net/ipv4/sysctl_net_ipv4.c

   1 // SPDX-License-Identifier: GPL-2.0
   2 /*
   3  * sysctl_net_ipv4.c: sysctl interface to net IPV4 subsystem.
   4  *
   5  * Begun April 1, 1996, Mike Shaver.
   6  * Added /proc/sys/net/ipv4 directory entry (empty =) ). [MS]
   7  */
   8
   9 #include <linux/mm.h>
  10 #include <linux/module.h>
  11 #include <linux/sysctl.h>
  12 #include <linux/igmp.h>
  13 #include <linux/inetdevice.h>
  14 #include <linux/seqlock.h>
  15 #include <linux/init.h>
  16 #include <linux/slab.h>
  17 #include <linux/nsproxy.h>
  18 #include <linux/swap.h>
  19 #include <net/snmp.h>
  20 #include <net/icmp.h>
  21 #include <net/ip.h>
  22 #include <net/route.h>
  23 #include <net/tcp.h>
  24 #include <net/udp.h>
  25 #include <net/cipso_ipv4.h>
  26 #include <net/inet_frag.h>
  27 #include <net/ping.h>
  28 #include <net/protocol.h>
  29
  30 static int zero;
  31 static int one = 1;
  32 static int four = 4;
  33 static int thousand = 1000;
  34 static int gso_max_segs = GSO_MAX_SEGS;
  35 static int tcp_retr1_max = 255;
  36 static int ip_local_port_range_min[] = { 1, 1 };
  37 static int ip_local_port_range_max[] = { 65535, 65535 };
  38 static int tcp_adv_win_scale_min = -31;
  39 static int tcp_adv_win_scale_max = 31;
  40 static int tcp_min_snd_mss_min = TCP_MIN_SND_MSS;
  41 static int tcp_min_snd_mss_max = 65535;
  42 static int ip_privileged_port_min;
  43 static int ip_privileged_port_max = 65535;
  44 static int ip_ttl_min = 1;
  45 static int ip_ttl_max = 255;
  46 static int tcp_syn_retries_min = 1;
  47 static int tcp_syn_retries_max = MAX_TCP_SYNCNT;
  48 static int ip_ping_group_range_min[] = { 0, 0 };
  49 static int ip_ping_group_range_max[] = { GID_T_MAX, GID_T_MAX };
  50 static int one_day_secs = 24 * 3600;
  51
  52 /* obsolete */
  53 static int sysctl_tcp_low_latency __read_mostly;
  54
  55 /* Update system visible IP port range */
  56 static void set_local_port_range(struct net *net, int range[2])
  57 {
  58         bool same_parity = !((range[0] ^ range[1]) & 1);
  59
  60         write_seqlock_bh(&net->ipv4.ip_local_ports.lock);
  61         if (same_parity && !net->ipv4.ip_local_ports.warned) {
  62                 net->ipv4.ip_local_ports.warned = true;
  63                 pr_err_ratelimited("ip_local_port_range: prefer different parity for start/end values.\n");
  64         }
  65         net->ipv4.ip_local_ports.range[0] = range[0];
  66         net->ipv4.ip_local_ports.range[1] = range[1];
  67         write_sequnlock_bh(&net->ipv4.ip_local_ports.lock);
  68 }
  69
  70 /* Validate changes from /proc interface. */
  71 static int ipv4_local_port_range(struct ctl_table *table, int write,
  72                                  void __user *buffer,
  73                                  size_t *lenp, loff_t *ppos)
  74 {
  75         struct net *net =
  76                 container_of(table->data, struct net, ipv4.ip_local_ports.range);
  77         int ret;
  78         int range[2];
  79         struct ctl_table tmp = {
  80                 .data = &range,
  81                 .maxlen = sizeof(range),
  82                 .mode = table->mode,
  83                 .extra1 = &ip_local_port_range_min,
  84                 .extra2 = &ip_local_port_range_max,
  85         };
  86
  87         inet_get_local_port_range(net, &range[0], &range[1]);
  88
  89         ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
  90
  91         if (write && ret == 0) {
  92                 /* Ensure that the upper limit is not smaller than the lower,
  93                  * and that the lower does not encroach upon the privileged
  94                  * port limit.
  95                  */
  96                 if ((range[1] < range[0]) ||
  97                     (range[0] < net->ipv4.sysctl_ip_prot_sock))
  98                         ret = -EINVAL;
  99                 else
 100                         set_local_port_range(net, range);
 101         }
 102
 103         return ret;
 104 }
 105
 106 /* Validate changes from /proc interface. */
 107 static int ipv4_privileged_ports(struct ctl_table *table, int write,
 108                                 void __user *buffer, size_t *lenp, loff_t *ppos)
 109 {
 110         struct net *net = container_of(table->data, struct net,
 111             ipv4.sysctl_ip_prot_sock);
 112         int ret;
 113         int pports;
 114         int range[2];
 115         struct ctl_table tmp = {
 116                 .data = &pports,
 117                 .maxlen = sizeof(pports),
 118                 .mode = table->mode,
 119                 .extra1 = &ip_privileged_port_min,
 120                 .extra2 = &ip_privileged_port_max,
 121         };
 122
 123         pports = net->ipv4.sysctl_ip_prot_sock;
 124
 125         ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
 126
 127         if (write && ret == 0) {
 128                 inet_get_local_port_range(net, &range[0], &range[1]);
 129                 /* Ensure that the local port range doesn't overlap with the
 130                  * privileged port range.
 131                  */
 132                 if (range[0] < pports)
 133                         ret = -EINVAL;
 134                 else
 135                         net->ipv4.sysctl_ip_prot_sock = pports;
 136         }
 137
 138         return ret;
 139 }
 140
 141 static void inet_get_ping_group_range_table(struct ctl_table *table, kgid_t *low, kgid_t *high)
 142 {
 143         kgid_t *data = table->data;
 144         struct net *net =
 145                 container_of(table->data, struct net, ipv4.ping_group_range.range);
 146         unsigned int seq;
 147         do {
 148                 seq = read_seqbegin(&net->ipv4.ping_group_range.lock);
 149
 150                 *low = data[0];
 151                 *high = data[1];
 152         } while (read_seqretry(&net->ipv4.ping_group_range.lock, seq));
 153 }
 154
 155 /* Update system visible IP port range */
 156 static void set_ping_group_range(struct ctl_table *table, kgid_t low, kgid_t high)
 157 {
 158         kgid_t *data = table->data;
 159         struct net *net =
 160                 container_of(table->data, struct net, ipv4.ping_group_range.range);
 161         write_seqlock(&net->ipv4.ping_group_range.lock);
 162         data[0] = low;
 163         data[1] = high;
 164         write_sequnlock(&net->ipv4.ping_group_range.lock);
 165 }
 166
 167 /* Validate changes from /proc interface. */
 168 static int ipv4_ping_group_range(struct ctl_table *table, int write,
 169                                  void __user *buffer,
 170                                  size_t *lenp, loff_t *ppos)
 171 {
 172         struct user_namespace *user_ns = current_user_ns();
 173         int ret;
 174         gid_t urange[2];
 175         kgid_t low, high;
 176         struct ctl_table tmp = {
 177                 .data = &urange,
 178                 .maxlen = sizeof(urange),
 179                 .mode = table->mode,
 180                 .extra1 = &ip_ping_group_range_min,
 181                 .extra2 = &ip_ping_group_range_max,
 182         };
 183
 184         inet_get_ping_group_range_table(table, &low, &high);
 185         urange[0] = from_kgid_munged(user_ns, low);
 186         urange[1] = from_kgid_munged(user_ns, high);
 187         ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
 188
 189         if (write && ret == 0) {
 190                 low = make_kgid(user_ns, urange[0]);
 191                 high = make_kgid(user_ns, urange[1]);
 192                 if (!gid_valid(low) || !gid_valid(high))
 193                         return -EINVAL;
 194                 if (urange[1] < urange[0] || gid_lt(high, low)) {
 195                         low = make_kgid(&init_user_ns, 1);
 196                         high = make_kgid(&init_user_ns, 0);
 197                 }
 198                 set_ping_group_range(table, low, high);
 199         }
 200
 201         return ret;
 202 }
 203
 204 static int proc_tcp_congestion_control(struct ctl_table *ctl, int write,
 205                                        void __user *buffer, size_t *lenp, loff_t *ppos)
 206 {
 207         char val[TCP_CA_NAME_MAX];
 208         struct ctl_table tbl = {
 209                 .data = val,
 210                 .maxlen = TCP_CA_NAME_MAX,
 211         };
 212         int ret;
 213
 214         tcp_get_default_congestion_control(val);
 215
 216         ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
 217         if (write && ret == 0)
 218                 ret = tcp_set_default_congestion_control(val);
 219         return ret;
 220 }
 221
 222 static int proc_tcp_available_congestion_control(struct ctl_table *ctl,
 223                                                  int write,
 224                                                  void __user *buffer, size_t *lenp,
 225                                                  loff_t *ppos)
 226 {
 227         struct ctl_table tbl = { .maxlen = TCP_CA_BUF_MAX, };
 228         int ret;
 229
 230         tbl.data = kmalloc(tbl.maxlen, GFP_USER);
 231         if (!tbl.data)
 232                 return -ENOMEM;
 233         tcp_get_available_congestion_control(tbl.data, TCP_CA_BUF_MAX);
 234         ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
 235         kfree(tbl.data);
 236         return ret;
 237 }
 238
 239 static int proc_allowed_congestion_control(struct ctl_table *ctl,
 240                                            int write,
 241                                            void __user *buffer, size_t *lenp,
 242                                            loff_t *ppos)
 243 {
 244         struct ctl_table tbl = { .maxlen = TCP_CA_BUF_MAX };
 245         int ret;
 246
 247         tbl.data = kmalloc(tbl.maxlen, GFP_USER);
 248         if (!tbl.data)
 249                 return -ENOMEM;
 250
 251         tcp_get_allowed_congestion_control(tbl.data, tbl.maxlen);
 252         ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
 253         if (write && ret == 0)
 254                 ret = tcp_set_allowed_congestion_control(tbl.data);
 255         kfree(tbl.data);
 256         return ret;
 257 }
 258
 259 static int proc_tcp_fastopen_key(struct ctl_table *ctl, int write,
 260                                  void __user *buffer, size_t *lenp,
 261                                  loff_t *ppos)
 262 {
 263         struct ctl_table tbl = { .maxlen = (TCP_FASTOPEN_KEY_LENGTH * 2 + 10) };
 264         struct tcp_fastopen_context *ctxt;
 265         u32  user_key[4]; /* 16 bytes, matching TCP_FASTOPEN_KEY_LENGTH */
 266         __le32 key[4];
 267         int ret, i;
 268
 269         tbl.data = kmalloc(tbl.maxlen, GFP_KERNEL);
 270         if (!tbl.data)
 271                 return -ENOMEM;
 272
 273         rcu_read_lock();
 274         ctxt = rcu_dereference(tcp_fastopen_ctx);
 275         if (ctxt)
 276                 memcpy(key, ctxt->key, TCP_FASTOPEN_KEY_LENGTH);
 277         else
 278                 memset(key, 0, sizeof(key));
 279         rcu_read_unlock();
 280
 281         for (i = 0; i < ARRAY_SIZE(key); i++)
 282                 user_key[i] = le32_to_cpu(key[i]);
 283
 284         snprintf(tbl.data, tbl.maxlen, "%08x-%08x-%08x-%08x",
 285                 user_key[0], user_key[1], user_key[2], user_key[3]);
 286         ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
 287
 288         if (write && ret == 0) {
 289                 if (sscanf(tbl.data, "%x-%x-%x-%x", user_key, user_key + 1,
 290                            user_key + 2, user_key + 3) != 4) {
 291                         ret = -EINVAL;
 292                         goto bad_key;
 293                 }
 294                 /* Generate a dummy secret but don't publish it. This
 295                  * is needed so we don't regenerate a new key on the
 296                  * first invocation of tcp_fastopen_cookie_gen
 297                  */
 298                 tcp_fastopen_init_key_once(false);
 299
 300                 for (i = 0; i < ARRAY_SIZE(user_key); i++)
 301                         key[i] = cpu_to_le32(user_key[i]);
 302
 303                 tcp_fastopen_reset_cipher(key, TCP_FASTOPEN_KEY_LENGTH);
 304         }
 305
 306 bad_key:
 307         pr_debug("proc FO key set 0x%x-%x-%x-%x <- 0x%s: %u\n",
 308                  user_key[0], user_key[1], user_key[2], user_key[3],
 309                (char *)tbl.data, ret);
 310         kfree(tbl.data);
 311         return ret;
 312 }
 313
 314 static int proc_tfo_blackhole_detect_timeout(struct ctl_table *table,
 315                                              int write,
 316                                              void __user *buffer,
 317                                              size_t *lenp, loff_t *ppos)
 318 {
 319         int ret;
 320
 321         ret = proc_dointvec_minmax(table, write, buffer, lenp, ppos);
 322         if (write && ret == 0)
 323                 tcp_fastopen_active_timeout_reset();
 324
 325         return ret;
 326 }
 327
 328 static int proc_tcp_available_ulp(struct ctl_table *ctl,
 329                                   int write,
 330                                   void __user *buffer, size_t *lenp,
 331                                   loff_t *ppos)
 332 {
 333         struct ctl_table tbl = { .maxlen = TCP_ULP_BUF_MAX, };
 334         int ret;
 335
 336         tbl.data = kmalloc(tbl.maxlen, GFP_USER);
 337         if (!tbl.data)
 338                 return -ENOMEM;
 339         tcp_get_available_ulp(tbl.data, TCP_ULP_BUF_MAX);
 340         ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
 341         kfree(tbl.data);
 342
 343         return ret;
 344 }
 345
 346 static struct ctl_table ipv4_table[] = {
 347         {
 348                 .procname       = "tcp_retrans_collapse",
 349                 .data           = &sysctl_tcp_retrans_collapse,
 350                 .maxlen         = sizeof(int),
 351                 .mode           = 0644,
 352                 .proc_handler   = proc_dointvec
 353         },
 354         {
 355                 .procname       = "tcp_max_orphans",
 356                 .data           = &sysctl_tcp_max_orphans,
 357                 .maxlen         = sizeof(int),
 358                 .mode           = 0644,
 359                 .proc_handler   = proc_dointvec
 360         },
 361         {
 362                 .procname       = "tcp_fastopen",
 363                 .data           = &sysctl_tcp_fastopen,
 364                 .maxlen         = sizeof(int),
 365                 .mode           = 0644,
 366                 .proc_handler   = proc_dointvec,
 367         },
 368         {
 369                 .procname       = "tcp_fastopen_key",
 370                 .mode           = 0600,
 371                 .maxlen         = ((TCP_FASTOPEN_KEY_LENGTH * 2) + 10),
 372                 .proc_handler   = proc_tcp_fastopen_key,
 373         },
 374         {
 375                 .procname       = "tcp_fastopen_blackhole_timeout_sec",
 376                 .data           = &sysctl_tcp_fastopen_blackhole_timeout,
 377                 .maxlen         = sizeof(int),
 378                 .mode           = 0644,
 379                 .proc_handler   = proc_tfo_blackhole_detect_timeout,
 380                 .extra1         = &zero,
 381         },
 382         {
 383                 .procname       = "tcp_abort_on_overflow",
 384                 .data           = &sysctl_tcp_abort_on_overflow,
 385                 .maxlen         = sizeof(int),
 386                 .mode           = 0644,
 387                 .proc_handler   = proc_dointvec
 388         },
 389         {
 390                 .procname       = "tcp_stdurg",
 391                 .data           = &sysctl_tcp_stdurg,
 392                 .maxlen         = sizeof(int),
 393                 .mode           = 0644,
 394                 .proc_handler   = proc_dointvec
 395         },
 396         {
 397                 .procname       = "tcp_rfc1337",
 398                 .data           = &sysctl_tcp_rfc1337,
 399                 .maxlen         = sizeof(int),
 400                 .mode           = 0644,
 401                 .proc_handler   = proc_dointvec
 402         },
 403         {
 404                 .procname       = "inet_peer_threshold",
 405                 .data           = &inet_peer_threshold,
 406                 .maxlen         = sizeof(int),
 407                 .mode           = 0644,
 408                 .proc_handler   = proc_dointvec
 409         },
 410         {
 411                 .procname       = "inet_peer_minttl",
 412                 .data           = &inet_peer_minttl,
 413                 .maxlen         = sizeof(int),
 414                 .mode           = 0644,
 415                 .proc_handler   = proc_dointvec_jiffies,
 416         },
 417         {
 418                 .procname       = "inet_peer_maxttl",
 419                 .data           = &inet_peer_maxttl,
 420                 .maxlen         = sizeof(int),
 421                 .mode           = 0644,
 422                 .proc_handler   = proc_dointvec_jiffies,
 423         },
 424         {
 425                 .procname       = "tcp_fack",
 426                 .data           = &sysctl_tcp_fack,
 427                 .maxlen         = sizeof(int),
 428                 .mode           = 0644,
 429                 .proc_handler   = proc_dointvec
 430         },
 431         {
 432                 .procname       = "tcp_recovery",
 433                 .data           = &sysctl_tcp_recovery,
 434                 .maxlen         = sizeof(int),
 435                 .mode           = 0644,
 436                 .proc_handler   = proc_dointvec,
 437         },
 438         {
 439                 .procname       = "tcp_max_reordering",
 440                 .data           = &sysctl_tcp_max_reordering,
 441                 .maxlen         = sizeof(int),
 442                 .mode           = 0644,
 443                 .proc_handler   = proc_dointvec
 444         },
 445         {
 446                 .procname       = "tcp_dsack",
 447                 .data           = &sysctl_tcp_dsack,
 448                 .maxlen         = sizeof(int),
 449                 .mode           = 0644,
 450                 .proc_handler   = proc_dointvec
 451         },
 452         {
 453                 .procname       = "tcp_mem",
 454                 .maxlen         = sizeof(sysctl_tcp_mem),
 455                 .data           = &sysctl_tcp_mem,
 456                 .mode           = 0644,
 457                 .proc_handler   = proc_doulongvec_minmax,
 458         },
 459         {
 460                 .procname       = "tcp_wmem",
 461                 .data           = &sysctl_tcp_wmem,
 462                 .maxlen         = sizeof(sysctl_tcp_wmem),
 463                 .mode           = 0644,
 464                 .proc_handler   = proc_dointvec_minmax,
 465                 .extra1         = &one,
 466         },
 467         {
 468                 .procname       = "tcp_rmem",
 469                 .data           = &sysctl_tcp_rmem,
 470                 .maxlen         = sizeof(sysctl_tcp_rmem),
 471                 .mode           = 0644,
 472                 .proc_handler   = proc_dointvec_minmax,
 473                 .extra1         = &one,
 474         },
 475         {
 476                 .procname       = "tcp_app_win",
 477                 .data           = &sysctl_tcp_app_win,
 478                 .maxlen         = sizeof(int),
 479                 .mode           = 0644,
 480                 .proc_handler   = proc_dointvec
 481         },
 482         {
 483                 .procname       = "tcp_adv_win_scale",
 484                 .data           = &sysctl_tcp_adv_win_scale,
 485                 .maxlen         = sizeof(int),
 486                 .mode           = 0644,
 487                 .proc_handler   = proc_dointvec_minmax,
 488                 .extra1         = &tcp_adv_win_scale_min,
 489                 .extra2         = &tcp_adv_win_scale_max,
 490         },
 491         {
 492                 .procname       = "tcp_frto",
 493                 .data           = &sysctl_tcp_frto,
 494                 .maxlen         = sizeof(int),
 495                 .mode           = 0644,
 496                 .proc_handler   = proc_dointvec
 497         },
 498         {
 499                 .procname       = "tcp_min_rtt_wlen",
 500                 .data           = &sysctl_tcp_min_rtt_wlen,
 501                 .maxlen         = sizeof(int),
 502                 .mode           = 0644,
 503                 .proc_handler   = proc_dointvec_minmax,
 504                 .extra1         = &zero,
 505                 .extra2         = &one_day_secs
 506         },
 507         {
 508                 .procname       = "tcp_low_latency",
 509                 .data           = &sysctl_tcp_low_latency,
 510                 .maxlen         = sizeof(int),
 511                 .mode           = 0644,
 512                 .proc_handler   = proc_dointvec
 513         },
 514         {
 515                 .procname       = "tcp_no_metrics_save",
 516                 .data           = &sysctl_tcp_nometrics_save,
 517                 .maxlen         = sizeof(int),
 518                 .mode           = 0644,
 519                 .proc_handler   = proc_dointvec,
 520         },
 521         {
 522                 .procname       = "tcp_moderate_rcvbuf",
 523                 .data           = &sysctl_tcp_moderate_rcvbuf,
 524                 .maxlen         = sizeof(int),
 525                 .mode           = 0644,
 526                 .proc_handler   = proc_dointvec,
 527         },
 528         {
 529                 .procname       = "tcp_tso_win_divisor",
 530                 .data           = &sysctl_tcp_tso_win_divisor,
 531                 .maxlen         = sizeof(int),
 532                 .mode           = 0644,
 533                 .proc_handler   = proc_dointvec,
 534         },
 535         {
 536                 .procname       = "tcp_congestion_control",
 537                 .mode           = 0644,
 538                 .maxlen         = TCP_CA_NAME_MAX,
 539                 .proc_handler   = proc_tcp_congestion_control,
 540         },
 541         {
 542                 .procname       = "tcp_workaround_signed_windows",
 543                 .data           = &sysctl_tcp_workaround_signed_windows,
 544                 .maxlen         = sizeof(int),
 545                 .mode           = 0644,
 546                 .proc_handler   = proc_dointvec
 547         },
 548         {
 549                 .procname       = "tcp_limit_output_bytes",
 550                 .data           = &sysctl_tcp_limit_output_bytes,
 551                 .maxlen         = sizeof(int),
 552                 .mode           = 0644,
 553                 .proc_handler   = proc_dointvec
 554         },
 555         {
 556                 .procname       = "tcp_challenge_ack_limit",
 557                 .data           = &sysctl_tcp_challenge_ack_limit,
 558                 .maxlen         = sizeof(int),
 559                 .mode           = 0644,
 560                 .proc_handler   = proc_dointvec
 561         },
 562         {
 563                 .procname       = "tcp_slow_start_after_idle",
 564                 .data           = &sysctl_tcp_slow_start_after_idle,
 565                 .maxlen         = sizeof(int),
 566                 .mode           = 0644,
 567                 .proc_handler   = proc_dointvec
 568         },
 569 #ifdef CONFIG_NETLABEL
 570         {
 571                 .procname       = "cipso_cache_enable",
 572                 .data           = &cipso_v4_cache_enabled,
 573                 .maxlen         = sizeof(int),
 574                 .mode           = 0644,
 575                 .proc_handler   = proc_dointvec,
 576         },
 577         {
 578                 .procname       = "cipso_cache_bucket_size",
 579                 .data           = &cipso_v4_cache_bucketsize,
 580                 .maxlen         = sizeof(int),
 581                 .mode           = 0644,
 582                 .proc_handler   = proc_dointvec,
 583         },
 584         {
 585                 .procname       = "cipso_rbm_optfmt",
 586                 .data           = &cipso_v4_rbm_optfmt,
 587                 .maxlen         = sizeof(int),
 588                 .mode           = 0644,
 589                 .proc_handler   = proc_dointvec,
 590         },
 591         {
 592                 .procname       = "cipso_rbm_strictvalid",
 593                 .data           = &cipso_v4_rbm_strictvalid,
 594                 .maxlen         = sizeof(int),
 595                 .mode           = 0644,
 596                 .proc_handler   = proc_dointvec,
 597         },
 598 #endif /* CONFIG_NETLABEL */
 599         {
 600                 .procname       = "tcp_available_congestion_control",
 601                 .maxlen         = TCP_CA_BUF_MAX,
 602                 .mode           = 0444,
 603                 .proc_handler   = proc_tcp_available_congestion_control,
 604         },
 605         {
 606                 .procname       = "tcp_allowed_congestion_control",
 607                 .maxlen         = TCP_CA_BUF_MAX,
 608                 .mode           = 0644,
 609                 .proc_handler   = proc_allowed_congestion_control,
 610         },
 611         {
 612                 .procname       = "tcp_thin_linear_timeouts",
 613                 .data           = &sysctl_tcp_thin_linear_timeouts,
 614                 .maxlen         = sizeof(int),
 615                 .mode           = 0644,
 616                 .proc_handler   = proc_dointvec
 617         },
 618         {
 619                 .procname       = "tcp_early_retrans",
 620                 .data           = &sysctl_tcp_early_retrans,
 621                 .maxlen         = sizeof(int),
 622                 .mode           = 0644,
 623                 .proc_handler   = proc_dointvec_minmax,
 624                 .extra1         = &zero,
 625                 .extra2         = &four,
 626         },
 627         {
 628                 .procname       = "tcp_min_tso_segs",
 629                 .data           = &sysctl_tcp_min_tso_segs,
 630                 .maxlen         = sizeof(int),
 631                 .mode           = 0644,
 632                 .proc_handler   = proc_dointvec_minmax,
 633                 .extra1         = &one,
 634                 .extra2         = &gso_max_segs,
 635         },
 636         {
 637                 .procname       = "tcp_pacing_ss_ratio",
 638                 .data           = &sysctl_tcp_pacing_ss_ratio,
 639                 .maxlen         = sizeof(int),
 640                 .mode           = 0644,
 641                 .proc_handler   = proc_dointvec_minmax,
 642                 .extra1         = &zero,
 643                 .extra2         = &thousand,
 644         },
 645         {
 646                 .procname       = "tcp_pacing_ca_ratio",
 647                 .data           = &sysctl_tcp_pacing_ca_ratio,
 648                 .maxlen         = sizeof(int),
 649                 .mode           = 0644,
 650                 .proc_handler   = proc_dointvec_minmax,
 651                 .extra1         = &zero,
 652                 .extra2         = &thousand,
 653         },
 654         {
 655                 .procname       = "tcp_autocorking",
 656                 .data           = &sysctl_tcp_autocorking,
 657                 .maxlen         = sizeof(int),
 658                 .mode           = 0644,
 659                 .proc_handler   = proc_dointvec_minmax,
 660                 .extra1         = &zero,
 661                 .extra2         = &one,
 662         },
 663         {
 664                 .procname       = "tcp_invalid_ratelimit",
 665                 .data           = &sysctl_tcp_invalid_ratelimit,
 666                 .maxlen         = sizeof(int),
 667                 .mode           = 0644,
 668                 .proc_handler   = proc_dointvec_ms_jiffies,
 669         },
 670         {
 671                 .procname       = "tcp_available_ulp",
 672                 .maxlen         = TCP_ULP_BUF_MAX,
 673                 .mode           = 0444,
 674                 .proc_handler   = proc_tcp_available_ulp,
 675         },
 676         {
 677                 .procname       = "icmp_msgs_per_sec",
 678                 .data           = &sysctl_icmp_msgs_per_sec,
 679                 .maxlen         = sizeof(int),
 680                 .mode           = 0644,
 681                 .proc_handler   = proc_dointvec_minmax,
 682                 .extra1         = &zero,
 683         },
 684         {
 685                 .procname       = "icmp_msgs_burst",
 686                 .data           = &sysctl_icmp_msgs_burst,
 687                 .maxlen         = sizeof(int),
 688                 .mode           = 0644,
 689                 .proc_handler   = proc_dointvec_minmax,
 690                 .extra1         = &zero,
 691         },
 692         {
 693                 .procname       = "udp_mem",
 694                 .data           = &sysctl_udp_mem,
 695                 .maxlen         = sizeof(sysctl_udp_mem),
 696                 .mode           = 0644,
 697                 .proc_handler   = proc_doulongvec_minmax,
 698         },
 699         {
 700                 .procname       = "udp_rmem_min",
 701                 .data           = &sysctl_udp_rmem_min,
 702                 .maxlen         = sizeof(sysctl_udp_rmem_min),
 703                 .mode           = 0644,
 704                 .proc_handler   = proc_dointvec_minmax,
 705                 .extra1         = &one
 706         },
 707         {
 708                 .procname       = "udp_wmem_min",
 709                 .data           = &sysctl_udp_wmem_min,
 710                 .maxlen         = sizeof(sysctl_udp_wmem_min),
 711                 .mode           = 0644,
 712                 .proc_handler   = proc_dointvec_minmax,
 713                 .extra1         = &one
 714         },
 715         { }
 716 };
 717
 718 static struct ctl_table ipv4_net_table[] = {
 719         {
 720                 .procname       = "icmp_echo_ignore_all",
 721                 .data           = &init_net.ipv4.sysctl_icmp_echo_ignore_all,
 722                 .maxlen         = sizeof(int),
 723                 .mode           = 0644,
 724                 .proc_handler   = proc_dointvec
 725         },
 726         {
 727                 .procname       = "icmp_echo_ignore_broadcasts",
 728                 .data           = &init_net.ipv4.sysctl_icmp_echo_ignore_broadcasts,
 729                 .maxlen         = sizeof(int),
 730                 .mode           = 0644,
 731                 .proc_handler   = proc_dointvec
 732         },
 733         {
 734                 .procname       = "icmp_ignore_bogus_error_responses",
 735                 .data           = &init_net.ipv4.sysctl_icmp_ignore_bogus_error_responses,
 736                 .maxlen         = sizeof(int),
 737                 .mode           = 0644,
 738                 .proc_handler   = proc_dointvec
 739         },
 740         {
 741                 .procname       = "icmp_errors_use_inbound_ifaddr",
 742                 .data           = &init_net.ipv4.sysctl_icmp_errors_use_inbound_ifaddr,
 743                 .maxlen         = sizeof(int),
 744                 .mode           = 0644,
 745                 .proc_handler   = proc_dointvec
 746         },
 747         {
 748                 .procname       = "icmp_ratelimit",
 749                 .data           = &init_net.ipv4.sysctl_icmp_ratelimit,
 750                 .maxlen         = sizeof(int),
 751                 .mode           = 0644,
 752                 .proc_handler   = proc_dointvec_ms_jiffies,
 753         },
 754         {
 755                 .procname       = "icmp_ratemask",
 756                 .data           = &init_net.ipv4.sysctl_icmp_ratemask,
 757                 .maxlen         = sizeof(int),
 758                 .mode           = 0644,
 759                 .proc_handler   = proc_dointvec
 760         },
 761         {
 762                 .procname       = "ping_group_range",
 763                 .data           = &init_net.ipv4.ping_group_range.range,
 764                 .maxlen         = sizeof(gid_t)*2,
 765                 .mode           = 0644,
 766                 .proc_handler   = ipv4_ping_group_range,
 767         },
 768         {
 769                 .procname       = "tcp_ecn",
 770                 .data           = &init_net.ipv4.sysctl_tcp_ecn,
 771                 .maxlen         = sizeof(int),
 772                 .mode           = 0644,
 773                 .proc_handler   = proc_dointvec
 774         },
 775         {
 776                 .procname       = "tcp_ecn_fallback",
 777                 .data           = &init_net.ipv4.sysctl_tcp_ecn_fallback,
 778                 .maxlen         = sizeof(int),
 779                 .mode           = 0644,
 780                 .proc_handler   = proc_dointvec
 781         },
 782         {
 783                 .procname       = "ip_dynaddr",
 784                 .data           = &init_net.ipv4.sysctl_ip_dynaddr,
 785                 .maxlen         = sizeof(int),
 786                 .mode           = 0644,
 787                 .proc_handler   = proc_dointvec
 788         },
 789         {
 790                 .procname       = "ip_early_demux",
 791                 .data           = &init_net.ipv4.sysctl_ip_early_demux,
 792                 .maxlen         = sizeof(int),
 793                 .mode           = 0644,
 794                 .proc_handler   = proc_dointvec
 795         },
 796         {
 797                 .procname       = "udp_early_demux",
 798                 .data           = &init_net.ipv4.sysctl_udp_early_demux,
 799                 .maxlen         = sizeof(int),
 800                 .mode           = 0644,
 801                 .proc_handler   = proc_douintvec,
 802         },
 803         {
 804                 .procname       = "tcp_early_demux",
 805                 .data           = &init_net.ipv4.sysctl_tcp_early_demux,
 806                 .maxlen         = sizeof(int),
 807                 .mode           = 0644,
 808                 .proc_handler   = proc_douintvec,
 809         },
 810         {
 811                 .procname       = "ip_default_ttl",
 812                 .data           = &init_net.ipv4.sysctl_ip_default_ttl,
 813                 .maxlen         = sizeof(int),
 814                 .mode           = 0644,
 815                 .proc_handler   = proc_dointvec_minmax,
 816                 .extra1         = &ip_ttl_min,
 817                 .extra2         = &ip_ttl_max,
 818         },
 819         {
 820                 .procname       = "ip_local_port_range",
 821                 .maxlen         = sizeof(init_net.ipv4.ip_local_ports.range),
 822                 .data           = &init_net.ipv4.ip_local_ports.range,
 823                 .mode           = 0644,
 824                 .proc_handler   = ipv4_local_port_range,
 825         },
 826         {
 827                 .procname       = "ip_local_reserved_ports",
 828                 .data           = &init_net.ipv4.sysctl_local_reserved_ports,
 829                 .maxlen         = 65536,
 830                 .mode           = 0644,
 831                 .proc_handler   = proc_do_large_bitmap,
 832         },
 833         {
 834                 .procname       = "ip_no_pmtu_disc",
 835                 .data           = &init_net.ipv4.sysctl_ip_no_pmtu_disc,
 836                 .maxlen         = sizeof(int),
 837                 .mode           = 0644,
 838                 .proc_handler   = proc_dointvec
 839         },
 840         {
 841                 .procname       = "ip_forward_use_pmtu",
 842                 .data           = &init_net.ipv4.sysctl_ip_fwd_use_pmtu,
 843                 .maxlen         = sizeof(int),
 844                 .mode           = 0644,
 845                 .proc_handler   = proc_dointvec,
 846         },
 847         {
 848                 .procname       = "ip_nonlocal_bind",
 849                 .data           = &init_net.ipv4.sysctl_ip_nonlocal_bind,
 850                 .maxlen         = sizeof(int),
 851                 .mode           = 0644,
 852                 .proc_handler   = proc_dointvec
 853         },
 854         {
 855                 .procname       = "fwmark_reflect",
 856                 .data           = &init_net.ipv4.sysctl_fwmark_reflect,
 857                 .maxlen         = sizeof(int),
 858                 .mode           = 0644,
 859                 .proc_handler   = proc_dointvec,
 860         },
 861         {
 862                 .procname       = "tcp_fwmark_accept",
 863                 .data           = &init_net.ipv4.sysctl_tcp_fwmark_accept,
 864                 .maxlen         = sizeof(int),
 865                 .mode           = 0644,
 866                 .proc_handler   = proc_dointvec,
 867         },
 868 #ifdef CONFIG_NET_L3_MASTER_DEV
 869         {
 870                 .procname       = "tcp_l3mdev_accept",
 871                 .data           = &init_net.ipv4.sysctl_tcp_l3mdev_accept,
 872                 .maxlen         = sizeof(int),
 873                 .mode           = 0644,
 874                 .proc_handler   = proc_dointvec_minmax,
 875                 .extra1         = &zero,
 876                 .extra2         = &one,
 877         },
 878 #endif
 879         {
 880                 .procname       = "tcp_mtu_probing",
 881                 .data           = &init_net.ipv4.sysctl_tcp_mtu_probing,
 882                 .maxlen         = sizeof(int),
 883                 .mode           = 0644,
 884                 .proc_handler   = proc_dointvec,
 885         },
 886         {
 887                 .procname       = "tcp_base_mss",
 888                 .data           = &init_net.ipv4.sysctl_tcp_base_mss,
 889                 .maxlen         = sizeof(int),
 890                 .mode           = 0644,
 891                 .proc_handler   = proc_dointvec,
 892         },
 893         {
 894                 .procname       = "tcp_min_snd_mss",
 895                 .data           = &init_net.ipv4.sysctl_tcp_min_snd_mss,
 896                 .maxlen         = sizeof(int),
 897                 .mode           = 0644,
 898                 .proc_handler   = proc_dointvec_minmax,
 899                 .extra1         = &tcp_min_snd_mss_min,
 900                 .extra2         = &tcp_min_snd_mss_max,
 901         },
 902         {
 903                 .procname       = "tcp_probe_threshold",
 904                 .data           = &init_net.ipv4.sysctl_tcp_probe_threshold,
 905                 .maxlen         = sizeof(int),
 906                 .mode           = 0644,
 907                 .proc_handler   = proc_dointvec,
 908         },
 909         {
 910                 .procname       = "tcp_probe_interval",
 911                 .data           = &init_net.ipv4.sysctl_tcp_probe_interval,
 912                 .maxlen         = sizeof(int),
 913                 .mode           = 0644,
 914                 .proc_handler   = proc_dointvec,
 915         },
 916         {
 917                 .procname       = "igmp_link_local_mcast_reports",
 918                 .data           = &init_net.ipv4.sysctl_igmp_llm_reports,
 919                 .maxlen         = sizeof(int),
 920                 .mode           = 0644,
 921                 .proc_handler   = proc_dointvec
 922         },
 923         {
 924                 .procname       = "igmp_max_memberships",
 925                 .data           = &init_net.ipv4.sysctl_igmp_max_memberships,
 926                 .maxlen         = sizeof(int),
 927                 .mode           = 0644,
 928                 .proc_handler   = proc_dointvec
 929         },
 930         {
 931                 .procname       = "igmp_max_msf",
 932                 .data           = &init_net.ipv4.sysctl_igmp_max_msf,
 933                 .maxlen         = sizeof(int),
 934                 .mode           = 0644,
 935                 .proc_handler   = proc_dointvec
 936         },
 937 #ifdef CONFIG_IP_MULTICAST
 938         {
 939                 .procname       = "igmp_qrv",
 940                 .data           = &init_net.ipv4.sysctl_igmp_qrv,
 941                 .maxlen         = sizeof(int),
 942                 .mode           = 0644,
 943                 .proc_handler   = proc_dointvec_minmax,
 944                 .extra1         = &one
 945         },
 946 #endif
 947         {
 948                 .procname       = "tcp_keepalive_time",
 949                 .data           = &init_net.ipv4.sysctl_tcp_keepalive_time,
 950                 .maxlen         = sizeof(int),
 951                 .mode           = 0644,
 952                 .proc_handler   = proc_dointvec_jiffies,
 953         },
 954         {
 955                 .procname       = "tcp_keepalive_probes",
 956                 .data           = &init_net.ipv4.sysctl_tcp_keepalive_probes,
 957                 .maxlen         = sizeof(int),
 958                 .mode           = 0644,
 959                 .proc_handler   = proc_dointvec
 960         },
 961         {
 962                 .procname       = "tcp_keepalive_intvl",
 963                 .data           = &init_net.ipv4.sysctl_tcp_keepalive_intvl,
 964                 .maxlen         = sizeof(int),
 965                 .mode           = 0644,
 966                 .proc_handler   = proc_dointvec_jiffies,
 967         },
 968         {
 969                 .procname       = "tcp_syn_retries",
 970                 .data           = &init_net.ipv4.sysctl_tcp_syn_retries,
 971                 .maxlen         = sizeof(int),
 972                 .mode           = 0644,
 973                 .proc_handler   = proc_dointvec_minmax,
 974                 .extra1         = &tcp_syn_retries_min,
 975                 .extra2         = &tcp_syn_retries_max
 976         },
 977         {
 978                 .procname       = "tcp_synack_retries",
 979                 .data           = &init_net.ipv4.sysctl_tcp_synack_retries,
 980                 .maxlen         = sizeof(int),
 981                 .mode           = 0644,
 982                 .proc_handler   = proc_dointvec
 983         },
 984 #ifdef CONFIG_SYN_COOKIES
 985         {
 986                 .procname       = "tcp_syncookies",
 987                 .data           = &init_net.ipv4.sysctl_tcp_syncookies,
 988                 .maxlen         = sizeof(int),
 989                 .mode           = 0644,
 990                 .proc_handler   = proc_dointvec
 991         },
 992 #endif
 993         {
 994                 .procname       = "tcp_reordering",
 995                 .data           = &init_net.ipv4.sysctl_tcp_reordering,
 996                 .maxlen         = sizeof(int),
 997                 .mode           = 0644,
 998                 .proc_handler   = proc_dointvec
 999         },
1000         {
1001                 .procname       = "tcp_retries1",
1002                 .data           = &init_net.ipv4.sysctl_tcp_retries1,
1003                 .maxlen         = sizeof(int),
1004                 .mode           = 0644,
1005                 .proc_handler   = proc_dointvec_minmax,
1006                 .extra2         = &tcp_retr1_max
1007         },
1008         {
1009                 .procname       = "tcp_retries2",
1010                 .data           = &init_net.ipv4.sysctl_tcp_retries2,
1011                 .maxlen         = sizeof(int),
1012                 .mode           = 0644,
1013                 .proc_handler   = proc_dointvec
1014         },
1015         {
1016                 .procname       = "tcp_orphan_retries",
1017                 .data           = &init_net.ipv4.sysctl_tcp_orphan_retries,
1018                 .maxlen         = sizeof(int),
1019                 .mode           = 0644,
1020                 .proc_handler   = proc_dointvec
1021         },
1022         {
1023                 .procname       = "tcp_fin_timeout",
1024                 .data           = &init_net.ipv4.sysctl_tcp_fin_timeout,
1025                 .maxlen         = sizeof(int),
1026                 .mode           = 0644,
1027                 .proc_handler   = proc_dointvec_jiffies,
1028         },
1029         {
1030                 .procname       = "tcp_notsent_lowat",
1031                 .data           = &init_net.ipv4.sysctl_tcp_notsent_lowat,
1032                 .maxlen         = sizeof(unsigned int),
1033                 .mode           = 0644,
1034                 .proc_handler   = proc_douintvec,
1035         },
1036         {
1037                 .procname       = "tcp_tw_reuse",
1038                 .data           = &init_net.ipv4.sysctl_tcp_tw_reuse,
1039                 .maxlen         = sizeof(int),
1040                 .mode           = 0644,
1041                 .proc_handler   = proc_dointvec
1042         },
1043         {
1044                 .procname       = "tcp_max_tw_buckets",
1045                 .data           = &init_net.ipv4.tcp_death_row.sysctl_max_tw_buckets,
1046                 .maxlen         = sizeof(int),
1047                 .mode           = 0644,
1048                 .proc_handler   = proc_dointvec
1049         },
1050         {
1051                 .procname       = "tcp_max_syn_backlog",
1052                 .data           = &init_net.ipv4.sysctl_max_syn_backlog,
1053                 .maxlen         = sizeof(int),
1054                 .mode           = 0644,
1055                 .proc_handler   = proc_dointvec
1056         },
1057 #ifdef CONFIG_IP_ROUTE_MULTIPATH
1058         {
1059                 .procname       = "fib_multipath_use_neigh",
1060                 .data           = &init_net.ipv4.sysctl_fib_multipath_use_neigh,
1061                 .maxlen         = sizeof(int),
1062                 .mode           = 0644,
1063                 .proc_handler   = proc_dointvec_minmax,
1064                 .extra1         = &zero,
1065                 .extra2         = &one,
1066         },
1067         {
1068                 .procname       = "fib_multipath_hash_policy",
1069                 .data           = &init_net.ipv4.sysctl_fib_multipath_hash_policy,
1070                 .maxlen         = sizeof(int),
1071                 .mode           = 0644,
1072                 .proc_handler   = proc_dointvec_minmax,
1073                 .extra1         = &zero,
1074                 .extra2         = &one,
1075         },
1076 #endif
1077         {
1078                 .procname       = "ip_unprivileged_port_start",
1079                 .maxlen         = sizeof(int),
1080                 .data           = &init_net.ipv4.sysctl_ip_prot_sock,
1081                 .mode           = 0644,
1082                 .proc_handler   = ipv4_privileged_ports,
1083         },
1084 #ifdef CONFIG_NET_L3_MASTER_DEV
1085         {
1086                 .procname       = "udp_l3mdev_accept",
1087                 .data           = &init_net.ipv4.sysctl_udp_l3mdev_accept,
1088                 .maxlen         = sizeof(int),
1089                 .mode           = 0644,
1090                 .proc_handler   = proc_dointvec_minmax,
1091                 .extra1         = &zero,
1092                 .extra2         = &one,
1093         },
1094 #endif
1095         {
1096                 .procname       = "tcp_sack",
1097                 .data           = &init_net.ipv4.sysctl_tcp_sack,
1098                 .maxlen         = sizeof(int),
1099                 .mode           = 0644,
1100                 .proc_handler   = proc_dointvec
1101         },
1102         {
1103                 .procname       = "tcp_window_scaling",
1104                 .data           = &init_net.ipv4.sysctl_tcp_window_scaling,
1105                 .maxlen         = sizeof(int),
1106                 .mode           = 0644,
1107                 .proc_handler   = proc_dointvec
1108         },
1109         {
1110                 .procname       = "tcp_timestamps",
1111                 .data           = &init_net.ipv4.sysctl_tcp_timestamps,
1112                 .maxlen         = sizeof(int),
1113                 .mode           = 0644,
1114                 .proc_handler   = proc_dointvec
1115         },
1116         { }
1117 };
1118
1119 static __net_init int ipv4_sysctl_init_net(struct net *net)
1120 {
1121         struct ctl_table *table;
1122
1123         table = ipv4_net_table;
1124         if (!net_eq(net, &init_net)) {
1125                 int i;
1126
1127                 table = kmemdup(table, sizeof(ipv4_net_table), GFP_KERNEL);
1128                 if (!table)
1129                         goto err_alloc;
1130
1131                 /* Update the variables to point into the current struct net */
1132                 for (i = 0; i < ARRAY_SIZE(ipv4_net_table) - 1; i++)
1133                         table[i].data += (void *)net - (void *)&init_net;
1134         }
1135
1136         net->ipv4.ipv4_hdr = register_net_sysctl(net, "net/ipv4", table);
1137         if (!net->ipv4.ipv4_hdr)
1138                 goto err_reg;
1139
1140         net->ipv4.sysctl_local_reserved_ports = kzalloc(65536 / 8, GFP_KERNEL);
1141         if (!net->ipv4.sysctl_local_reserved_ports)
1142                 goto err_ports;
1143
1144         return 0;
1145
1146 err_ports:
1147         unregister_net_sysctl_table(net->ipv4.ipv4_hdr);
1148 err_reg:
1149         if (!net_eq(net, &init_net))
1150                 kfree(table);
1151 err_alloc:
1152         return -ENOMEM;
1153 }
1154
1155 static __net_exit void ipv4_sysctl_exit_net(struct net *net)
1156 {
1157         struct ctl_table *table;
1158
1159         kfree(net->ipv4.sysctl_local_reserved_ports);
1160         table = net->ipv4.ipv4_hdr->ctl_table_arg;
1161         unregister_net_sysctl_table(net->ipv4.ipv4_hdr);
1162         kfree(table);
1163 }
1164
1165 static __net_initdata struct pernet_operations ipv4_sysctl_ops = {
1166         .init = ipv4_sysctl_init_net,
1167         .exit = ipv4_sysctl_exit_net,
1168 };
1169
1170 static __init int sysctl_ipv4_init(void)
1171 {
1172         struct ctl_table_header *hdr;
1173
1174         hdr = register_net_sysctl(&init_net, "net/ipv4", ipv4_table);
1175         if (!hdr)
1176                 return -ENOMEM;
1177
1178         if (register_pernet_subsys(&ipv4_sysctl_ops)) {
1179                 unregister_net_sysctl_table(hdr);
1180                 return -ENOMEM;
1181         }
1182
1183         return 0;
1184 }
1185
1186 __initcall(sysctl_ipv4_init);