net/ipv4/sysctl_net_ipv4.c

   1 /*
   2  * sysctl_net_ipv4.c: sysctl interface to net IPV4 subsystem.
   3  *
   4  * Begun April 1, 1996, Mike Shaver.
   5  * Added /proc/sys/net/ipv4 directory entry (empty =) ). [MS]
   6  */
   7
   8 #include <linux/mm.h>
   9 #include <linux/module.h>
  10 #include <linux/sysctl.h>
  11 #include <linux/igmp.h>
  12 #include <linux/inetdevice.h>
  13 #include <linux/seqlock.h>
  14 #include <linux/init.h>
  15 #include <linux/slab.h>
  16 #include <linux/nsproxy.h>
  17 #include <linux/swap.h>
  18 #include <net/snmp.h>
  19 #include <net/icmp.h>
  20 #include <net/ip.h>
  21 #include <net/route.h>
  22 #include <net/tcp.h>
  23 #include <net/udp.h>
  24 #include <net/cipso_ipv4.h>
  25 #include <net/inet_frag.h>
  26 #include <net/ping.h>
  27 #include <net/tcp_memcontrol.h>
  28
  29 static int zero;
  30 static int one = 1;
  31 static int four = 4;
  32 static int gso_max_segs = GSO_MAX_SEGS;
  33 static int tcp_retr1_max = 255;
  34 static int ip_local_port_range_min[] = { 1, 1 };
  35 static int ip_local_port_range_max[] = { 65535, 65535 };
  36 static int tcp_adv_win_scale_min = -31;
  37 static int tcp_adv_win_scale_max = 31;
  38 static int ip_ttl_min = 1;
  39 static int ip_ttl_max = 255;
  40 static int tcp_syn_retries_min = 1;
  41 static int tcp_syn_retries_max = MAX_TCP_SYNCNT;
  42 static int ip_ping_group_range_min[] = { 0, 0 };
  43 static int ip_ping_group_range_max[] = { GID_T_MAX, GID_T_MAX };
  44
  45 /* Update system visible IP port range */
  46 static void set_local_port_range(struct net *net, int range[2])
  47 {
  48         write_seqlock(&net->ipv4.sysctl_local_ports.lock);
  49         net->ipv4.sysctl_local_ports.range[0] = range[0];
  50         net->ipv4.sysctl_local_ports.range[1] = range[1];
  51         write_sequnlock(&net->ipv4.sysctl_local_ports.lock);
  52 }
  53
  54 /* Validate changes from /proc interface. */
  55 static int ipv4_local_port_range(struct ctl_table *table, int write,
  56                                  void __user *buffer,
  57                                  size_t *lenp, loff_t *ppos)
  58 {
  59         struct net *net =
  60                 container_of(table->data, struct net, ipv4.sysctl_local_ports.range);
  61         int ret;
  62         int range[2];
  63         struct ctl_table tmp = {
  64                 .data = &range,
  65                 .maxlen = sizeof(range),
  66                 .mode = table->mode,
  67                 .extra1 = &ip_local_port_range_min,
  68                 .extra2 = &ip_local_port_range_max,
  69         };
  70
  71         inet_get_local_port_range(net, &range[0], &range[1]);
  72
  73         ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
  74
  75         if (write && ret == 0) {
  76                 if (range[1] < range[0])
  77                         ret = -EINVAL;
  78                 else
  79                         set_local_port_range(net, range);
  80         }
  81
  82         return ret;
  83 }
  84
  85
  86 static void inet_get_ping_group_range_table(struct ctl_table *table, kgid_t *low, kgid_t *high)
  87 {
  88         kgid_t *data = table->data;
  89         struct net *net =
  90                 container_of(table->data, struct net, ipv4.sysctl_ping_group_range);
  91         unsigned int seq;
  92         do {
  93                 seq = read_seqbegin(&net->ipv4.sysctl_local_ports.lock);
  94
  95                 *low = data[0];
  96                 *high = data[1];
  97         } while (read_seqretry(&net->ipv4.sysctl_local_ports.lock, seq));
  98 }
  99
 100 /* Update system visible IP port range */
 101 static void set_ping_group_range(struct ctl_table *table, kgid_t low, kgid_t high)
 102 {
 103         kgid_t *data = table->data;
 104         struct net *net =
 105                 container_of(table->data, struct net, ipv4.sysctl_ping_group_range);
 106         write_seqlock(&net->ipv4.sysctl_local_ports.lock);
 107         data[0] = low;
 108         data[1] = high;
 109         write_sequnlock(&net->ipv4.sysctl_local_ports.lock);
 110 }
 111
 112 /* Validate changes from /proc interface. */
 113 static int ipv4_ping_group_range(struct ctl_table *table, int write,
 114                                  void __user *buffer,
 115                                  size_t *lenp, loff_t *ppos)
 116 {
 117         struct user_namespace *user_ns = current_user_ns();
 118         int ret;
 119         gid_t urange[2];
 120         kgid_t low, high;
 121         struct ctl_table tmp = {
 122                 .data = &urange,
 123                 .maxlen = sizeof(urange),
 124                 .mode = table->mode,
 125                 .extra1 = &ip_ping_group_range_min,
 126                 .extra2 = &ip_ping_group_range_max,
 127         };
 128
 129         inet_get_ping_group_range_table(table, &low, &high);
 130         urange[0] = from_kgid_munged(user_ns, low);
 131         urange[1] = from_kgid_munged(user_ns, high);
 132         ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
 133
 134         if (write && ret == 0) {
 135                 low = make_kgid(user_ns, urange[0]);
 136                 high = make_kgid(user_ns, urange[1]);
 137                 if (!gid_valid(low) || !gid_valid(high) ||
 138                     (urange[1] < urange[0]) || gid_lt(high, low)) {
 139                         low = make_kgid(&init_user_ns, 1);
 140                         high = make_kgid(&init_user_ns, 0);
 141                 }
 142                 set_ping_group_range(table, low, high);
 143         }
 144
 145         return ret;
 146 }
 147
 148 static int proc_tcp_congestion_control(struct ctl_table *ctl, int write,
 149                                        void __user *buffer, size_t *lenp, loff_t *ppos)
 150 {
 151         char val[TCP_CA_NAME_MAX];
 152         struct ctl_table tbl = {
 153                 .data = val,
 154                 .maxlen = TCP_CA_NAME_MAX,
 155         };
 156         int ret;
 157
 158         tcp_get_default_congestion_control(val);
 159
 160         ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
 161         if (write && ret == 0)
 162                 ret = tcp_set_default_congestion_control(val);
 163         return ret;
 164 }
 165
 166 static int proc_tcp_available_congestion_control(struct ctl_table *ctl,
 167                                                  int write,
 168                                                  void __user *buffer, size_t *lenp,
 169                                                  loff_t *ppos)
 170 {
 171         struct ctl_table tbl = { .maxlen = TCP_CA_BUF_MAX, };
 172         int ret;
 173
 174         tbl.data = kmalloc(tbl.maxlen, GFP_USER);
 175         if (!tbl.data)
 176                 return -ENOMEM;
 177         tcp_get_available_congestion_control(tbl.data, TCP_CA_BUF_MAX);
 178         ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
 179         kfree(tbl.data);
 180         return ret;
 181 }
 182
 183 static int proc_allowed_congestion_control(struct ctl_table *ctl,
 184                                            int write,
 185                                            void __user *buffer, size_t *lenp,
 186                                            loff_t *ppos)
 187 {
 188         struct ctl_table tbl = { .maxlen = TCP_CA_BUF_MAX };
 189         int ret;
 190
 191         tbl.data = kmalloc(tbl.maxlen, GFP_USER);
 192         if (!tbl.data)
 193                 return -ENOMEM;
 194
 195         tcp_get_allowed_congestion_control(tbl.data, tbl.maxlen);
 196         ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
 197         if (write && ret == 0)
 198                 ret = tcp_set_allowed_congestion_control(tbl.data);
 199         kfree(tbl.data);
 200         return ret;
 201 }
 202
 203 static int ipv4_tcp_mem(struct ctl_table *ctl, int write,
 204                            void __user *buffer, size_t *lenp,
 205                            loff_t *ppos)
 206 {
 207         ctl->data = &current->nsproxy->net_ns->ipv4.sysctl_tcp_mem;
 208         return proc_doulongvec_minmax(ctl, write, buffer, lenp, ppos);
 209 }
 210
 211 static int proc_tcp_fastopen_key(struct ctl_table *ctl, int write,
 212                                  void __user *buffer, size_t *lenp,
 213                                  loff_t *ppos)
 214 {
 215         struct ctl_table tbl = { .maxlen = (TCP_FASTOPEN_KEY_LENGTH * 2 + 10) };
 216         struct tcp_fastopen_context *ctxt;
 217         int ret;
 218         u32  user_key[4]; /* 16 bytes, matching TCP_FASTOPEN_KEY_LENGTH */
 219
 220         tbl.data = kmalloc(tbl.maxlen, GFP_KERNEL);
 221         if (!tbl.data)
 222                 return -ENOMEM;
 223
 224         rcu_read_lock();
 225         ctxt = rcu_dereference(tcp_fastopen_ctx);
 226         if (ctxt)
 227                 memcpy(user_key, ctxt->key, TCP_FASTOPEN_KEY_LENGTH);
 228         else
 229                 memset(user_key, 0, sizeof(user_key));
 230         rcu_read_unlock();
 231
 232         snprintf(tbl.data, tbl.maxlen, "%08x-%08x-%08x-%08x",
 233                 user_key[0], user_key[1], user_key[2], user_key[3]);
 234         ret = proc_dostring(&tbl, write, buffer, lenp, ppos);
 235
 236         if (write && ret == 0) {
 237                 if (sscanf(tbl.data, "%x-%x-%x-%x", user_key, user_key + 1,
 238                            user_key + 2, user_key + 3) != 4) {
 239                         ret = -EINVAL;
 240                         goto bad_key;
 241                 }
 242                 /* Generate a dummy secret but don't publish it. This
 243                  * is needed so we don't regenerate a new key on the
 244                  * first invocation of tcp_fastopen_cookie_gen
 245                  */
 246                 tcp_fastopen_init_key_once(false);
 247                 tcp_fastopen_reset_cipher(user_key, TCP_FASTOPEN_KEY_LENGTH);
 248         }
 249
 250 bad_key:
 251         pr_debug("proc FO key set 0x%x-%x-%x-%x <- 0x%s: %u\n",
 252                user_key[0], user_key[1], user_key[2], user_key[3],
 253                (char *)tbl.data, ret);
 254         kfree(tbl.data);
 255         return ret;
 256 }
 257
 258 static struct ctl_table ipv4_table[] = {
 259         {
 260                 .procname       = "tcp_timestamps",
 261                 .data           = &sysctl_tcp_timestamps,
 262                 .maxlen         = sizeof(int),
 263                 .mode           = 0644,
 264                 .proc_handler   = proc_dointvec
 265         },
 266         {
 267                 .procname       = "tcp_window_scaling",
 268                 .data           = &sysctl_tcp_window_scaling,
 269                 .maxlen         = sizeof(int),
 270                 .mode           = 0644,
 271                 .proc_handler   = proc_dointvec
 272         },
 273         {
 274                 .procname       = "tcp_sack",
 275                 .data           = &sysctl_tcp_sack,
 276                 .maxlen         = sizeof(int),
 277                 .mode           = 0644,
 278                 .proc_handler   = proc_dointvec
 279         },
 280         {
 281                 .procname       = "tcp_retrans_collapse",
 282                 .data           = &sysctl_tcp_retrans_collapse,
 283                 .maxlen         = sizeof(int),
 284                 .mode           = 0644,
 285                 .proc_handler   = proc_dointvec
 286         },
 287         {
 288                 .procname       = "ip_default_ttl",
 289                 .data           = &sysctl_ip_default_ttl,
 290                 .maxlen         = sizeof(int),
 291                 .mode           = 0644,
 292                 .proc_handler   = proc_dointvec_minmax,
 293                 .extra1         = &ip_ttl_min,
 294                 .extra2         = &ip_ttl_max,
 295         },
 296         {
 297                 .procname       = "ip_no_pmtu_disc",
 298                 .data           = &ipv4_config.no_pmtu_disc,
 299                 .maxlen         = sizeof(int),
 300                 .mode           = 0644,
 301                 .proc_handler   = proc_dointvec
 302         },
 303         {
 304                 .procname       = "ip_nonlocal_bind",
 305                 .data           = &sysctl_ip_nonlocal_bind,
 306                 .maxlen         = sizeof(int),
 307                 .mode           = 0644,
 308                 .proc_handler   = proc_dointvec
 309         },
 310         {
 311                 .procname       = "tcp_syn_retries",
 312                 .data           = &sysctl_tcp_syn_retries,
 313                 .maxlen         = sizeof(int),
 314                 .mode           = 0644,
 315                 .proc_handler   = proc_dointvec_minmax,
 316                 .extra1         = &tcp_syn_retries_min,
 317                 .extra2         = &tcp_syn_retries_max
 318         },
 319         {
 320                 .procname       = "tcp_synack_retries",
 321                 .data           = &sysctl_tcp_synack_retries,
 322                 .maxlen         = sizeof(int),
 323                 .mode           = 0644,
 324                 .proc_handler   = proc_dointvec
 325         },
 326         {
 327                 .procname       = "tcp_max_orphans",
 328                 .data           = &sysctl_tcp_max_orphans,
 329                 .maxlen         = sizeof(int),
 330                 .mode           = 0644,
 331                 .proc_handler   = proc_dointvec
 332         },
 333         {
 334                 .procname       = "tcp_max_tw_buckets",
 335                 .data           = &tcp_death_row.sysctl_max_tw_buckets,
 336                 .maxlen         = sizeof(int),
 337                 .mode           = 0644,
 338                 .proc_handler   = proc_dointvec
 339         },
 340         {
 341                 .procname       = "ip_early_demux",
 342                 .data           = &sysctl_ip_early_demux,
 343                 .maxlen         = sizeof(int),
 344                 .mode           = 0644,
 345                 .proc_handler   = proc_dointvec
 346         },
 347         {
 348                 .procname       = "ip_dynaddr",
 349                 .data           = &sysctl_ip_dynaddr,
 350                 .maxlen         = sizeof(int),
 351                 .mode           = 0644,
 352                 .proc_handler   = proc_dointvec
 353         },
 354         {
 355                 .procname       = "tcp_keepalive_time",
 356                 .data           = &sysctl_tcp_keepalive_time,
 357                 .maxlen         = sizeof(int),
 358                 .mode           = 0644,
 359                 .proc_handler   = proc_dointvec_jiffies,
 360         },
 361         {
 362                 .procname       = "tcp_keepalive_probes",
 363                 .data           = &sysctl_tcp_keepalive_probes,
 364                 .maxlen         = sizeof(int),
 365                 .mode           = 0644,
 366                 .proc_handler   = proc_dointvec
 367         },
 368         {
 369                 .procname       = "tcp_keepalive_intvl",
 370                 .data           = &sysctl_tcp_keepalive_intvl,
 371                 .maxlen         = sizeof(int),
 372                 .mode           = 0644,
 373                 .proc_handler   = proc_dointvec_jiffies,
 374         },
 375         {
 376                 .procname       = "tcp_retries1",
 377                 .data           = &sysctl_tcp_retries1,
 378                 .maxlen         = sizeof(int),
 379                 .mode           = 0644,
 380                 .proc_handler   = proc_dointvec_minmax,
 381                 .extra2         = &tcp_retr1_max
 382         },
 383         {
 384                 .procname       = "tcp_retries2",
 385                 .data           = &sysctl_tcp_retries2,
 386                 .maxlen         = sizeof(int),
 387                 .mode           = 0644,
 388                 .proc_handler   = proc_dointvec
 389         },
 390         {
 391                 .procname       = "tcp_fin_timeout",
 392                 .data           = &sysctl_tcp_fin_timeout,
 393                 .maxlen         = sizeof(int),
 394                 .mode           = 0644,
 395                 .proc_handler   = proc_dointvec_jiffies,
 396         },
 397 #ifdef CONFIG_SYN_COOKIES
 398         {
 399                 .procname       = "tcp_syncookies",
 400                 .data           = &sysctl_tcp_syncookies,
 401                 .maxlen         = sizeof(int),
 402                 .mode           = 0644,
 403                 .proc_handler   = proc_dointvec
 404         },
 405 #endif
 406         {
 407                 .procname       = "tcp_fastopen",
 408                 .data           = &sysctl_tcp_fastopen,
 409                 .maxlen         = sizeof(int),
 410                 .mode           = 0644,
 411                 .proc_handler   = proc_dointvec,
 412         },
 413         {
 414                 .procname       = "tcp_fastopen_key",
 415                 .mode           = 0600,
 416                 .maxlen         = ((TCP_FASTOPEN_KEY_LENGTH * 2) + 10),
 417                 .proc_handler   = proc_tcp_fastopen_key,
 418         },
 419         {
 420                 .procname       = "tcp_tw_recycle",
 421                 .data           = &tcp_death_row.sysctl_tw_recycle,
 422                 .maxlen         = sizeof(int),
 423                 .mode           = 0644,
 424                 .proc_handler   = proc_dointvec
 425         },
 426         {
 427                 .procname       = "tcp_abort_on_overflow",
 428                 .data           = &sysctl_tcp_abort_on_overflow,
 429                 .maxlen         = sizeof(int),
 430                 .mode           = 0644,
 431                 .proc_handler   = proc_dointvec
 432         },
 433         {
 434                 .procname       = "tcp_stdurg",
 435                 .data           = &sysctl_tcp_stdurg,
 436                 .maxlen         = sizeof(int),
 437                 .mode           = 0644,
 438                 .proc_handler   = proc_dointvec
 439         },
 440         {
 441                 .procname       = "tcp_rfc1337",
 442                 .data           = &sysctl_tcp_rfc1337,
 443                 .maxlen         = sizeof(int),
 444                 .mode           = 0644,
 445                 .proc_handler   = proc_dointvec
 446         },
 447         {
 448                 .procname       = "tcp_max_syn_backlog",
 449                 .data           = &sysctl_max_syn_backlog,
 450                 .maxlen         = sizeof(int),
 451                 .mode           = 0644,
 452                 .proc_handler   = proc_dointvec
 453         },
 454         {
 455                 .procname       = "ip_local_reserved_ports",
 456                 .data           = NULL, /* initialized in sysctl_ipv4_init */
 457                 .maxlen         = 65536,
 458                 .mode           = 0644,
 459                 .proc_handler   = proc_do_large_bitmap,
 460         },
 461         {
 462                 .procname       = "igmp_max_memberships",
 463                 .data           = &sysctl_igmp_max_memberships,
 464                 .maxlen         = sizeof(int),
 465                 .mode           = 0644,
 466                 .proc_handler   = proc_dointvec
 467         },
 468         {
 469                 .procname       = "igmp_max_msf",
 470                 .data           = &sysctl_igmp_max_msf,
 471                 .maxlen         = sizeof(int),
 472                 .mode           = 0644,
 473                 .proc_handler   = proc_dointvec
 474         },
 475         {
 476                 .procname       = "inet_peer_threshold",
 477                 .data           = &inet_peer_threshold,
 478                 .maxlen         = sizeof(int),
 479                 .mode           = 0644,
 480                 .proc_handler   = proc_dointvec
 481         },
 482         {
 483                 .procname       = "inet_peer_minttl",
 484                 .data           = &inet_peer_minttl,
 485                 .maxlen         = sizeof(int),
 486                 .mode           = 0644,
 487                 .proc_handler   = proc_dointvec_jiffies,
 488         },
 489         {
 490                 .procname       = "inet_peer_maxttl",
 491                 .data           = &inet_peer_maxttl,
 492                 .maxlen         = sizeof(int),
 493                 .mode           = 0644,
 494                 .proc_handler   = proc_dointvec_jiffies,
 495         },
 496         {
 497                 .procname       = "tcp_orphan_retries",
 498                 .data           = &sysctl_tcp_orphan_retries,
 499                 .maxlen         = sizeof(int),
 500                 .mode           = 0644,
 501                 .proc_handler   = proc_dointvec
 502         },
 503         {
 504                 .procname       = "tcp_fack",
 505                 .data           = &sysctl_tcp_fack,
 506                 .maxlen         = sizeof(int),
 507                 .mode           = 0644,
 508                 .proc_handler   = proc_dointvec
 509         },
 510         {
 511                 .procname       = "tcp_reordering",
 512                 .data           = &sysctl_tcp_reordering,
 513                 .maxlen         = sizeof(int),
 514                 .mode           = 0644,
 515                 .proc_handler   = proc_dointvec
 516         },
 517         {
 518                 .procname       = "tcp_dsack",
 519                 .data           = &sysctl_tcp_dsack,
 520                 .maxlen         = sizeof(int),
 521                 .mode           = 0644,
 522                 .proc_handler   = proc_dointvec
 523         },
 524         {
 525                 .procname       = "tcp_wmem",
 526                 .data           = &sysctl_tcp_wmem,
 527                 .maxlen         = sizeof(sysctl_tcp_wmem),
 528                 .mode           = 0644,
 529                 .proc_handler   = proc_dointvec_minmax,
 530                 .extra1         = &one,
 531         },
 532         {
 533                 .procname       = "tcp_notsent_lowat",
 534                 .data           = &sysctl_tcp_notsent_lowat,
 535                 .maxlen         = sizeof(sysctl_tcp_notsent_lowat),
 536                 .mode           = 0644,
 537                 .proc_handler   = proc_dointvec,
 538         },
 539         {
 540                 .procname       = "tcp_rmem",
 541                 .data           = &sysctl_tcp_rmem,
 542                 .maxlen         = sizeof(sysctl_tcp_rmem),
 543                 .mode           = 0644,
 544                 .proc_handler   = proc_dointvec_minmax,
 545                 .extra1         = &one,
 546         },
 547         {
 548                 .procname       = "tcp_app_win",
 549                 .data           = &sysctl_tcp_app_win,
 550                 .maxlen         = sizeof(int),
 551                 .mode           = 0644,
 552                 .proc_handler   = proc_dointvec
 553         },
 554         {
 555                 .procname       = "tcp_adv_win_scale",
 556                 .data           = &sysctl_tcp_adv_win_scale,
 557                 .maxlen         = sizeof(int),
 558                 .mode           = 0644,
 559                 .proc_handler   = proc_dointvec_minmax,
 560                 .extra1         = &tcp_adv_win_scale_min,
 561                 .extra2         = &tcp_adv_win_scale_max,
 562         },
 563         {
 564                 .procname       = "tcp_tw_reuse",
 565                 .data           = &sysctl_tcp_tw_reuse,
 566                 .maxlen         = sizeof(int),
 567                 .mode           = 0644,
 568                 .proc_handler   = proc_dointvec
 569         },
 570         {
 571                 .procname       = "tcp_frto",
 572                 .data           = &sysctl_tcp_frto,
 573                 .maxlen         = sizeof(int),
 574                 .mode           = 0644,
 575                 .proc_handler   = proc_dointvec
 576         },
 577         {
 578                 .procname       = "tcp_low_latency",
 579                 .data           = &sysctl_tcp_low_latency,
 580                 .maxlen         = sizeof(int),
 581                 .mode           = 0644,
 582                 .proc_handler   = proc_dointvec
 583         },
 584         {
 585                 .procname       = "tcp_no_metrics_save",
 586                 .data           = &sysctl_tcp_nometrics_save,
 587                 .maxlen         = sizeof(int),
 588                 .mode           = 0644,
 589                 .proc_handler   = proc_dointvec,
 590         },
 591         {
 592                 .procname       = "tcp_moderate_rcvbuf",
 593                 .data           = &sysctl_tcp_moderate_rcvbuf,
 594                 .maxlen         = sizeof(int),
 595                 .mode           = 0644,
 596                 .proc_handler   = proc_dointvec,
 597         },
 598         {
 599                 .procname       = "tcp_tso_win_divisor",
 600                 .data           = &sysctl_tcp_tso_win_divisor,
 601                 .maxlen         = sizeof(int),
 602                 .mode           = 0644,
 603                 .proc_handler   = proc_dointvec,
 604         },
 605         {
 606                 .procname       = "tcp_congestion_control",
 607                 .mode           = 0644,
 608                 .maxlen         = TCP_CA_NAME_MAX,
 609                 .proc_handler   = proc_tcp_congestion_control,
 610         },
 611         {
 612                 .procname       = "tcp_mtu_probing",
 613                 .data           = &sysctl_tcp_mtu_probing,
 614                 .maxlen         = sizeof(int),
 615                 .mode           = 0644,
 616                 .proc_handler   = proc_dointvec,
 617         },
 618         {
 619                 .procname       = "tcp_base_mss",
 620                 .data           = &sysctl_tcp_base_mss,
 621                 .maxlen         = sizeof(int),
 622                 .mode           = 0644,
 623                 .proc_handler   = proc_dointvec,
 624         },
 625         {
 626                 .procname       = "tcp_workaround_signed_windows",
 627                 .data           = &sysctl_tcp_workaround_signed_windows,
 628                 .maxlen         = sizeof(int),
 629                 .mode           = 0644,
 630                 .proc_handler   = proc_dointvec
 631         },
 632         {
 633                 .procname       = "tcp_limit_output_bytes",
 634                 .data           = &sysctl_tcp_limit_output_bytes,
 635                 .maxlen         = sizeof(int),
 636                 .mode           = 0644,
 637                 .proc_handler   = proc_dointvec
 638         },
 639         {
 640                 .procname       = "tcp_challenge_ack_limit",
 641                 .data           = &sysctl_tcp_challenge_ack_limit,
 642                 .maxlen         = sizeof(int),
 643                 .mode           = 0644,
 644                 .proc_handler   = proc_dointvec
 645         },
 646 #ifdef CONFIG_NET_DMA
 647         {
 648                 .procname       = "tcp_dma_copybreak",
 649                 .data           = &sysctl_tcp_dma_copybreak,
 650                 .maxlen         = sizeof(int),
 651                 .mode           = 0644,
 652                 .proc_handler   = proc_dointvec
 653         },
 654 #endif
 655         {
 656                 .procname       = "tcp_slow_start_after_idle",
 657                 .data           = &sysctl_tcp_slow_start_after_idle,
 658                 .maxlen         = sizeof(int),
 659                 .mode           = 0644,
 660                 .proc_handler   = proc_dointvec
 661         },
 662 #ifdef CONFIG_NETLABEL
 663         {
 664                 .procname       = "cipso_cache_enable",
 665                 .data           = &cipso_v4_cache_enabled,
 666                 .maxlen         = sizeof(int),
 667                 .mode           = 0644,
 668                 .proc_handler   = proc_dointvec,
 669         },
 670         {
 671                 .procname       = "cipso_cache_bucket_size",
 672                 .data           = &cipso_v4_cache_bucketsize,
 673                 .maxlen         = sizeof(int),
 674                 .mode           = 0644,
 675                 .proc_handler   = proc_dointvec,
 676         },
 677         {
 678                 .procname       = "cipso_rbm_optfmt",
 679                 .data           = &cipso_v4_rbm_optfmt,
 680                 .maxlen         = sizeof(int),
 681                 .mode           = 0644,
 682                 .proc_handler   = proc_dointvec,
 683         },
 684         {
 685                 .procname       = "cipso_rbm_strictvalid",
 686                 .data           = &cipso_v4_rbm_strictvalid,
 687                 .maxlen         = sizeof(int),
 688                 .mode           = 0644,
 689                 .proc_handler   = proc_dointvec,
 690         },
 691 #endif /* CONFIG_NETLABEL */
 692         {
 693                 .procname       = "tcp_available_congestion_control",
 694                 .maxlen         = TCP_CA_BUF_MAX,
 695                 .mode           = 0444,
 696                 .proc_handler   = proc_tcp_available_congestion_control,
 697         },
 698         {
 699                 .procname       = "tcp_allowed_congestion_control",
 700                 .maxlen         = TCP_CA_BUF_MAX,
 701                 .mode           = 0644,
 702                 .proc_handler   = proc_allowed_congestion_control,
 703         },
 704         {
 705                 .procname       = "tcp_max_ssthresh",
 706                 .data           = &sysctl_tcp_max_ssthresh,
 707                 .maxlen         = sizeof(int),
 708                 .mode           = 0644,
 709                 .proc_handler   = proc_dointvec,
 710         },
 711         {
 712                 .procname       = "tcp_thin_linear_timeouts",
 713                 .data           = &sysctl_tcp_thin_linear_timeouts,
 714                 .maxlen         = sizeof(int),
 715                 .mode           = 0644,
 716                 .proc_handler   = proc_dointvec
 717         },
 718         {
 719                 .procname       = "tcp_thin_dupack",
 720                 .data           = &sysctl_tcp_thin_dupack,
 721                 .maxlen         = sizeof(int),
 722                 .mode           = 0644,
 723                 .proc_handler   = proc_dointvec
 724         },
 725         {
 726                 .procname       = "tcp_early_retrans",
 727                 .data           = &sysctl_tcp_early_retrans,
 728                 .maxlen         = sizeof(int),
 729                 .mode           = 0644,
 730                 .proc_handler   = proc_dointvec_minmax,
 731                 .extra1         = &zero,
 732                 .extra2         = &four,
 733         },
 734         {
 735                 .procname       = "tcp_min_tso_segs",
 736                 .data           = &sysctl_tcp_min_tso_segs,
 737                 .maxlen         = sizeof(int),
 738                 .mode           = 0644,
 739                 .proc_handler   = proc_dointvec_minmax,
 740                 .extra1         = &zero,
 741                 .extra2         = &gso_max_segs,
 742         },
 743         {
 744                 .procname       = "udp_mem",
 745                 .data           = &sysctl_udp_mem,
 746                 .maxlen         = sizeof(sysctl_udp_mem),
 747                 .mode           = 0644,
 748                 .proc_handler   = proc_doulongvec_minmax,
 749         },
 750         {
 751                 .procname       = "udp_rmem_min",
 752                 .data           = &sysctl_udp_rmem_min,
 753                 .maxlen         = sizeof(sysctl_udp_rmem_min),
 754                 .mode           = 0644,
 755                 .proc_handler   = proc_dointvec_minmax,
 756                 .extra1         = &one
 757         },
 758         {
 759                 .procname       = "udp_wmem_min",
 760                 .data           = &sysctl_udp_wmem_min,
 761                 .maxlen         = sizeof(sysctl_udp_wmem_min),
 762                 .mode           = 0644,
 763                 .proc_handler   = proc_dointvec_minmax,
 764                 .extra1         = &one
 765         },
 766         { }
 767 };
 768
 769 static struct ctl_table ipv4_net_table[] = {
 770         {
 771                 .procname       = "icmp_echo_ignore_all",
 772                 .data           = &init_net.ipv4.sysctl_icmp_echo_ignore_all,
 773                 .maxlen         = sizeof(int),
 774                 .mode           = 0644,
 775                 .proc_handler   = proc_dointvec
 776         },
 777         {
 778                 .procname       = "icmp_echo_ignore_broadcasts",
 779                 .data           = &init_net.ipv4.sysctl_icmp_echo_ignore_broadcasts,
 780                 .maxlen         = sizeof(int),
 781                 .mode           = 0644,
 782                 .proc_handler   = proc_dointvec
 783         },
 784         {
 785                 .procname       = "icmp_ignore_bogus_error_responses",
 786                 .data           = &init_net.ipv4.sysctl_icmp_ignore_bogus_error_responses,
 787                 .maxlen         = sizeof(int),
 788                 .mode           = 0644,
 789                 .proc_handler   = proc_dointvec
 790         },
 791         {
 792                 .procname       = "icmp_errors_use_inbound_ifaddr",
 793                 .data           = &init_net.ipv4.sysctl_icmp_errors_use_inbound_ifaddr,
 794                 .maxlen         = sizeof(int),
 795                 .mode           = 0644,
 796                 .proc_handler   = proc_dointvec
 797         },
 798         {
 799                 .procname       = "icmp_ratelimit",
 800                 .data           = &init_net.ipv4.sysctl_icmp_ratelimit,
 801                 .maxlen         = sizeof(int),
 802                 .mode           = 0644,
 803                 .proc_handler   = proc_dointvec_ms_jiffies,
 804         },
 805         {
 806                 .procname       = "icmp_ratemask",
 807                 .data           = &init_net.ipv4.sysctl_icmp_ratemask,
 808                 .maxlen         = sizeof(int),
 809                 .mode           = 0644,
 810                 .proc_handler   = proc_dointvec
 811         },
 812         {
 813                 .procname       = "ping_group_range",
 814                 .data           = &init_net.ipv4.sysctl_ping_group_range,
 815                 .maxlen         = sizeof(gid_t)*2,
 816                 .mode           = 0644,
 817                 .proc_handler   = ipv4_ping_group_range,
 818         },
 819         {
 820                 .procname       = "tcp_ecn",
 821                 .data           = &init_net.ipv4.sysctl_tcp_ecn,
 822                 .maxlen         = sizeof(int),
 823                 .mode           = 0644,
 824                 .proc_handler   = proc_dointvec
 825         },
 826         {
 827                 .procname       = "ip_local_port_range",
 828                 .maxlen         = sizeof(init_net.ipv4.sysctl_local_ports.range),
 829                 .data           = &init_net.ipv4.sysctl_local_ports.range,
 830                 .mode           = 0644,
 831                 .proc_handler   = ipv4_local_port_range,
 832         },
 833         {
 834                 .procname       = "tcp_mem",
 835                 .maxlen         = sizeof(init_net.ipv4.sysctl_tcp_mem),
 836                 .mode           = 0644,
 837                 .proc_handler   = ipv4_tcp_mem,
 838         },
 839         { }
 840 };
 841
 842 static __net_init int ipv4_sysctl_init_net(struct net *net)
 843 {
 844         struct ctl_table *table;
 845
 846         table = ipv4_net_table;
 847         if (!net_eq(net, &init_net)) {
 848                 table = kmemdup(table, sizeof(ipv4_net_table), GFP_KERNEL);
 849                 if (table == NULL)
 850                         goto err_alloc;
 851
 852                 table[0].data =
 853                         &net->ipv4.sysctl_icmp_echo_ignore_all;
 854                 table[1].data =
 855                         &net->ipv4.sysctl_icmp_echo_ignore_broadcasts;
 856                 table[2].data =
 857                         &net->ipv4.sysctl_icmp_ignore_bogus_error_responses;
 858                 table[3].data =
 859                         &net->ipv4.sysctl_icmp_errors_use_inbound_ifaddr;
 860                 table[4].data =
 861                         &net->ipv4.sysctl_icmp_ratelimit;
 862                 table[5].data =
 863                         &net->ipv4.sysctl_icmp_ratemask;
 864                 table[6].data =
 865                         &net->ipv4.sysctl_ping_group_range;
 866                 table[7].data =
 867                         &net->ipv4.sysctl_tcp_ecn;
 868                 table[8].data =
 869                         &net->ipv4.sysctl_local_ports.range;
 870
 871                 /* Don't export sysctls to unprivileged users */
 872                 if (net->user_ns != &init_user_ns)
 873                         table[0].procname = NULL;
 874         }
 875
 876         /*
 877          * Sane defaults - nobody may create ping sockets.
 878          * Boot scripts should set this to distro-specific group.
 879          */
 880         net->ipv4.sysctl_ping_group_range[0] = make_kgid(&init_user_ns, 1);
 881         net->ipv4.sysctl_ping_group_range[1] = make_kgid(&init_user_ns, 0);
 882
 883         /*
 884          * Set defaults for local port range
 885          */
 886         seqlock_init(&net->ipv4.sysctl_local_ports.lock);
 887         net->ipv4.sysctl_local_ports.range[0] =  32768;
 888         net->ipv4.sysctl_local_ports.range[1] =  61000;
 889
 890         tcp_init_mem(net);
 891
 892         net->ipv4.ipv4_hdr = register_net_sysctl(net, "net/ipv4", table);
 893         if (net->ipv4.ipv4_hdr == NULL)
 894                 goto err_reg;
 895
 896         return 0;
 897
 898 err_reg:
 899         if (!net_eq(net, &init_net))
 900                 kfree(table);
 901 err_alloc:
 902         return -ENOMEM;
 903 }
 904
 905 static __net_exit void ipv4_sysctl_exit_net(struct net *net)
 906 {
 907         struct ctl_table *table;
 908
 909         table = net->ipv4.ipv4_hdr->ctl_table_arg;
 910         unregister_net_sysctl_table(net->ipv4.ipv4_hdr);
 911         kfree(table);
 912 }
 913
 914 static __net_initdata struct pernet_operations ipv4_sysctl_ops = {
 915         .init = ipv4_sysctl_init_net,
 916         .exit = ipv4_sysctl_exit_net,
 917 };
 918
 919 static __init int sysctl_ipv4_init(void)
 920 {
 921         struct ctl_table_header *hdr;
 922         struct ctl_table *i;
 923
 924         for (i = ipv4_table; i->procname; i++) {
 925                 if (strcmp(i->procname, "ip_local_reserved_ports") == 0) {
 926                         i->data = sysctl_local_reserved_ports;
 927                         break;
 928                 }
 929         }
 930         if (!i->procname)
 931                 return -EINVAL;
 932
 933         hdr = register_net_sysctl(&init_net, "net/ipv4", ipv4_table);
 934         if (hdr == NULL)
 935                 return -ENOMEM;
 936
 937         if (register_pernet_subsys(&ipv4_sysctl_ops)) {
 938                 unregister_net_sysctl_table(hdr);
 939                 return -ENOMEM;
 940         }
 941
 942         return 0;
 943 }
 944
 945 __initcall(sysctl_ipv4_init);