_handler   = proc_dointvec,
 560         },
 561 #endif /* CONFIG_NETLABEL */
 562         {
 563                 .procname       = "tcp_available_congestion_control",
 564                 .maxlen         = TCP_CA_BUF_MAX,
 565                 .mode           = 0444,
 566                 .proc_handler   = proc_tcp_available_congestion_control,
 567         },
 568         {
 569                 .procname       = "tcp_allowed_congestion_control",
 570                 .maxlen         = TCP_CA_BUF_MAX,
 571                 .mode           = 0644,
 572                 .proc_handler   = proc_allowed_congestion_control,
 573         },
 574         {
 575                 .procname       = "tcp_thin_linear_timeouts",
 576                 .data           = &sysctl_tcp_thin_linear_timeouts,
 577                 .maxlen         = sizeof(int),
 578                 .mode           = 0644,
 579                 .proc_handler   = proc_dointvec
 580         },
 581         {
 582                 .procname       = "tcp_thin_dupack",
 583                 .data           = &sysctl_tcp_thin_dupack,
 584                 .maxlen         = sizeof(int),
 585                 .mode           = 0644,
 586                 .proc_handler   = proc_dointvec
 587         },
 588         {
 589                 .procname       = "tcp_early_retrans",
 590                 .data           = &sysctl_tcp_early_retrans,
 591                 .maxlen         = sizeof(int),
 592                 .mode           = 0644,
 593                 .proc_handler   = proc_dointvec_minmax,
 594                 .extra1         = &zero,
 595                 .extra2         = &four,
 596         },
 597         {
 598                 .procname       = "tcp_min_tso_segs",
 599                 .data           = &sysctl_tcp_min_tso_segs,
 600                 .maxlen         = sizeof(int),
 601                 .mode           = 0644,
 602                 .proc_handler   = proc_dointvec_minmax,
 603                 .extra1         = &one,
 604                 .extra2         = &gso_max_segs,
 605         },
 606         {
 607                 .procname       = "tcp_pacing_ss_ratio",
 608                 .data           = &sysctl_tcp_pacing_ss_ratio,
 609                 .maxlen         = sizeof(int),
 610                 .mode           = 0644,
 611                 .proc_handler   = proc_dointvec_minmax,
 612                 .extra1         = &zero,
 613                 .extra2         = &thousand,
 614         },
 615         {
 616                 .procname       = "tcp_pacing_ca_ratio",
 617                 .data           = &sysctl_tcp_pacing_ca_ratio,
 618                 .maxlen         = sizeof(int),
 619                 .mode           = 0644,
 620                 .proc_handler   = proc_dointvec_minmax,
 621                 .extra1         = &zero,
 622                 .extra2         = &thousand,
 623         },
 624         {
 625                 .procname       = "tcp_autocorking",
 626                 .data           = &sysctl_tcp_autocorking,
 627                 .maxlen         = sizeof(int),
 628                 .mode           = 0644,
 629                 .proc_handler   = proc_dointvec_minmax,
 630                 .extra1         = &zero,
 631                 .extra2         = &one,
 632         },
 633         {
 634                 .procname       = "tcp_invalid_ratelimit",
 635                 .data           = &sysctl_tcp_invalid_ratelimit,
 636                 .maxlen         = sizeof(int),
 637                 .mode           = 0644,
 638                 .proc_handler   = proc_dointvec_ms_jiffies,
 639         },
 640         {
 641                 .procname       = "icmp_msgs_per_sec",
 642                 .data           = &sysctl_icmp_msgs_per_sec,
 643                 .maxlen         = sizeof(int),
 644                 .mode           = 0644,
 645                 .proc_handler   = proc_dointvec_minmax,
 646                 .extra1         = &zero,
 647         },
 648         {
 649                 .procname       = "icmp_msgs_burst",
 650                 .data           = &sysctl_icmp_msgs_burst,
 651                 .maxlen         = sizeof(int),
 652                 .mode           = 0644,
 653                 .proc_handler   = proc_dointvec_minmax,
 654                 .extra1         = &zero,
 655         },
 656         {
 657                 .procname       = "udp_mem",
 658                 .data           = &sysctl_udp_mem,
 659                 .maxlen         = sizeof(sysctl_udp_mem),
 660                 .mode           = 0644,
 661                 .proc_handler   = proc_doulongvec_minmax,
 662         },
 663         {
 664                 .procname       = "udp_rmem_min",
 665                 .data           = &sysctl_udp_rmem_min,
 666                 .maxlen         = sizeof(sysctl_udp_rmem_min),
 667                 .mode           = 0644,
 668                 .proc_handler   = proc_dointvec_minmax,
 669                 .extra1         = &one
 670         },
 671         {
 672                 .procname       = "udp_wmem_min",
 673                 .data           = &sysctl_udp_wmem_min,
 674                 .maxlen         = sizeof(sysctl_udp_wmem_min),
 675                 .mode           = 0644,
 676                 .proc_handler   = proc_dointvec_minmax,
 677                 .extra1         = &one
 678         },
 679         { }
 680 };
 681
 682 static struct ctl_table ipv4_net_table[] = {
 683         {
 684                 .procname       = "icmp_echo_ignore_all",
 685                 .data           = &init_net.ipv4.sysctl_icmp_echo_ignore_all,
 686                 .maxlen         = sizeof(int),
 687                 .mode           = 0644,
 688                 .proc_handler   = proc_dointvec
 689         },
 690         {
 691                 .procname       = "icmp_echo_ignore_broadcasts",
 692                 .data           = &init_net.ipv4.sysctl_icmp_echo_ignore_broadcasts,
 693                 .maxlen         = sizeof(int),
 694                 .mode           = 0644,
 695                 .proc_handler   = proc_dointvec
 696         },
 697         {
 698                 .procname       = "icmp_ignore_bogus_error_responses",
 699                 .data           = &init_net.ipv4.sysctl_icmp_ignore_bogus_error_responses,
 700                 .maxlen         = sizeof(int),
 701                 .mode           = 0644,
 702                 .proc_handler   = proc_dointvec
 703         },
 704         {
 705                 .procname       = "icmp_errors_use_inbound_ifaddr",
 706                 .data           = &init_net.ipv4.sysctl_icmp_errors_use_inbound_ifaddr,
 707                 .maxlen         = sizeof(int),
 708                 .mode           = 0644,
 709                 .proc_handler   = proc_dointvec
 710         },
 711         {
 712                 .procname       = "icmp_ratelimit",
 713                 .data           = &init_net.ipv4.sysctl_icmp_ratelimit,
 714                 .maxlen         = sizeof(int),
 715                 .mode           = 0644,
 716                 .proc_handler   = proc_dointvec_ms_jiffies,
 717         },
 718         {
 719                 .procname       = "icmp_ratemask",
 720                 .data           = &init_net.ipv4.sysctl_icmp_ratemask,
 721                 .maxlen         = sizeof(int),
 722                 .mode           = 0644,
 723                 .proc_handler   = proc_dointvec
 724         },
 725         {
 726                 .procname       = "ping_group_range",
 727                 .data           = &init_net.ipv4.ping_group_range.range,
 728                 .maxlen         = sizeof(gid_t)*2,
 729                 .mode           = 0644,
 730                 .proc_handler   = ipv4_ping_group_range,
 731         },
 732         {
 733                 .procname       = "tcp_ecn",
 734                 .data           = &init_net.ipv4.sysctl_tcp_ecn,
 735                 .maxlen         = sizeof(int),
 736                 .mode           = 0644,
 737                 .proc_handler   = proc_dointvec
 738         },
 739         {
 740                 .procname       = "tcp_ecn_fallback",
 741                 .data           = &init_net.ipv4.sysctl_tcp_ecn_fallback,
 742                 .maxlen         = sizeof(int),
 743                 .mode           = 0644,
 744                 .proc_handler   = proc_dointvec
 745         },
 746         {
 747                 .procname       = "ip_dynaddr",
 748                 .data           = &init_net.ipv4.sysctl_ip_dynaddr,
 749                 .maxlen         = sizeof(int),
 750                 .mode           = 0644,
 751                 .proc_handler   = proc_dointvec
 752         },
 753         {
 754                 .procname       = "ip_early_demux",
 755                 .data           = &init_net.ipv4.sysctl_ip_early_demux,
 756                 .maxlen         = sizeof(int),
 757                 .mode           = 0644,
 758                 .proc_handler   = proc_dointvec
 759         },
 760         {
 761                 .procname       = "ip_default_ttl",
 762                 .data           = &init_net.ipv4.sysctl_ip_default_ttl,
 763                 .maxlen         = sizeof(int),
 764                 .mode           = 0644,
 765                 .proc_handler   = proc_dointvec_minmax,
 766                 .extra1         = &ip_ttl_min,
 767                 .extra2         = &ip_ttl_max,
 768         },
 769         {
 770                 .procname       = "ip_local_port_range",
 771                 .maxlen         = sizeof(init_net.ipv4.ip_local_ports.range),
 772                 .data           = &init_net.ipv4.ip_local_ports.range,
 773                 .mode           = 0644,
 774                 .proc_handler   = ipv4_local_port_range,
 775         },
 776         {
 777                 .procname       = "ip_local_reserved_ports",
 778                 .data           = &init_net.ipv4.sysctl_local_reserved_ports,
 779                 .maxlen         = 65536,
 780                 .mode           = 0644,
 781                 .proc_handler   = proc_do_large_bitmap,
 782         },
 783         {
 784                 .procname       = "ip_no_pmtu_disc",
 785                 .data           = &init_net.ipv4.sysctl_ip_no_pmtu_disc,
 786                 .maxlen         = sizeof(int),
 787                 .mode           = 0644,
 788                 .proc_handler   = proc_dointvec
 789         },
 790         {
 791                 .procname       = "ip_forward_use_pmtu",
 792                 .data           = &init_net.ipv4.sysctl_ip_fwd_use_pmtu,
 793                 .maxlen         = sizeof(int),
 794                 .mode           = 0644,
 795                 .proc_handler   = proc_dointvec,
 796         },
 797         {
 798                 .procname       = "ip_nonlocal_bind",
 799                 .data           = &init_net.ipv4.sysctl_ip_nonlocal_bind,
 800                 .maxlen         = sizeof(int),
 801                 .mode           = 0644,
 802                 .proc_handler   = proc_dointvec
 803         },
 804         {
 805                 .procname       = "fwmark_reflect",
 806                 .data           = &init_net.ipv4.sysctl_fwmark_reflect,
 807                 .maxlen         = sizeof(int),
 808                 .mode           = 0644,
 809                 .proc_handler   = proc_dointvec,
 810         },
 811         {
 812                 .procname       = "tcp_fwmark_accept",
 813                 .data           = &init_net.ipv4.sysctl_tcp_fwmark_accept,
 814                 .maxlen         = sizeof(int),
 815                 .mode           = 0644,
 816                 .proc_handler   = proc_dointvec,
 817         },
 818 #ifdef CONFIG_NET_L3_MASTER_DEV
 819         {
 820                 .procname       = "tcp_l3mdev_accept",
 821                 .data           = &init_net.ipv4.sysctl_tcp_l3mdev_accept,
 822                 .maxlen         = sizeof(int),
 823                 .mode           = 0644,
 824                 .proc_handler   = proc_dointvec_minmax,
 825                 .extra1         = &zero,
 826                 .extra2         = &one,
 827         },
 828 #endif
 829         {
 830                 .procname       = "tcp_mtu_probing",
 831                 .data           = &init_net.ipv4.sysctl_tcp_mtu_probing,
 832                 .maxlen         = sizeof(int),
 833                 .mode           = 0644,
 834                 .proc_handler   = proc_dointvec,
 835         },
 836         {
 837                 .procname       = "tcp_base_mss",
 838                 .data           = &init_net.ipv4.sysctl_tcp_base_mss,
 839                 .maxlen         = sizeof(int),
 840                 .mode           = 0644,
 841                 .proc_handler   = proc_dointvec,
 842         },
 843         {
 844                 .procname       = "tcp_min_snd_mss",
 845                 .data           = &init_net.ipv4.sysctl_tcp_min_snd_mss,
 846                 .maxlen         = sizeof(int),
 847                 .mode           = 0644,
 848                 .proc_handler   = proc_dointvec_minmax,
 849                 .extra1         = &tcp_min_snd_mss_min,
 850                 .extra2         = &tcp_min_snd_mss_max,
 851         },
 852         {
 853                 .procname       = "tcp_probe_threshold",
 854                 .data           = &init_net.ipv4.sysctl_tcp_probe_threshold,
 855                 .maxlen         = sizeof(int),
 856                 .mode           = 0644,
 857                 .proc_handler   = proc_dointvec,
 858         },
 859         {
 860                 .procname       = "tcp_probe_interval",
 861                 .data           = &init_net.ipv4.sysctl_tcp_probe_interval,
 862                 .maxlen         = sizeof(int),
 863                 .mode           = 0644,
 864                 .proc_handler   = proc_dointvec,
 865         },
 866         {
 867                 .procname       = "igmp_link_local_mcast_reports",
 868                 .data           = &init_net.ipv4.sysctl_igmp_llm_reports,
 869                 .maxlen         = sizeof(int),
 870                 .mode           = 0644,
 871                 .proc_handler   = proc_dointvec
 872         },
 873         {
 874                 .procname       = "igmp_max_memberships",
 875                 .data           = &init_net.ipv4.sysctl_igmp_max_memberships,
 876                 .maxlen         = sizeof(int),
 877                 .mode           = 0644,
 878                 .proc_handler   = proc_dointvec
 879         },
 880         {
 881                 .procname       = "igmp_max_msf",
 882                 .data           = &init_net.ipv4.sysctl_igmp_max_msf,
 883                 .maxlen         = sizeof(int),
 884                 .mode           = 0644,
 885                 .proc_handler   = proc_dointvec
 886         },
 887 #ifdef CONFIG_IP_MULTICAST
 888         {
 889                 .procname       = "igmp_qrv",
 890                 .data           = &init_net.ipv4.sysctl_igmp_qrv,
 891                 .maxlen         = sizeof(int),
 892                 .mode           = 0644,
 893                 .proc_handler   = proc_dointvec_minmax,
 894                 .extra1         = &one
 895         },
 896 #endif
 897         {
 898                 .procname       = "tcp_keepalive_time",
 899                 .data           = &init_net.ipv4.sysctl_tcp_keepalive_time,
 900                 .maxlen         = sizeof(int),
 901                 .mode           = 0644,
 902                 .proc_handler   = proc_dointvec_jiffies,
 903         },
 904         {
 905                 .procname       = "tcp_keepalive_probes",
 906                 .data           = &init_net.ipv4.sysctl_tcp_keepalive_probes,
 907                 .maxlen         = sizeof(int),
 908                 .mode           = 0644,
 909                 .proc_handler   = proc_dointvec
 910         },
 911         {
 912                 .procname       = "tcp_keepalive_intvl",
 913                 .data           = &init_net.ipv4.sysctl_tcp_keepalive_intvl,
 914                 .maxlen         = sizeof(int),
 915                 .mode           = 0644,
 916                 .proc_handler   = proc_dointvec_jiffies,
 917         },
 918         {
 919                 .procname       = "tcp_syn_retries",
 920                 .data           = &init_net.ipv4.sysctl_tcp_syn_retries,
 921                 .maxlen         = sizeof(int),
 922                 .mode           = 0644,
 923                 .proc_handler   = proc_dointvec_minmax,
 924                 .extra1         = &tcp_syn_retries_min,
 925                 .extra2         = &tcp_syn_retries_max
 926         },
 927         {
 928                 .procname       = "tcp_synack_retries",
 929                 .data           = &init_net.ipv4.sysctl_tcp_synack_retries,
 930                 .maxlen         = sizeof(int),
 931                 .mode           = 0644,
 932                 .proc_handler   = proc_dointvec
 933         },
 934 #ifdef CONFIG_SYN_COOKIES
 935         {
 936                 .procname       = "tcp_syncookies",
 937                 .data           = &init_net.ipv4.sysctl_tcp_syncookies,
 938                 .maxlen         = sizeof(int),
 939                 .mode           = 0644,
 940                 .proc_handler   = proc_dointvec
 941         },
 942 #endif
 943         {
 944                 .procname       = "tcp_reordering",
 945                 .data           = &init_net.ipv4.sysctl_tcp_reordering,
 946                 .maxlen         = sizeof(int),
 947                 .mode           = 0644,
 948                 .proc_handler   = proc_dointvec
 949         },
 950         {
 951                 .procname       = "tcp_retries1",
 952                 .data           = &init_net.ipv4.sysctl_tcp_retries1,
 953                 .maxlen         = sizeof(int),
 954                 .mode           = 0644,
 955                 .proc_handler   = proc_dointvec_minmax,
 956                 .extra2         = &tcp_retr1_max
 957         },
 958         {
 959                 .procname       = "tcp_retries2",
 960                 .data           = &init_net.ipv4.sysctl_tcp_retries2,
 961                 .maxlen         = sizeof(int),
 962                 .mode           = 0644,
 963                 .proc_handler   = proc_dointvec
 964         },
 965         {
 966                 .procname       = "tcp_orphan_retries",
 967                 .data           = &init_net.ipv4.sysctl_tcp_orphan_retries,
 968                 .maxlen         = sizeof(int),
 969                 .mode           = 0644,
 970                 .proc_handler   = proc_dointvec
 971         },
 972         {
 973                 .procname       = "tcp_fin_timeout",
 974                 .data           = &init_net.ipv4.sysctl_tcp_fin_timeout,
 975                 .maxlen         = sizeof(int),
 976                 .mode           = 0644,
 977                 .proc_handler   = proc_dointvec_jiffies,
 978         },
 979         {
 980                 .procname       = "tcp_notsent_lowat",
 981                 .data           = &init_net.ipv4.sysctl_tcp_notsent_lowat,
 982                 .maxlen         = sizeof(unsigned int),
 983                 .mode           = 0644,
 984                 .proc_handler   = proc_douintvec,
 985         },
 986 #ifdef CONFIG_IP_ROUTE_MULTIPATH
 987         {
 988                 .procname       = "fib_multipath_use_neigh",
 989                 .data           = &init_net.ipv4.sysctl_fib_multipath_use_neigh,
 990                 .maxlen         = sizeof(int),
 991                 .mode           = 0644,
 992                 .proc_handler   = proc_dointvec_minmax,
 993                 .extra1         = &zero,
 994                 .extra2         = &one,
 995         },
 996 #endif
 997         { }
 998 };
 999
1000 static __net_init int ipv4_sysctl_init_net(struct net *net)
1001 {
1002         struct ctl_table *table;
1003
1004         table = ipv4_net_table;
1005         if (!net_eq(net, &init_net)) {
1006                 int i;
1007
1008                 table = kmemdup(table, sizeof(ipv4_net_table), GFP_KERNEL);
1009                 if (!table)
1010                         goto err_alloc;
1011
1012                 /* Update the variables to point into the current struct net */
1013                 for (i = 0; i < ARRAY_SIZE(ipv4_net_table) - 1; i++)
1014                         table[i].data += (void *)net - (void *)&init_net;
1015         }
1016
1017         net->ipv4.ipv4_hdr = register_net_sysctl(net, "net/ipv4", table);
1018         if (!net->ipv4.ipv4_hdr)
1019                 goto err_reg;
1020
1021         net->ipv4.sysctl_local_reserved_ports = kzalloc(65536 / 8, GFP_KERNEL);
1022         if (!net->ipv4.sysctl_local_reserved_ports)
1023                 goto err_ports;
1024
1025         return 0;
1026
1027 err_ports:
1028         unregister_net_sysctl_table(net->ipv4.ipv4_hdr);
1029 err_reg:
1030         if (!net_eq(net, &init_net))
1031                 kfree(table);
1032 err_alloc:
1033         return -ENOMEM;
1034 }
1035
1036 static __net_exit void ipv4_sysctl_exit_net(struct net *net)
1037 {
1038         struct ctl_table *table;
1039
1040         kfree(net->ipv4.sysctl_local_reserved_ports);
1041         table = net->ipv4.ipv4_hdr->ctl_table_arg;
1042         unregister_net_sysctl_table(net->ipv4.ipv4_hdr);
1043         kfree(table);
1044 }
1045
1046 static __net_initdata struct pernet_operations ipv4_sysctl_ops = {
1047         .init = ipv4_sysctl_init_net,
1048         .exit = ipv4_sysctl_exit_net,
1049 };
1050
1051 static __init int sysctl_ipv4_init(void)
1052 {
1053         struct ctl_table_header *hdr;
1054
1055         hdr = register_net_sysctl(&init_net, "net/ipv4", ipv4_table);
1056         if (!hdr)
1057                 return -ENOMEM;
1058
1059         if (register_pernet_subsys(&ipv4_sysctl_ops)) {
1060                 unregister_net_sysctl_table(hdr);
1061                 return -ENOMEM;
1062         }
1063
1064         return 0;
1065 }
1066
1067 __initcall(sysctl_ipv4_init);